The idea that a field can be reformed by making it worse until it suddenly faces a reckoning and emerges much better is ... I don't know where people keep getting the idea that this might work. It has never worked in any field ever in history.
The thing that can happen: fields gradually split into rigorous and non-rigorous camps. Like with evidence-based medicine, or chemistry/alchemy. Depending on the field, either might prevail in the market. Medical research and bridge-building are mostly rigorous, programming is mostly non-rigorous.
AI/ML has a range of rigor levels, from fairly good to total crap. I think people in the field have a reasonable idea which is which. It's frustrating to people outside the field that think they can just take the technique with highest reported performance numbers and expect good results.
To me, this did not read as a serious proposal to actually commit more fraud. Rather, it was a "modest proposal" in the tradition of Swift, in which the actual call to action was for the field to be more critical of papers, especially to be on the lookout for all fraud, both the obvious kind and the more subtle variants, the latter of which also do great damage to the field.
The tip: humor like this is fun and appreciated by people who run in the same circles as the author, but an essay like this will be read by a diverse cross-section of people. Some won't have the cultural references, some won't have English as a first language, etc. Almost always when I've snuck jokes into my writing, I've found it causes confusion.
So I might have written this a slightly different way, something along the lines of: the community is structurally more equipped to deal with blatant than subtle fraud. Ironically, now that we're seeing more egregious examples of fraud, there's a better chance that things will get better; we would have tolerated the subtle types for a long time, as lots of people benefit from the status quo.
If the author were actually legitimately calling for more fraud, then I apologize for misunderstanding.
I assumed it was tongue in cheek, but the end of the article really convinced me that the author is serious about his ‘let it get worse so it can get better proposal’ stance.
“Widespread fraud would force us to re-strengthen our community’s academic norms, transforming the way we do research, and improving our collective ability to progress humanity’s knowledge.
“So this is a call to action: please commit more academic fraud.
[...]
“Let’s make explicit academic fraud commonplace enough to cast doubt into the minds of every scientist reading an AI paper. Overall, science will benefit.”
Given the number of people in academics and politics and software and just about any field who argue from the ‘let’s burn it down and rebuild it’ position, even if it is satire, it’s basically impossible to tell.
Author here -- it's intended as satire, don't worry, haha. I know that nobody would ever actually do this so I decided to just lean into it. If I had any real power by which to fix things, I would have a much more nuanced take.
Well, I wonder if this could force people to start publishing their code so others can replicate the results.
This would help greatly with the kinds of issues that are being discussed:
- It would be much harder to hide makeup results and cherry-picked seeds
- Useless research would not be as easy to hide
- The real-world impact of the research would be more valued.
- Researchers would be forced to recheck their code for it to be presentable, which could be a net benefit in terms of finding mistakes
- and so on..
I do imagine this could harm the number of replications of the research.
However, I think there can be a net positive effect from such a policy. What do you think?
These discussions remind me of how there's a years long delay before these ideas first reach boards like HN and then popular magazines and the taxi driver.
Yes, code releasing is a big talking point and most major groups release code. It doesn't solve the subtler problems,like claiming 1% improvements as a solid contribution, unequal hyperparameter tuning, cherry picking datasets, overclaiming the novelty, salami publishing, lacking ablations, not comparing against better methods etc etc. Releasing code is a step but not a panacea. At least it partially protects against fully made up numbers which is not nothing, but it's just scratching the surface. In many cases the code is not so easy to use and the original authors may not be willing to help you with reproducing it, ignoring emails and GitHub issues. If you can't reproduce it, they can always claim something. In one case I was told they rewrote the codebase in a new framework since they generated the results in the paper and there may be small discrepancies due to that. Apparently that was enough change that they were no longer the best in the benchmark comparison. But nobody cares at this point, the paper was published, they are on to their next project. People also have the attitude that it doesn't matter anymore, the field has moved on already, a retraction doesn't change much just hurts the author and the group. The vast majority of papers isn't used anywhere actually in practice, the field is moving so fast that any method can only remain on top for a couple of months.
Also, I don't think AI really is worse at this. It's rather that we are more open about it because we have less to lose and professors don't have an iron grip, due to the existence of industry. In some of academia your whole future depends on whose ass you kiss, it's better in CS. It's a bit like sexism accusations directed at nerds, as if they particularly bad.
The molochian cancer is everywhere. The less you see it may just indicate its more effectively hidden.
I'm active in the security and privacy field, and some conferences have started "artifact reviews" [1][2], where authors can optionally submit the data and code that go along with their work.
It's definitely an improvement, but the devil is in the details. Should these be mandatory? If they're mandatory, then what quality standards should artifacts meet? Who's ensuring those standards are followed? How should proprietary code and sensitive human-subjects research data be handled? There's also the question of what code should be made public: Is it just the analysis that produced the metrics on the paper (e.g., R scripts, Jupyter notebooks, etc.)? Or should it also include the data collection and pre-processing? How about the code that didn't produce immediately publishable results, but could be useful for future work?
> where authors can optionally submit the data and code that go along with their work.
So cool. I wish there were more steps in this direction.
> Should these be mandatory?
As an outsider, I would say no. However that should have a detrimental effect on the ability to get published if you are from an academic institution. I can totally see a gray area when it comes to non-academic publishers.
> There's also the question of what code should be made public
Shouldn't it be as much as possible? The way I see it is that the easier to replicate a paper, the more people can focus on its true utility. That is, by abstracting ourselves away over the details of the paper, what does this really contribute? From my point of view, publishing as much as possible maximizes the real-world impact of the research.
Publishing the code may not be enough. Part of the problem is that replication of a study isn’t usually rewarded in the academic sphere because of the obsession with novelty within publications. Most researchers want to spend their finite time on something they find interesting and also may help their career. One way may be to create a movement to foster good replication studies that can be published and rewarded.
Novelty is in the job description though. I think the problem, especially in AI/ML is the obsession with "State of the Art" algorithms. This can be gamed by making claims with numerical "experiments". If an idea had to stand on its own, and be interesting even if it may not make a better pattern-recognition product, most of them probably wouldn't make it. Indeed that's essentially the requirements presented by reviewers and editors: "prove we should read about this otherwise-uninteresting method by giving numerical test results where it wins".
I’m not stating novelty isn’t be part of the aim of research, I just think we need to be aware of the perverse incentives it creates when it becomes the sole underlying criteria by which merit is judged. I think there’s room for expanding the definition of what is good research to include replication of prior art because the field in general will be better off for it. I imagine novel research would still be more prized but it wouldn’t relegate replication to being considered a waste of time.
> Well, I wonder if this could force people to start publishing their code so others can replicate the results. This would help greatly with the kinds of issues that are being discussed:
The real solution to the problem is to set a rule that says if you use a lab's published code base AT ALL for new work, you must cite the people who produced it along with the PI of that lab. No exceptions; acknowledgements don't count. Labs (especially those outside of CS such as life sciences labs) view code as IP and they don't want to release their code bases for two reasons: First, they don't think they will be properly cited when people use it (valid concern). Second, they think that they will lose their monopoly on future publications using the code base (less valid concern). Mandatory citation would solve both of those problems. Until a rule like this is in place, most labs will fight you tooth and nail on releasing code.
AI seems to me as quite experimental. Designing better benchmarks (that is less cheatable) will help. Publishing code will go against IP policies, that should be handled at the political level : do you research for profit or for knowledge ?
But yeah, code adds transparency and a whole lot of cheats... The funny thing is that any well organized coder will automate report production, so it's not like it is hard to do...
In the steel-manned implementation, I would like the plots and tables to be auto-re-generated on publication, from submitted code. That would ensure that the results can have no “hidden state” and the paper cannot claim anything more than what their publicly shared implementation guarantees.
This is par for the course in theoretical work; we should strive to apply the same standards of rigor to experimental work.
They require the full means for replication to be open access, and reviewers take this very seriously! They will run your code, open github issues if they detect bugs etc.
In politics and other areas, the concept is known as "accelerationism."
It tends to be associated with political extremism... "bring on the revolution/war/etc." As an angry or disillusioned response, I think it's a close relative of nihilism. Cover for being destructive.
That's not to say "worse before better" isn't a thing... it's just not a thing we can do usually.
Is AI as a field not destined for destruction? There were good stuff, but it's been out for a while and now it's just mountains upon mountains of crap.
In any case, I wasn't arguing that accelerationism is incorrect... just that it tends to be associated with angry feelings, political, professional or otherwise. Also, while I can think of many things that got worse and then better... I can't think of any cases where it happened deliberately.
Since animal psychology is mostly design to keep animals safe (not, eg., to flourish), we can persist in severely bad circumstances merely because they arent unsafe.
What "getting to rock bottom" does is drive a person to acute unsafety so they are able to overcome the inherent risk of change.
This "personal accelerationism" does work, indeed, it may be the only thing that works.
I do agree that this likely does not generalise to institutions.
I loved your first paragraph but I choked your second.
All modern fields of study use quantitative measurement so splits in practice now would have to do with different ideas of rigor rather than rigor-nonrigor.
The problem of AI/ML isn't the blatant cheating but the way that the goal is often getting n% higher than soto on X benchmark. Just chasing benchmarks makes any connection to broad dubious imo. It might or might-not give you something practically useful but definitely give you something career-wise useful.
But just as much, when the field is just a giant race where no cares about any broader understanding, cynicism seems like a natural result. The ideal of academia, for all it's failings, is to give people some amount of space to speculate and explore wider vistas.
It would be good if X number of people had the space to explore a variety of visions of "AI" other than the dominant one. But despite the vast number of people being sucked into the field, my guess is this is getting harder, not easier. And, of course, the mere appearance of "rigor", of quantitative measurement, is not helping things, again in my opinion as someone of no authority at all in field.
This struck me mostly as tongue in cheek, and the the author's main point is the blatant fraud is just the tip of the iceberg, the subtle fraud is happening under the surface and affecting almost every field.
It's interesting that programming used to be rigorous. Whole programs were written out and verified before anyone would invest the computer time. IBM had a range of flowchart templates for it! Then the growing availability of computing resources flipped the cost-benefit calculation in favor of getting it done at any cost, and quality control processes haven't been able to keep up.
It's funny you mention that. I'm often surprised how little literature there is out there on Quality in general, and how often I have to depart from Software in general and head into manufacturing/metrology to get any new/useful insight on Quality Control methodologies for software. There is, however, more to the story than Moore's Law at work.
#1 In the early days, you had very little between you and the executing machine. Your programs were more ways of doing things to ensure a particular machine would get you a reasonable answer, and it was far easier to communicate the totality of the stack, software+hardware combined.
#2 Compare the picture of the 60's-90's programmer to the programmer of today. You had to know the hardware, and do cartwheels to decompose your problem to be solvable within the constraints of the machine which you had to run it on. You were a professional optimization problem solver+a studious cross-referencer (libraries and dependency management were not as mature as they are now, nor were there as many Virtual Machine constructs to foster write once, run anywhere.
#3 Libraries were a case of build it yourself, or you ported something else by figuring out the toolchain + operating in a much smaller network of professionals to reach out to for guidance.
#4 There weren't many if any concessions to programmer "comfort" (IDE's), less static analysis (as far as I'm aware), and good luck finding documentation without paying for it.
Now:
>Many programmers are blissfully unaware of cache coherency, memory hierarchy, or the quirks of the hardware/filesystems they run on. The hardware is "the compiler writer's job" or "those driver writer's problem" (smh?), the filesystem is Someone Else's Problem...
>More and more, the solution looks more and more like "throw more hardware" at the problem to create more abstraction, which requires more intermediate steps, which takes more compute...
If for no other reason, software is disgustingly hard to Quality Control for because nowadays, it's more about having access to development talent to make the core system architecture, no matter how inefficient or bloated, maintain it, and extend it rather than boiling things down to least computtational overhead.
When you've got a thing built on a constantly shifting Tower of Babel, where your artisan knows less than a 10th of what their total set of dependencies are doing, and oftentimes are selected for their willingness to sit down, crank out the requirements, and not balk; it truly is a miracle when Quality software actually happens. Test coverage alone isn't it. Nor is refactorability, or readability. You have to have the right Software for the right people, at the right time, for the right costs, to bring about the right constellation of jobs done, to create value. None of that value, in modern thinking, should be intrinsically tied to the people making it. In fact, I have a theory that the market is trying to get away from that by favoring designs where everything you need is in the automation code itself, as it obviates the concept of the "heroic wizard coder" as we know it. Part of this Quality too, is you have to actually care about the people you are implementing for. Any programmer can create a program that works (painfully). It takes a sharp one to make something intuitive, quick to learn, and a pleasure to work with without being godawful slow.
I'd say Quality software is not something you set out to build (at least the way the Market driven ecosystem we have today works), but a happy accident when you combine all the right factors to get the job done you set out to do and keep it that way. But the roadmap to getting there is the hard part.
If anyone ever wrote a dynamic code generator that worked like a database optimizer (imagine Select implementation FROM C Where INPUTS(x1,..,xn) AND OUTPUTS (y1,...,yn) WITH CONSTRAINTS (z1,...,zn)), we'd all be out of a job, possibly all the healthier for it.
Who am I kidding, we'd all just be query writers and optimization engine babysitters for it.
If medical research is rigorous then rigor isn’t worth much. It’s one of the fields most affected by the replication crisis, along with social science. https://en.wikipedia.org/wiki/Replication_crisis
It's worth noting: In most of these cases, even when caught, the guilty parties and the bad papers remain unnamed. There are known-to-the-editors fraudulent papers in major journals still being cited.
The consequences of committing academic fraud are minor, and the consequences of not doing so generally mean no academic jobs or tenure.
There isn't a 99.99% innocent claim here (as at the end of the article). This stuff is widespread. This is much more accurate:
"because everybody is complicit in this subtle fraud, nobody is willing to acknowledge its existence"
Littman is at Brown, where the majority of the CS department engages in this !@#$%. I'm not trying to single out Brown. Parts of MIT, Stanford, etc. are even worse.
With laws, there are full time police officers, investigators, and prosecutors working to catch people. And even then, most crimes go unsolved and unpunished.
Even without the complicity, stopping fraud and bad behavior is a difficult problem. Part time reviewers and reputation seem woefully insufficient counters.
I don't think people understand how hard this would be, or how much it would impede science. To give an analogy to coding: imagine a company with thousands of programmers, all using different languages and writing different types of code (mobile apps, back-end server apps, mainframe code, device firmware). Now assign one or two poorly-paid employees whose job is to review all of their code and find issues with it.
Can you imagine how annoying this would be? How much time you would spend simply explaining to these people what your code does, so they could understand the basics of what your program is even doing? How ineffective they would be at detecting actual fraud, and how quickly their processes would turn into yet another annoying layer of bureaucracy and checklist compliance?
There are a lot of things Universities could do to assist scientists in producing better output. This is not one of them.
after having spent a third of my life in academia i don't see this happening: everyone want as many as possible articles, if possible in the most high impact journals. detecting fraudulent or irrelevant pieces is against that goal.
Yes, I'm very skeptical of Buckman's claim that more fraud will help. Seems similar to a claim around 2015-2016 that voting for Donald Trump will inspire America to clean up its act politically. Institutions aren't biological systems that operate according to mysterious hormetic processes. Institutions are created by humans and thrive or decay based on whether they effectively channel human effort. Be the change you want to see in the world. And if you wreck something, and others apply blood sweat and tears to recreate it in a way that's better than the original, you don't get to take any credit as the wrecker.
I think the overwhelming problem is the amount of trust people have an academia, and especially, in elite academia. A Harvard research paper is trusted. The harm of people trusting fake science is high. I see it every day in my field.
1) If people realize it's more like a Facebook post, that will be better. People will be able to push back.
2) Alternatively, if we clean up this mess, that will be even better. Academia ought to be trustworthy.
I don't see a path to #2 without a lot of dishonest people with tenure being laid off. At elite institutions, most people hired in the past decade or two cheated at least a little bit. I don't see a path to get there without a high-profile scandal.
I manage an applied research group in the machine learning space, particularly AutoML and synthetic data generation. That means is we read the latest literature and create applications. Sometimes we build new research, most of the time we patent something new. But the goal is to get research to production as fast and as impactful as possible.
The first thing I do when sorting research (papers, blog posts, presentations, etc) is sort the research by “B.S.”, “maybe B.S.”, “probably real”.
If you look at research papers in computer science, half are what have the following problem: “running a simulation over and over until the authors get the results They like”.
I can’t tell you how many papers with thousands of citations are just blatantly wrong. They don’t run corrections based on the number of simulations they run, they don’t take into account other variables, etc
This isn’t limited to CS either. Biology, economics, environmental sciences, etc all suffer the same fate.
The worst part about it, is it’s PEER REVIEWED. Meaning, others agree this is the way to do things, which is why I don’t trust academia almost at all.
> They don’t run corrections based on the number of simulations they run, they don’t take into account other variables, etc
I think this looks like a bigger problem specifically because you are in AutoML.
Suppose you are training a GAN. There's notoriously a certain amount of luck involved in traditional GAN training, because you need the adversary and the generator to balance each other just right. So people try many times until they succeed. Probably they were not even recording each attempt, so they do not report how many times they had to run before getting good results.
From an AutoML point of view, this is BS work - the training procedure cannot be automated, and (apart from using the actual seeds) the work cannot be reproduced.
But from the point of view of everyone else, maybe it is fine. They get a generator model at the end, it works, other people can run it.
>But from the point of view of everyone else, maybe it is fine
I think from a practical perspective, it is fine. You want results and you have a black box algorithm that produces them, fine.
From an academic perspective, AI research is a mess. The reason you try something is not from a logical theory, but due from a "hunch" or replicatinga similar algorithm applied in a parallel area. If it does not work, you change some parameters and run it more times. Still not working so maybe you extend the network to include some more inputs and hope for better results.
I did my thesis in machine learning and was very disappointed with the state of the field.
I don't think there's necessarily a problem with trying things on a hunch, some of the best results in science have been due to a hunch or even an accident. The problem comes from trying a dozen hunches and only writing up one, or like you say completely cherry picking hyperparameters.
I despised how data sets and code were considered a gold-mine and researchers would often refuse to release them lest they give up a serious paper publishing advantage for their group. More often than not, i found code full of serious errors and completely lacking basic test cases and sanity. Sometimes the data sets were much larger than the subset chosen for publication.
Say you run 5 simulations and when comparing to the standard approach as a percent the samples are [-1%, -1%, +2%, +2%, +%6]. Your new method is not 6% better the standard it's at best +1.6% better.
If you keep track of these categories and have amassed a large enough dataset, it sounds like a fun experiment to see if a NN can learn some feature of the papers text to predict if a paper is BS. Some combinations of citations may even be a telling feature.
I watched a lecture by an MIT database professor who recommends that the top 10 schools only allow up to 3 papers on CVs for junior faculty applications and 10 for tenure applications.
He thinks this will be enough to start a culture shift towards quality over quantity, which could go some way towards addressing the fraud and collusion ring issues by removing the incentive for these behaviors.
If the universities wanted the problem fixed it would be fixed. I truly don't believe research universities are focused on producing value (other than to themselves).
Apparently it wasn't always like this, and it used to be common for junior faculty hires to have only 0 or 1 publications coming out of their PhD program.
Why did the culture deteriorate so much and become so myopically focused on weakly informative metrics like publication count?
When fields are small, they can rely on much higher fidelity signals for hiring (having significant interactions with the individuals at conferences and workshops, for instance). I did my PhD in a sub field of nuclear physics, and even though I only published one paper in my PhD, I got a job offer at an Ivy League university because I had done extensive work with their group in collaboration, and the group leaders liked me.
However, now that many research fields are so massive that it is impossible to personally know the majority of individuals, institutions need other ways of judging individuals. The number of papers published is a weak signal, but it’s better than nothing. Now that it’s being so heavily gamed by so many individuals, that signal strength is decreasing even more.
There’s also a second strong corrupting factor that many, if not most, of these individuals do not want to become professors, they want to get a high paying job in industry, which means their short term output is far more important to them than their long term reputation in the field.
I honestly don’t know what can be done to fix this that wouldn’t have negative side effects. But perhaps the side effects would be better than the situation we are in now.
1) due to population effects, academic positions are much more competitive now than they were in say 1970; if you figure that the top 50 research universities are not generally expanding the number of professors, and that new professors generally also come from those top 50 research universities, then on average a top-50-research-university professor will generate one new such professor in a career, despite having 10-100x as many graduate students (this was different in the 70s when the university system was rapidly expanding).
2) the increasing desire for fairness in hiring and promotion (by itself, a good thing) means that you need to be able to resolve hiring and promotion disputes with something both objective and external to the university (in the same way some undergraduate institutions put more admissions weight on external and objective metrics like standardized tests compared to more easily game-able internal metrics like high school class grades)
Because government grants demand it. Private patrons can trust their own judgment when deciding who to fund, but when it's taxpayer money being handed out people are understandably going to demand objective metrics to guard against corruption. In academia the objective metrics of choice are publication and citation count, so here we are.
Was it Mike Stonebraker's talk "My 10 Fears about the Future of the DBMS Field"? (2018 Donald B. Gillies Memorial Lecture, presented at the University of Illinois): https://www.youtube.com/watch?v=DJFKl_5JTnA
- He thinks it's not an overly difficult problem to solve. It just needs some grassroots push from a handful of people in each of the top CS departments. It's just that not many are really trying to push for it. (I assume there would be people staked in the current system, though, who want to maintain the status quo.)
- Nowadays he sees people trying to split research up into Least Publishable Units in order to maximize the number of papers.
- Coming up with Postgres would've been impossible in today's climate since the time it took would entail an insufficient number of papers to get hired or promoted.
- He thinks part of the reason that the culture has deteriorated is that Western universities have adopted cultural norms of East Asian universities, for whatever reasons.
I'm really happy to see this article here. I quit my PhD where I was working on AI partly because of the "mundane day-to-day fraud" that the author and I observed in the field. Once you've read enough papers, you can easily see a kind of recipe to an academic paper. Present a new method, compare against a baseline, discuss the differences in the results found. To me, some researchers just follow this recipe and do not think about the steps critically. You see figures that plot two categories of points that are purportedly following different distributions but the error bars (or even more of the distributions) overlap. But a p-value says they are significant. So that's good enough to publish! I felt a lot of anxiety about sharing my research in this environment and felt pressure to make my work seem like it was the best solution for every problem when in reality there are more nuances.
> I felt a lot of anxiety about sharing my research in this environment and felt pressure to make my work seem like it was the best solution for every problem when in reality there are more nuances.
I am curious what kind of pressure you are referring to. It's your work and your decision how to present it to the world.
As far as other people's research is concerned, academics can express their opinions by participating in peer review and expressing their opinions vocally at program-committee meetings (for instance, proposing artifact evaluation as a part of the review process).
I had the same experience as GP. I dropped out of a PhD program after witnessing widespread low-level fraud.
I published a single paper with my advisor. I asked questions that a peer reviewer would have asked, and my concerns were basically ignored. It was more important to publish than be accurate and precise with wording (partly due to length limits imposed by the publication). Pushing back harder likely would have had a deleterious effect on my progress in the program. It would have been career suicide.
Worse, the topic was nothing more than rehashed results from a paper he published years earlier. There was really nothing even worth publishing. This was but one example that showed me that academia is endemic with fraud.
Eventually I killed that career path, because I could never participate in such a fundamentally corrupt system. There will be no reform here without Revolution.
It is true that as the author of a research paper it is my decision how to present it. However, if you're too far off the mark, you are just going to be rejected by peer review or they will ask for revisions. The fact is as a PhD student, you are trying to join the research community. You aren't in a position to change that community. You have to tailor your work and statements to fit the mold. Students who are outspoken voicing these concerns, especially if they rise to the level of abuse or research misconduct, must tread carefully. A case I was made aware of in the study of research ethics is of Anil Potti [1], [2]. From [2], the statement from the whistleblower, "In raising these concerns, I have nothing to gain and much to lose" is apt.
For more mild cases, like those mentioned in the article, the stakes are lower but there is also more plausible deniability. If Duke University tried to bury even the blatant abuse, you can imagine how it is also hard to confront the article's so-called "mundane, day-to-day fraud".
Best part of the article is the footnote, where the author characterizes his own papers:
> This paper is bullshit, this paper (a NeurIPS oral) is bullshit, this paper is complete bullshit, this paper is mostly good science but also has a sprinkling of bullshit. Apologies to my co-authors.
In undergrad, my professor explained to me in very candid terms (from the position of being a generous mentor) that the optimal path to success in the Arts and Social Sciences is:
1. Find a niche only a couple of people operate in.
2. Make friends with them at all costs and work in their area.
3. Review each others work, amassing enormous citations in highly respected (albeit sometimes niche) journals.
I saw something similar in an academic's blogged advice ten or fifteen ago: "At graduate level, you should cultivate [peers and a concentration] such that your intellectual correspondence is publishable."
I took that as plausible enough at the time to mention the advice once to a graduate student in philosophy. I did not take it at the time as immediately translatable into a [simply translated?] phrase I have heard since: "friendship corruption."
There are some video lectures from a writing consultant employed at UChicago. He accepts that concentrations within fields are to some extent self-defined in the academic game -- in only slightly more generous terms. IIRC his strategic advice for those failing to publish is: learn to frame your abstracts as respecting but advancing the discussion. Perhaps in small enough niches ingratiate yourself would need to be mentioned? Or show that you will play by being selectively generous with citations??
Citations. The academic version of SEO and currency more valuable than money? (Recall where page rank came from.) One kind of power behind institution-sanctioned monsters?
> "At graduate level, you should cultivate [peers and a concentration] such that your intellectual correspondence is publishable." ... I did not take it at the time as immediately translatable into a [simply translated?] phrase I have heard since: "friendship corruption."
This seems misguided, and I certainly hope that this 'friendship corruption' concept never catches on. There are great papers that started as letters and were later completed by the sender, recipient or both. No one should feel ashamed about that, and no one should feel ashamed of developing friendships with their colleagues.
Well said. My comment was phrased in response to its parent comment looking askance at "amassing enormous citations." There is love of truth and not all is corruption.
> When a measure becomes a target, it ceases to be a good measure. - Goodhart's Law
> The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor. - Campbell's Law
This is just another proof that you can't trust metrics. Build up a network of people that you can hold personally responsible for, and if any of them ever recommends low quality work then let them go. Let that network know whenever you discover a network of individuals who put out low effort work so that you and your network can disregard the entire network.
Bottom up, not top down. Do not rely on citation count. Do not rely on views. Do not rely on upvotes or reviews from people you do not know. Do not rely on if they went to a prestigious university. Do not rely on the ability to pass a standardized test.
The problem here is that there is no bright-line distinction between "friendship corruption" and "doing high-quality collaborative research in a small field where colleagues are cordial."
I'm not saying that one can't bleed into the other -- sometimes they do. I'm saying that any approach you take that disincentivizes the bad behavior is also likely to harm good scientific collaborations as well. It's one of the downsides of peer review.
I wonder if part of the current problem is not that the fields are too big.
In a small field where people know each other, reputation becomes more important, and you can't just misrepresent your results, since everyone will read it and know you did it.
Meanwhile, in a big field, it is easier to hide in the crowd. Write some papers that get accepted, but not read, and have some friends cite them.
I think we need to rethink the core concept of Universities having publish or perish as a part of being a Professor. The reason why fraud exists is because of the incentive structure and due to the nature of research you have intelligent people that will game the system. So, we have to figure out how to change the incentive structure so that professors don't attempt to perform Academic Fraud or we h ave to figure out if fraud does exist after the fact.
I have tried to reproduce ML / Deep learning research. I have a few heuristics that don't really tell me if fraud exists but either they just can't release the data due to special circumstances that I can understand or it isn't kosher. The biggest one is that you have to register or ask them for their test data. Another one is that the paper has been published for more than two years and there are no citations from other work.
Keeping test data private is good practice actually. It keeps from hyperparameter optimization on the test set. Yes theoretically the dataset authors might still do it, but it's still better for the field as a whole to keep some test data secret and only use it for evaluation.
Registering before getting data is common, so they can make you agree to some terms. It's not suspicious in itself.
There are no hard and fast rules of thumb. You can look at the group, the people involved, the country, the writing style, the quality and thoroughness of ablation, is there code available etc. But nothing is 100%.
> So, we have to figure out how to change the incentive structure so that professors don't attempt to perform Academic Fraud or we h ave to figure out if fraud does exist after the fact.
One possible way: increase funding to the arts and sciences. The total NSF budget was ~$8B, or 0.03% of US GDP. We could double that and no one would even notice. Increased funding means more money to go around, means higher acceptance rates for grants, means less incentive to write fraudulent papers.
The best part is, you have the power to actually change this. Write your congressperson, vote for a congressperson who pledges to do this, run for congress yourself, or organize other people to do this with you.
Increasing the budget doesn’t solve the problem. You would need to create a set of well funded auditors that have domain experience but the problem with that is no one would want the job.
I ran into these issues in an adjacent field a generation ago. Many candidly spoke to me about having to be a ‘team player’.
This is all due to the criteria for survival and graduation. Let’s be honest, not every idea will work and this does not reflect badly on the researcher who came up with it. But the only way to graduate and put food on the table is to keep getting funding through novel papers that have great results.
Well, does it reflect badly on a startup founder if their idea doesn't work out? But since things are subtler in science, it's possible to sell stuff and pollute the literature even if it didn't really work out. To do otherwise is altruistic but ultimately outcompeted by those with less qualms about it.
> Well, does it reflect badly on a startup founder if their idea doesn't work out?
Maybe not in SV, but I have heard plenty of people tell startup founders to "get a real job" or refer to trying a startup as a "figuring it out phase."
When a startup some friends founded failed, they had to deal with people wondering if they chose the startup route as they couldn't get real jobs upon graduation.
Science involves a lot of risks but we don’t reward negative results, lessons learned from methodology problems, or failed reproductions except in rare studies. These issues come up all the time and many resort to tactics in the article to keep their funding / graduation plan alive.
Either way, even if the problem will be solved it will be the small guys who get screwed. The top dogs have built their careers on this bullshit, now they pull up the ladder and would want to make the playing field harder for anyone else.
Typical strategy also with startup companies that skirt the dark gray zone of the law when they are small, but then when they extract enough money, they suddenly demand regulation so the rest of the up and coming are smacked down.
That was my gut feeling about the ACM article that the author here refers to. The vague exposé in ACM is made by someone in the comfortable position of "Royce Family Professor of Teaching Excellence in Computer Science at Brown University" about people who aren't anywhere near that. It is like saying: please stop treading water so desperately, you are splashing those of us in the lifeboats.
I've met Jacob in person and we had a super interesting discussion about RL that sadly petered out during the covid lockdown. This blog post very much fits my impression of him and I give mad probs to shitting on your own work (particularly the thermometer encoding was a pet peeve of mine).
Wow, the amount of depravity in some academic circles is astonishing.
The article links [1] the case of a PhD student at the University of Florida who was forced to participate in such a publication collusion ring and was pressured to commit scientific fraud by fabricating results and submitting them to a conference [2], being threatened with physical harm should he decide to go public.
This student saw no way out and decided to kill himself.
Just a few days after the suicide, the department thought it would be appropriate for the student's own lab to have a "fun" excursion and to document it on Instagram [3].
How is it possible that professors with impeccable academic credentials get fired for jokes nearly instantly [1], yet this student's professor was allowed to carry on until he resigned two years later [2]?
Well... non-PC jokes are a liability risk for the institution at large. Dead students are not. We have an identical phenomenon in hospitals: there are huge campaigns against sexism and the like, while there are multiple young professionals
committing suicide from burn-out every year and noone does as much as bat an eyelash.
Contrary to victims of discrimination, dead people can't easily organize into coordinated legal action.
Said another way: Attention economy. There's no rhyme or reason for why any one particular outrage bubbles to the top and becomes today's cause célèbre.
I get the distinct impression that the Twitter sphere isn’t genuinely concerned about people of color, or else they would express concern over, say, inner city violence. Instead they work hard to brand any such concern as “far-right”.
Are Asians PoC? I thought they were not and that was literally the only reason the oft repeated phrase “black and other PoC” didn’t literally translate into “not white”.
Some woke people will argue that they are “white adjacent” when convenient (e.g., the Harvard admissions scandal) but also that they are people of color when convenient (e.g., the Atlanta spa killings).
My understanding is that the term "Latino" includes both people indigenous to Latin America and people whose ancestors were colonists from Spain. The latter group is generally the larger/more well-known one -- that's why "Latino", a Spanish word, is used to describe them.
They’ve recently started using the term “BIPOC” to clarify that Asians are excluded, and when they want to include Asians they’ll say “BIPOC and AAPI”.
No, by any good faith source that I have read, BIPOC does not at all exclude Asians. "The acronym BIPOC refers to black, indigenous, and other people of color and aims to emphasize the historic oppression of black and indigenous people."
Incidentally, one of the co-founders of the "BIPOC Project" is an Asian-American woman.
From the BIPOC project website: “We use the term BIPOC to highlight the unique relationship to whiteness that Indigenous and Black (African Americans) people have”
If it included Asians they would just keep using POC.
The "PC twitter sphere" as you describe them are only interested in one thing: doing what is easy and public for burnishing their own stature, and nothing else. No real problems will be solved by them, because they prefer having the public think they are solving them, rather than putting in the actual work and disciplined thinking to do it.
I would encourage you to think more about the social use of "jokes". Once we get past the level of knock-knocks, jokes very often have social meaning. Look at people like George Carlin, for example. His "7 dirty words" routine was deeply political. It was a full-on assault on American government censorship and the cultural elements that demanded it. And looking at history, he's won. Humor can be very powerful.
Jokes can also be used the other way, for social control. Growing up, I heard a lot of racist and sexist jokes, the practical effect of which was to demean: to create a place and put disfavored people in it. I'm old enough that nerds were a similar group, and I remember being the butt of a lot of jokes. When that happens, you're just supposed to take it; any objection to being demeaned is met with, "Why so sensitive! It's just a joke!"
So in the case you cite, the problem wasn't him telling a joke. There are whole books full of jokes for speakers he could have used. It was him invoking rank sexist stereotypes and suggesting the solution to his inability to manage his feelings was to push women out of the labs that they've been working for decades to get equal access to. And indeed, are still working on. At my alma mater, just this week a CS professor was just pushed out after dozens of women complained about sexual harassment in recent years. [1] It took dozens because early complaints were dismissed. And there are far more stories of professors like that then there are of ones being booted for "joking" misogyny.
Unpopular Opinion: people making these sexist jokes do actually think they are jokes. They aren't using them to push people down. Most people are nice people and don't realise they are being assholes and pushing people down.
On the other hand it's completely fair that people feel pushed down by them.
HOWEVER - this entire social justice movement is being used to outsource getting into conflict, and standing up for yourself. My guess is 80-90% of the time if you told someone who made a sexist joke that you are hurt by it, they would apologise (sincerely), and probably not do it again. But for that people actually need to get into a conflict situation, which is hard.
But it would make life a lot easier if we just sorted out these issues at the source, with two people, explaining what hurts and why to someone.
This modern solution of going to HR or to Twitter is not constructive to society, it creates massive divides, it also creates cowardly behaviour rather than encouraging actual people to talk to each other.
But this becomes a full-time job for minorities to explain what is bigoted to people they don't even want to be talking to, which
1) gets you attacked for seeing everything as bigoted - especially when you make mistakes because you can't know why everyone is doing everything, just see statistically stuff is happening to you and people who look like you more than everyone else, and
2) alienates you from your co-workers, who would prefer that you act according to the stereotypes they have of people like you and laugh at the jokes they're making about you (and your parents, and your parents parents, who were indisputably shat on.) They don't want to hang out with you because they can't relax around you. You're not going to get promoted unless the word comes from so far up you're going to get resented for it, and
0) it's just another burden to constantly be explaining how and why you're miserable to people, even (especially) the ones who consider it self-improvement to listen to you.
The temptation is just to coon for people, say what they want you to say and do what they want you to do, and just silently hate them and hate yourself.
> But it would make life a lot easier if we just sorted out these issues at the source, with two people, explaining what hurts and why to someone.
This is problematic thinking. For example, black people are 15% of the US population. It isn't one-on-one, it's one-on-five-and-a-half at best. And really, if you're a middle class professional (let's say programmer) where there's a lower proportion of black people that would be indicated by relative populations, it's one-on-a-small-army-20%-of-them-heavily-redpilled-and-angry.
I prefer to leave it to the twitter mob, although some of their positions are crazy, and it being twitter the people who are going to be the most vocal are going to have severe personality disorders (usually borderline.) It's still nice sometimes to have them deflect the belligerent white dude from you.
I once had a Black intern come back to his desk and I could see he was unhappy, which was far from his norm. I asked him what was up and there was this long moment of consideration, clearly deciding whether it was worth even explaining it to a white guy.
It turned out he'd left his badge at his desk and got trapped in the elevator lobby, something many people did going to and from the bathroom. When asking to be let in, a white guy gave him the third degree about who he was and whether he really belonged there. Nothing like that ever happened to me, even though my intern was a sharp dresser and I looked one notch up from a hobo.
After talking with my boss, I wrote this up for HR. At the intern's request, we never even named the interrogator. I explained that it was surely an unconscious bias incident and that I thought they'd want to keep track of things like that. In short order I get an email back from an HR lawyer denying, deflecting, defending. They did jack shit. So I also wrote the Black ERG heads. They jumped on it and raised it up to the level of the CEO, which I was very grateful for. It ended up being a reasonably positive experience for the intern on net; he felt heard and respected.
But it has always stuck with me how instantly this was dismissed by the normal power structure. They did not want to hear it, and no amount of me explaining "what hurts and why" would have made a lick of difference.
And that was to a (white) manager with the backing of his (white) manager. I totally get why people targeted with this stuff just keep their heads down most of the time.
> But it has always stuck with me how instantly this was dismissed by the normal power structure.
Your misunderstanding was believing HR is there to solve human interaction problems in the company. HR is there first and foremost to protect the company. In the case of legal, that’s even more the case.
But good on your for speaking up and trying to make the situation better. But I also believe your intern was smart to ask not to name the interlocutor. That would have only made enemies, and probably made things harder on him.
Thanks, but avoiding future lawsuits for racial bias is part of protecting the company. As is making it safe for all sorts of people to work there, as that aids recruitment, retention, and results. So this wasn't about protecting the company.
My take is that HR's first goal is to protect HR, and their second to protect the powerful people in the company. Protecting the actual company is low down on the list.
The phrase "outsource getting into conflict" caught my eye. To me, that is an interesting take/observation. I've been trying to sort out when a disagreement or argument becomes harassment and the closest I can come to is when there is a power imbalance between the parties. I'm inclined to agree that taking your grievance to the mob doesn't solve the real issue.
I think it depends on what you mean a lot by "think". Did they wake up thinking, "Hey, let's go for some misogyny today"? Probably not. But on the other hand, it's not like their behavior is random, patternless. People often do things without really understanding what it means. Indeed, given how hard actually understanding a global society and its history is, I'd say we almost never totally understand anything we do. It's a big world, and people are universes unto themselves.
One book you might read here is, Bancroft's "Why Does He Do That? Inside the Minds of Angry and Controlling Men". It's written by a therapist who mainly dealt with men in court-mandated therapy programs related to domestic abuse. Few of the men ever saw themselves as bad actors. There was always a reason they were justified in their abuse. He goes into great detail examining how abuse worked well for them.
I also think you misunderstand the systemic nature of things like misogyny and racism. You are effectively saying it's the job of women to fix sexists. Putting the burden there acts to support sexism. That's true anywhere, but it's especially true when we're talking about academia. Look at that UMich CS professor: he had many opportunities to understand his behavior was wrong. He surely heard it from women. He certainly heard it in trainings; "don't grope the students" is something every professor knows by now. Ignorance is not the problem, and suggesting that women with little power should educate men who can ruin their careers only helps abusive men.
Ignorance can be the problem in specific cases, of course. But even there, individuals are responsible for their own behavior. If men would like to not be sexist, they should study the topic. For the HN crowd, I might suggest Manne's "Down Girl: The Logic of Misogyny" and the follow-on book "Entitled". Both are sharp, readable, and very analytical looks at the topic.
I agree that jokes can be subversive; I don’t agree (perhaps you don’t either) that subversive speech (jokes or otherwise) merit termination, which is to say I’m not an authoritarian. With respect to racism, I don’t doubt that there are too many racist jokes, but the Twitter sphere tends to miss those in favor of jokes which are decidedly “antiracist” i.e., jokes which make fun of racism (whether left-wing racism or right-wing racism).
Taking the Wikipedia article at face value, and taking for granted the exact word it cites, I fail to see anything sexist in what Tim Hunt said in that conference. If anything, he was joking about sexism itself. The people who say are offended by his words either read a quote out of context, or did the cherry picking themselves.
Assuming you followed the same Wikipedia link I did, it would seam you are guilty of cherry picking. Let's explore this mistake together:
> Let me tell you about my trouble with girls. Three things happen when they are in the lab: you fall in love with them, they fall in love with you, and when you criticise them they cry. Perhaps we should make separate labs for boys and girls?
So at a first glance, it seems to be as you say. That would indeed be disgusting. Wait a minute though. In your haste, it seems you failed to read the words that preceded:
> It's strange that such a chauvinist monster like me has been asked to speak to women scientists.
…as well as the words that followed:
> Now, seriously, I'm impressed by the economic development of Korea. And women scientists played, without a doubt, an important role in it. Science needs women, and you should do science, despite all the obstacles, and despite monsters like me.
So first, he tells his audience to take the words that will follow with a grain of salt: "Hey, I'm a monster, don't be surprised if I say monstrous things!". Then he says the thing (with less than ideal words, he could have said "me" instead of using the generic "you"). And finally he explicitly signals that the joke is over ("seriously"), and go on encouraging women to do science, and fight the very misogyny he just incarnated in his joke.
Did he actually used rank sexist stereotypes to suggest that women should be pushed out of men's labs? Of course not, and you know this.
---
Now let's stop the cherry picking, and acknowledge that we have both false positives (people getting fired over misplaced outrage), and false negatives (people not getting prosecuted for serious offences). While I understand the need for getting fewer false negatives (rape prosecution rates for instance are appallingly low), the solution is not to move the cursor all the way to hair trigger sensitivity: you'd just end up with far too many false positives, and not enough attention left to solve the real issues.
He did literally both use rank sexist stereotypes and suggest women should be pushed out of men's labs. There's just no denying that. He did that as part of what he may have intended to be a joke. But when your audience doesn't laugh, it's a bad joke. If one is going to joke about a fraught topic, there's a strong obligation to succeed.
Even taking it as entirely sincere and well meant, something I don't think women in science are obliged to do, his "solution" is apparently for women to just ignore the sexism, something that places the burden for men's sexism on women. That is also sexist. So again, this looks like a failure to me.
I therefore think your theory this is a false positive is incorrect. I think the most that you can claim is that the level of outrage is disproportionate to the particular offense. But that analysis ignores the extent to which sexism is utterly commonplace in a society that has oppressed women for centuries and is still working its way out of it. So you can argue that this wasn't perfectly fair to this one guy, but it's not disproportionate to the problem this guy was part of. And a) that rings of himpathy to me, b) that ignores the much, much greater degree of unfairness caused by sexism, and c) that focus itself helps protect sexism. If fairness is really what's motivating you, your time is better spent on the many early-career women continually being harmed and pushed out of the sciences, not one old white guy who is already back doing what he wants to.
> his "solution" is apparently for women to just ignore the sexism, something that places the burden for men's sexism on women. That is also sexist. So again, this looks like a failure to me.
That I can concede.
> If fairness is really what's motivating you, your time is better spent on the many early-career women continually being harmed and pushed out of the sciences, not one old white guy who is already back doing what he wants to.
Agreed. This cuts both ways, though. Attention directed at slandering the guy on Twitter is attention not devoted to actually help discriminated women.
If slander were the point, sure. But for a lot of the people calling out sexism, etc, the point is not really the one offending guy. It's the caste of guys who have been supporting and benefiting from the problem that the current focus is symptomatic of. It's the system itself. But humans mostly don't think in systemic terms act to solve systemic issues. They work in terms of narrative, of example.
If the point is not the one offending guy, how do we justify his sacking? We could say we needed to make an example of someone, and that ended up being him, but I bet my hat it wasn't a conscious choice.
My problem with this whole thing is that our attention is focused on the visible things, instead of the real things. Harassment to name one tends to happen quietly, subtly, often away from witnesses. There's also a good chance that sexist jokes in public speeches are a consequence of a sexist atmosphere more than they are a cause. While they should indeed be addressed, we should be wary of fixing the symptom (looking good on the internet or in front of journalists), without addressing the actual cause (toxic work environment).
He was hired for leadership/advocacy positions and compromised his value as a leader/advocate, so I can see why he resigned.
I agree that his behavior was driven by the cultural artifacts of patriarchy, but the very best way to continue that culture is to have leaders who are comfortable with the those artifacts or are clueless about eliminating them.
I'm all for anything that fixes toxic work environments, of course. And there are many subtler approaches. But as the UMich situation shows, those approaches often fail. Existing systems are good at sweeping incidents under the rug. So I am also all for giant public reactions to failures that are so painful for the organizations involved that change actually happens.
As an example, consider Brock Turner's judge Aaron Persky. His prominent himpathy for a rapist made him an internationally known figure and cost him his judicial seat, the first CA judge to be recalled in 80 years. Was this fair? In one sense no, in that other judges were surely equally bad in going easy on rapists that they identified with. But on the other hand, he was part of an institution that had 170 years to get its act together on rape. So although I would have preferred that the CA legislative branch had fixed this problem at any time in the past, they hadn't. And you can bet that a lot of male judges who would have scoffed at, say, mandatory training have taken careful note of what happened to Persky.
If you think you can swing the creation of effective training programs or other more real interventions, definitely go for it. But if not, then I think you'll have to get comfortable with the big-failure-and-strong-reaction model. Because as a general rule, the people who have the power to drive systemic change are not doing much about America's endemic sexism and racism. Until that gets better, activists are going to keep using the powers they have, a big one of which is making big examples of visible problems.
Persky sounds like a good example of… being made an example of. Maybe not fair, but definitely useful: it very likely had a real effect on the remaining judges.
> If you think you can swing the creation of effective training programs or other more real interventions
Actually, I'm totally for utterly destroying the reputation of an individual, or an institution. It just have to be done for the right reasons. Take for instance rape culture as was prominent 5 years ago in several French business schools (it's not over yet), that I've learned about just this week. I confess I seriously had no idea. So the problem had (still has) several facets. First there's the general mood, which is blatantly sexist and demeaning to female students. Then there are a shocking string of sexual assaults, as well as many rapes and rape attempts (like 10% of all female students being victims of attempted or actual rape during their stay at the school —possibly more, probably no less).
It would appear the only way this can change is external pressure. What happens needs to be publicly visible, so we can have a scandal and apply popular pressure. The problem is what exactly what we should be scandalised by, how much, what punishments we wish to exact. If we fire a student or a professor for making a sexist joke, there's going to be serious backlash, downplay, and objections —including from me, see this whole thread above.
Which is why I believe it is best to punish the worst offences first: rape, attempted rape, failure to properly punish those, sexual assault… in roughly that order. Few will seriously object to a public scandal involving a strong presumption of rape. Few will seriously object to dissolving student organisations that routinely (or even just once) set up the conditions for repeated sexual assaults (to give one example: invite the girls first, give them free unlimited strong drinks for 2-3 hour, then let the boys in, all worked up and horny).
Once you get the ball rolling, it's easier to crack down on the sexist jokes. React with disgust, make them uncool, not fun any more, passé. And give a serious slap on the wrist if such happens at a public event (like a temporary suspension). Just focus on the rapes first, if only to gather sympathy.
I am not particularly referring to Hunt's "joke" there. There's a whole genre of "jokes" used to denigrate women.
I disagree that everything is "about" control or manipulation, but as status-oriented primates, humans certainly do inject it into everything. However, that's all the more reason we have to examine little things like "jokes" and be conscious of the effects we're having.
At a lunch for female journalists and scientists, Hunt gave a speech...
"It's strange that such a chauvinist monster like me has been asked to speak to women scientists. Let me tell you about my trouble with girls. Three things happen when they are in the lab: you fall in love with them, they fall in love with you, and when you criticise them they cry. Perhaps we should make separate labs for boys and girls? Now, seriously, I'm impressed by the economic development of Korea. And women scientists played, without a doubt, an important role in it. Science needs women, and you should do science, despite all the obstacles, and despite monsters like me."
People keep bringing out that this can be a joke, but he said this two days later
"I did mean the part about having trouble with girls. It is true that I have fallen in love with people in the lab, and that people in the lab have fallen in love with me, and it's very disruptive to the science. It's terribly important that, in the lab, people are on a level playing field. And I found these emotional entanglements made life very difficult. I mean, I'm really, really sorry that I caused any offence – that's awful. I certainly didn't mean – I just meant to be honest, actually." [1]
He was given a chance to clarify, he doubled down. While I'm not saying a joke should cause people to be fired, but this is clearly more than a joke.
So what part of falling in love with people in the lab, people in the lab falling in love with you, and that affecting concentration and productivity, is so evil as to get someone fired?
Sorry, honest question. Being from a different culture, I honestly fail to get this kind of outrage. For me it's just a description of humans being human...
It's painfully reductive and one-dimensional. What about envy and hate, hero worship, and other emotional attachments? Those have no effect on working environment?
What about men and women who aren't romantically attracted to (respectively) women and men? Are gay men relegated to the women's lab? But only one per batch, lest they fall in love with each other? (And bisexual people can only be trusted to do science on their own.)
In general don't we expect "professionalism" to include a level of managing your emotions? And this person is basically saying "I can't deal, therefore certain other people must be kept away so I don't get distracted". As well as tarring women in general as not being able to deal, which is unfair. I sure wouldn't want to work with this guy after hearing him say that.
Romantic comedies are fiction, though, intentionally exaggerated for entertainment. I don't deny there's a kernel of truth to that kind of story -- or else they wouldn't be interesting at all -- but I hope we're not taking them as a model of workplace behavior.
Right but my point is that love is something you fall into, not a conscious choice (as opposed to making inappropriate comments or non-consensually touching a colleague where you shouldnt. These are things professionals are reasonably expected to manage in the workplace)
Precisely. There's some sort of "if you have authority you must be better than me" feeling and "better than me" admits no flaws or human variety at all, apparently. Some folks want perfect Gods to follow and keep failing to make them from people made of meat.
It is difficult to distinguish genuine romantic feelings between two people, from the case of a superior using their position to get their genitals wet and a subordinate capitulating for fear of losing their job.
Since feelings are only a biological impulse, and we humans frequently suppress our impulses in the form of self control, it's much easier to look for that oxytocin fix in a more appropriate arena.
The military has been doing this for ages, forbidding officers from fraternizing with enlisted. And plenty of civilians abusing positions of power have proven the wisdom of such a policy.
It's difficult to distinguish genuine romantic feelings from exploitative lust everywhere. It's a constant of human experience and has very little to do with power dynamics.
They're saying people are assholes. For instance, in some American cultures (see my previous comment for which culture), it's a given that perhaps 40-60% of married individuals are cheating on their spouse. That's not love--that's doing what feels good, and then doing someone else that feels good.
Where there's a power imbalance, it's easier to ban a class of abuses than to figure out the small percentage of cases where both parties are genuinely afflicted by mutual biological imperatives.
Okay, I got a little off track. Let's say you're an HR person (or whoever is at legal risk if an employee decides they've been taken advantage of), and someone come in with just such a complaint.
How would you, an outside party, determine whether the superior was really [infatuated, in love, whatever], and not simply taking advantage of their situation? Or how would a judge determine that? Is it worth it to the company to work through that process every time it happens? What about the people who really were victimized, but the evidence is circumstantial and the court says otherwise? Isn't it easier to exclude the small pool of people that are subordinates and tell the supervisor to find romance anywhere else?
The controversial part is where he suggests that this is a "problem" and might be a good reason to exclude women from working in the lab alongside with males. That looks like he's starting out with a very sexist attitude and trying to justify it with flimsy excuses.
That part was the joke. He's saying it's a real problem, but that's clearly the wrong solution. For him, it's obviously a bad solution, and so worthy of ridicule.
As a policy it denies both women and men the opportunity of working together. If you are a man and a woman happens to be working on the same problem that you are investigating, would you like to be excluded from learning from her? And vice versa.
Exactly. So his proposal is not a practical solution to the problem of human behavior affecting productivity, but it’s not “sexist”. Many single-sex schools exist and have strong proponents, but are rarely described as “excluding” people.
> So what part of falling in love with people in the lab, people in the lab falling in love with you, and that affecting concentration and productivity, is so evil as to get someone fired?
Fundamentally, he is saying that since he has trouble keeping his emotions in check around women, the solution is to not allow women in the lab rather than developing his own managerial or social skills.
It's understandable why some would question the wisdom of having this person responsible for developing the skills of female scientists.
Should he have been fired? I don't know. Certainly not if tis joke was is only "offence", but I suspect there is a bit more history to the situation.
I guess for me the thing that's really crappy about this quote is that it shows his underlying attitude - that women are basically always potential romantic partners.
If he was into men, and he said he didn't want men in the lab because he might fall in love with them, you can sort of see how absurd it is, and how unpleasant it is to be the object of romantic fantasy when you're just trying to get on with your job.
> women are basically always potential romantic partners
Ok, why would that not be the case? Laws? PC? Age difference? Love/biology doesn't care about social rules, and this has been shown time and again in every possible situation you could think of.
> you can sort of see how absurd it is
Huh, no I can't. What makes it different when you reverse the situation?
The point is, any human can be a romantic partner to any other. Therefore, his argument should be that no pair of humans should work together ever for risk of romantic entanglements.
Except it doesn't work that way, because we're all really used to the idea that in the workplace, you treat your colleagues as colleagues, not as fantasy-future-partners.
This isn't PC. It's just basic common sense, that he's lost his grip on, because he sees women first as romantic partners or sex objects or whatever, and second as scientists.
Because of professionalism. The root of term used for “professions” like law, medicine, engineering etc. is that one professes to a code of ethics. That code should overrule base desires.
We generally wouldn’t accept a doctor who views and treats patients primarily as an income stream despite greed being a near-universal human drive and we shouldn’t expect a professor to view subordinates as potential romantic partners. Acknowledging the drive exists isn’t a reason to condone it.
Maybe you can help me understand the context better. From the GP post where he seems to advocate for separate male and female labs he seems to imply there isn’t enough professionalism present to have co-ed labs.
I’m saying that claim is more an implication of the person saying it and their (lack of) professional ethics than an indictment of the subordinates. It’s very similar in my mind to the recent arguments about gender in military units
The context is that he in essence says that this problem has no good solution, but he thinks that the co-ed labs are the best alternative even with all the shortcomings that go with them. Everyone will be perfectly professional until someone falls in love and the PC solution crumbles to dust. And FWIW, I think he's right.
>Everyone will be perfectly professional until someone falls in love
Isn’t this the case with everything? I.e., if “everything is fine until it isn’t” it’s not really saying much of anything except he doesn’t think he can create a culture of professionalism within his lab. Does this “welp, we can’t do anything about our base desires” extend outside romantic relations? Would it be acceptable to claim “well, physical altercations are just going to happen because you know people will get mad at each other from time to time”?
I’m not hiding behind professionalism, I’m saying it’s reasonable to acknowledge those base desires while also expecting a higher standard of behavior.
> Would it be acceptable to claim “well, physical altercations are just going to happen because you know people will get mad at each other from time to time”?
Are you willing to punish people with jail time or worse for falling in love and adopting the behaviour that goes with it? This is the other extreme of your argument, and there are many places in the world where this is the social norm.
The PI can do everything he/she wants, love will happen and people will behave accordingly. The point is acknowledging that this is not a problem that arises at a single point in time allowing you to fire the offender, but that it happens along a continuum that will constantly decrease lab efficiency.
>Are you willing to punish people with jail time or worse for falling in love and adopting the behaviour that goes with it?
No, because one is a criminal offense and the other is a breach of ethics. From that standpoint, it was a bad analogy. But I would hold someone accountable for being unprofessional in the workplace. To be clear, I’m not saying to punish people for falling in love, I’m saying you can hold them accountable for letting it affect the workplace and creating an unprofessional environment.
>The PI can do everything he/she wants, love will happen and people will behave accordingly.
This is probably where we disagree. I think the PI holds some responsibility for setting the tone of the work culture. You may not be able to control people’s feelings but you can make it clear that certain actions are not going to be tolerated. That’s especially necessary in cases of fraternization. It’s the PI’s job to maintain the professional standards of the lab.
Plenty and plenty, yes. Now, what's a solution that would work for everyone, males females and others alike?
This is very clearly a professor acknowledging the problems that arise due to interindividual biology in work environments. Unlike PC supporters hiding the issue under the blanket of professionalism.
I think an obvious solution would be to exclude anyone who is incapable of managing the bare minimum level of professionalism that's required in a workplace.
If someone is incapable of managing their feelings in a workplace then maybe they don't belong in one, and their colleagues should not be the ones who are punished for that.
When you're, like 2 years old, and you want a toy in the sandpit, you grab it. And if some other kid has it, too bad for them, because you want it, and that's what's important.
Somehow, most people manage to figure out that this kind of behavior isn't appropriate, and hide the issue under the blanket of being-a-decent-human-being.
I don't understand how dealing with sexual or romantic feelings is any different.
Not really, no. Human nature is what we're dealt, we must embrace both the positive and negative aspects of it. Perhaps the inevitability of two people gravitating towards one another can be leveraged? Perhaps the disparities can elucidate us on unseen proclivities in different populations, things that can also be leveraged and positively.
What we should avoid is cramming people into functionary roles and instruct them they must act as would a machine. No longer can they be compelling or compelled but only impelled as would be a gear turning in the insurmountable forces of the engine that drives.
Often, yes (though usually not to the point of actually dying). Overcoming survival fears is behind every act of physical courage, e.g., saving a friend or stranger, or exploring any new zone that will normally kill you (mountaineering, undersea, space, etc.).
You have got to suppress your naive survival instinct (or remain massively ignorant) to climb on top of a rocket with thousands of tons of explosive material...
> FUnny you leave out the part where he also said if you criticize a woman in the lab they cry...
Do you think women in the lab, when criticized, are more likely, less likely, or equally likely to cry?
Is the scientist’s comment mean spirited, or sexist, or just an observation?
I think it’s important to consider what the intent behind these jokes are. The Wikipedia article calls out statements from 29 other scientists that note how women (and men) were advanced within his lab and outside his lab.
So if this person thinks that the women he’s worked with cry when criticized, so we not want him to say that? It seems more like the goal should be to not stigmatize crying as that seems pretty reasonable for all genders, rather than to stigmatize talking about crying.
The rule is really simple: it makes people uncomfortable when you make generalizations about natural traits shared by the group they're in. Period. All groups (even groups people feel proud to be a part of), and all generalizations, even ones that sound positive or don't apply to the listener. I'm not going to list examples but if you're having a hard time thinking of them just imagine overhearing a conversation at a coffee shop about "those <something you are>, they're always <something you do or don't do>."
"WHen making terribly misogynistic comments his INTENT wasn't to be misogynistic, it's just based on his experience that women are driven primarily by their emotions and unable to handle pressure!"
I’m not defending him and it’s funny you think this is some pro/con situation.
That being said misogyny requires intent, right. It means someone who hates or dislikes women. So if you make a statement that every time you criticize a woman, she cries and don’t have ill intent toward women, then that isn’t a sign of misogyny.
I don’t think women are any more likely to cry than men, but if there’s research that shows it so, is that misogynistic?
If you say “women are shorter than men” is that misogynistic?
I think it largely depends on intent as if someone is trying to demean women or does hate women, that’s a big difference. Saying women are shorter than men as part of some overall argument on inferiority is clearly misogynistic.
That misogyny or bigotry requires "intent" is so ludicrously divorced from reality I don't know where it came from.
I've heard the most ridiculously hateful and racist things prefaced by "I'm not a racist but..."
Clearly in these people's own minds they aren't racist/misogynist, but when you are pushing plainly racist or misogynistic views it doesn't matter.
The statements are misogynistic. If you think you can convince your boss your intent wasn't to be bigoted so your bigoted comments aren't actually bigoted then be my guess. I am confident you are unlikely to succeed.
The definition of misogyny literally includes intent, perhaps you’re thinking of something other than misogyny.
Here’s the definition from Wikipedia [0]… “ Misogyny (/mɪˈsɒdʒɪni/) is the hatred of, contempt for, or prejudice against women or girls. It enforces sexism by punishing those who reject an inferior status for women and rewarding those who accept it.”
You also didn’t answer my question of what negative information you wouldn’t interpret as misogynistic. Is it possible to report anything negative based on biological sex or gender that you don’t interpret as misogynistic?
How can these differences be studied and discussed to learn more and overcome harm caused toward women?
The one about woman crying when they are criticized? That one is certainly untrue. While some people cry when they are criticized (I personally am severely effected and have cried in the past but afterwards), women have no extra tendency to do this.
I can't see it that way. He is open about his own weaknesses as well. We don't have to look up to him for saying this, but that doesn't make it your typical one-sided bashing.
One ambiguously offensive comment during a speech by an aging professor isn't something I'd particularly detest. Outraged cancel culture social media mobs screaming for someone's head at the slightest supposed provocation, on the other hand are thoroughly detestable.
The balance of the problem here isn't this professor, who otherwise spent a large part of his own career helping many women (many of who defended him after his speech) with their own scientific careers. Instead the much more insidious problem is a hysterical social tendency of people with influence in their support for an increasingly dominant social norm then mobbing together under said social banners to destroy entire careers and lives at even minuscule transgressions of their ideas of correct thinking.
The point of these attacks often isn't even about genuine justice for marginalized people. Instead it's about establishing dominance while virtue signalling as aggressively and righteously as possible.
I’ve noticed a pattern where, wherever someone feels the need to write (after a disparaging story about someone else) “They were not joking” or “They really said this”, it’s almost always false, and the disparaged person really was joking, or really did not say that, or it was taken wildly out of context, etc.
What's the missing context here? That he used an important public speaking opportunity to actually mock the women he was assigned to help, and we misunderstood it as a serious appeal? That he was being fake offensive because it would be funny or enlightening?
This wasn't an opportunistic pun. Humor is rooted in one's worldview.
The missing context is that many women he worked with defended him. That he has consistently worked with, hired, supported, and promoted women throughout his career. That he has done far more to benefit women in science and humanity generally than the whining Twitterati who denounced him ever will.
It sometimes surprises me how people don't seem to see that while all of this is part of a healthy (and seemingly normal) societal change, that it's unfortunate that not just are oldfashioned behaviors shunned, but that merely talking about struggles with them is so taboo. Is society really going to adapt better because people lash out so uncompromisingly?
I don't know any of the context here, but it seems to me that what the quote actually says is that the author admits that he is unable to function effectively with female colleagues, that there are other men like him, and that it's a problem that mustn't be allowed to hold women back from doing science.
The way I read this; when he says "you", it's the generic you: he's talking about the "chauvinist monsters" like himself, and it's in the context of three particular ways in which his own failings prevent him working with women.
He forms and encourages inappropriate emotional relationships with his female colleagues and it affects his ability to give criticism of their work effectively.
Now you have the right to read that a different way; but please respect that others are not just "memory-holing" misogynistic comments for some reason or another.
I'm not defending him: he definitely has a problem with his attitude to women, it sounds like it might absolutely create a hostile workplace, and it's almost certainly inappropriate to be talking about the subject in such a light-hearted manner - but this is fundamentally a mea culpa rather than a criticism of women in science.
Yes I imagined his idea of "criticizing" amounted to abusively yelling at people. From what I've seen, the way people win Nobel prizes is by working their lab 24/7 like a slavedriver. I actually have seen psychopathic professors yell at female lab members until they cry. I knew one guy who complained about it. He apparently got much more productivity out of yelling at the guys. Pretty dark humor if it's a joke.
"Now you have the right to read that a different way; but please respect that others are not just "memory-holing" misogynistic comments for some reason or another."
Almost every single one of the posters defending this misogynistic behavior are just deliberately leaving out his most incendiary and misogynistic remark and instead focusing on comments he made that are less objectionable.
Why do these people deserve "credit"? Am I suppose to believe only focusing on the least objectionable comments to paint him as some unfairly maligned martyr is just an accident?
I mean, who could prove him wrong? Maybe his idea of "criticizing" his fellow professionals involves a lot of yelling and throwing chairs around. There's certainly some people like that in the workplace.
I don’t know anything about this specific case. It might be the exception which proves the rule.
In general, I try to follow the HN guideline: “Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.”
He also implies that if you're too good of a student, at least around him, he might fall in love with you and become a worse teacher, and he's just being honest about it. You could mock men for being bad at suppressing feelings equally from the same comment.
For one, you're cherry-picking two examples. Some professors don't get fired for jokes nearly instantly and others get immediately fired for abusing their students.
Second, amplification plays a part. A story with "two sides" (such as the joke one) will draw far more public attention than another with only one side, just because there will be more debate and people using it to advance their agendas. In this case, it includes "PC Twitter" but also all of the people who scream "freedom of speech" whenever someone is criticized for saying something dumb or inappropriate.
Much easier to prove the case of terrible jokes or non-PC behavior. Twitter has streamlined that stuff for years now - and what's more, that kind of behavior is much more relevant for more people, rather than academic fraud which might be a very niche thing, only relevant to small circles of people.
One could ask - why is it that petty theft can land some people behind bars for years, while wealthy people committing tax fraud only get fined, or at most a couple of months in a cushy white-collar crime facility? Well - for one, the former crime is much easier to prove, especially if you're caught red-handed. The latter crime tends to be incredibly complex, and will cost a ton of resources to prosecute.
I guess the same goes for your questions. Professors incriminating themselves on twitter or youtube - easy as pie.
Collusion rings with respectable professors, that probably use students for dirty work and plausible deniability, and resources enough to fight their employer in the courts: hard fight
Not all targets and not all campaigns are created equal? If someone has a lot of power within the institution, and the campaign comes mainly from people the institution doesn't care about… you can fill in the gaps.
I seriously doubt that exactly what someone has done is the major factor. People in weak positions can easily lose them over the tiniest of things; people in strong positions can get away with murder!
Number one was a highly offensive “joke” minimizing a truly serious problem that was made to a number of journalists to whom it definitely wasn’t funny. Plenty of people are fired regularly for making complete fools out of themselves and the institutions they represent by offending important outsiders.
Comparing these personnel decisions from not only different universities and entirely different incidents but also an entirely different country seems unwise to me.
No tenure, no exploratory science and no moonshot. Newton was a dangerous asshole. Would you exchange differential calculus for squeaky-clean academia? I don't think so ...
There's an old quote to the effect that, if forced to choose one thing to destroy: either the Principia Mathematica or the Sistine Chapel*, choose the Principia. The reason? The Principia is a monumental achievement, but it's also universal, and it can be rediscovered. The paintings are the unique creation of one moment in time, unduplicatable.
*Or some other work of art; I can't remember the original exactly.
There's a lot of questions that we need to address, cause and effect in the case of tenure and the general priorities of academia.
First you implicitly assert that tenure effects the cause of emergent moonshots. How much evidence do we have to the contrary? Cursory research shows the modern application in the US dating to the vague "19th century" not a long timeframe. Tenure itself appears to have emerged in the same timeframe. The modern US application of tenure (secondary) was put in place in 1940.
I'll grant you that we have seen a good deal of progress, but I don't know that you could make a robust argument that without tenure, that progress would be absent. I would assert that it falls into inconclusivity, and that to form an argument would require speculation and conjecture. All things are not made equal, and so finding a suitable control group to compare against would be impossible.
We can look at history, though, and see that there was a plenitude of highly driven scientists publishing and advancing understanding prior to the advent of tenure. But to say that we can transpose that to the contemporary model itself is a conjecture.
Simply, we do not know, and can not know.
As to the priorities of academia, and tenured individuals, and the metrics that institutions use to enlighten themselves on the performance of individuals we seem to have come upon a system of perverse incentives. That is exactly everything, to me, it seems we had ought to avoid. Tenured academics can obviously be terminated, but not in frivolous contexts. They are expected to hold some degree of real responsibility. What tenure grants is their freedom of opinion, and the right to fail in their pursuit. As we know, science is the art of failing upwards in a controlled direction.
"In all lines of academic investigation it is of the utmost importance that the investigator should be absolutely free to follow the indications of truth wherever they may lead. Whatever may be the limitations which trammel inquiry elsewhere we believe the great state University of Wisconsin should ever encourage that continual and fearless sifting and winnowing by which alone the truth can be found."
--Theodore Herfurth, 1894
But the cutthroat competition, the model of "publish or perish" elimination, the perverted demand for conclusivity all stand to imperil the actual aims of science. This is all somewhat incentivized by the tenure system, by personal interest, and by the implicit obligation to "realized" progress - except this is of course not real progress. It isn't as concrete as the fundamentals, the traces are laid much finer these days and replication of research appears more often infrequent, while the quality of publications is increasingly called into question and an economy of debauchery contaminates data for capital and personal gain. And thus the bastion of humanity is corroded while evermore maintaining its authority outward. A real hazard if you ask me.
I believe it all needs reform and serious reflection to build it back better.
I had a professor in a chemistry lab that was proud of the amount of complaints he received against him.
He literally threw the two ring binder it at me while I was in his office. It must have been a hundred pages. He kept it on the wall like a trophy.
I talked to a Counselor at the school, and before I could complete my sentance, he said Dr. Berzergian? (I don't remember the exact spelling of his name.). The Conselor said he, and the Dr., almost got into fisticuffs over his attitude. He told me to take the course at another college.
I realized later all his "problem" students were young males.
Yes--I truly believe this was his twisted way of hitting on people.
A few years later, I was in a bar in San Francisco talking about this professor whom really gave me a bad time. By chance, he knew of the guy, and told me about him.
This was in the nineties, but oh boy if he acted this way today, and I stole that stack of complaints---well who knows?
This professor caused students to change majors, and even drop out.
Getting tenure means (ideally) that the system considers you have proven yourself as a competent researcher and therefore accepts to release a bit of pressure on publishing to allow you to pursue more exploratory objectives. It does not mean you're allowed to behave badly and (officially) does not protect oneself from the consequences of such behaviour. Remove tenure and you remove the last bastion of real research we have left in our industrialized and quasi-corporate western research institutions.
Removing tenure would completely trash western science and in practice yield total scientific leadership to eastern powers, who still have old style academic systems with strong tenure positions and less concern for academic mistreatment and "wokeness".
Where do you draw the line for tenure? So hitting colleagues with a binder is acceptable, okay got it. How about hitting them with a baseball bat? How much tenure we need for that? Should Einstein be allowed to kill physicists because he's a genius? If we already agreed to make tenure an excuse for doing bad stuff, let's make it official and negotiate the legal borders.
Removing tenure is indeed a very bad idea. However I think we're talking about using tenure as a hall pass for whatever asocial attitude some might practice. Because hitting your fellow researchers with a binder doesn't do much to advance science, that's for for sure... it's not like the tenured scientists had all the answers and by being aggressive they actually quench different - and possibly novel - ideas.
For that matter, there's nothing in tenure that requires universities to allow bad teachers to continue teaching. Tell him to go research something, and if anyone thinks his attitude is worth dealing with for his expertice, they can approach him volutarily.
Supply and demand. Everyone wants to do cutting edge research but nobody wants to pay for it. People who know how to get it paid for hold all the cards and can therefore get away with anything.
I definitely wonder if industry is any better here. At least in academia, the papers are public, so there's an opportunity for scrutiny. But I've heard tell of "AI" boondoggles in both large companies and small. E.g., the large corporate "AI" efforts burning millions without making any real improvements. And I wonder how many startups out there have standards that are in effect lower than academia, but instead of writing papers they are shipping products than harm people's lives when they go wrong.
As a previous Data 'Scientist' I can tell you that it absolutely is.
Garbage in garbage out is the norm, the models are long established, but they can't mine gold from dirt. But nobody except the engineers seem to understand this.
I'm so glad I left to become a regular software engineer. My code does not depend on a blackbox that is fed with crap, and can be reliably tested.
It is rampant in consulting business. At least in an internal project it's possible to pull the plug when the results are not promising.
In a consulting engagement, when the results are garbage, there's no revenue and no potential for an upsell. In effect the pressure to "find" significant results is enormous.
Of course a lot of abuse happens in private enterprise. But typically, senior management is much much more proactive about it than department chairs and provosts. There's very few Fortune 500 companies where you could publicly get away with the kind of abuse that's just part and parcel of being a grad student or post doc at a major research university.
It might still happen either because it's well concealed. Or it might happen at smaller or poorly managed companies. But at the typical functional, large corporation, egregious abuse of your subordinates essentially guarantees that you'll be terminated if/when it's brought to senior management's attention.
Academia is different because tenured professors are given far more independence and autonomy. By contrast middle managers are tightly monitored and controlled by their own line managers. The typical large corporation strives very hard to promulgate a homogenous corporate culture across the org. Whereas academia as a system encourages professors to be fiercely independent maverick. That has both pros and cons, but one of the major cons is that it tolerates a lot more abuse and dysfunctional management towards the non-tenured subordinates.
Your rosy picture of industry is surprising to me. It could be right, and the few takes I've heard (eg. recently, google's AI ethics mess) being only one side or exceptions that prove the rule.
But I have currently no reason to expect better from industry, and will want some proof before putting it on a pedestal.
Abuse happens to people who don't have other options. If you abuse someone who has a white collar job, they generally leave quickly.
If you abuse a poor grad student who has to complete research to pay back his loans which cannot be discharged through bankruptcy, he will generally put up with it.
You mean the number of suicides at foxconn? The forced labour that is likely contributing to many of the large mobile phone company bottomlines. The blatant benchmark fraud that happens all the time by all the large GPU and CPU manufacturers, I could go on. But I don't see the press all over this 24/7 t all, maybe some small niche outlets sometimes. In comparison the scrutiny that the academic world is under (if we relate to the affected people and effects) is much, much larger
There are fewer suicides at FoxConn than would be expected given the age profile and number of their employees. If you employ hundreds of thousands of people some of them will kill themselves for reasons entirely unrelated from work. If they live in work dormitories they’ll do it at work.
Yes, I agree. It seems to me that academia feels external to the power structure, so it rarely gets the same kind of attention as industry because it doesn't stir up resentment.
They are part of the power structure, they're just 'protected'. Most publications don't want to be seen as promoting a narrative that goes 'against science' even though of course it's not. There's no room for nuance in populism.
It's more the fact that academia is granted a greater degree of leniency because it is fundamentally something that is societally required for innovations to take place. Genuine ingenuity requires room to flourish, so in the West at least, the theory is you leave tge Boffins alone to do their wizardry.
The part I don't understand is how extorting the students to fund bloated Administrative processes, sport teams, stadiums, and executive staff ever became a thing, nevermind the incestuous relationship with academic publishing. Almost all great work I've seen came not out of gnashing of teeth and publish or perish, but out of a labor of love or an odd obsession with truly understanding something until you could practically get it across to a 5th grader.
"I can explain how, don't ask me why too many times, still figuring that part out."
The faculty have very little say in most of those aspects of universities you're complaining about.
The media is just a crazy sideshow that cherry picks a tiny subset of stories to run or people to destroy when it fits the right narratives. When it comes to real scrutiny, I'd say faculty are under vastly more than people in industry. Though yes, google as an entire entity will analyzed more than some random professor. But the rank-and-file professor is also probably in far more constant danger of being ruined than almost any individual in a comparable role in industry. Also more than most businesses that no one cares about.
Academia is not protected because of some intellectual notion of 'innovation' so much as they are on the right side of the bias presented in most publications. Most writers and commentators I think have venerable views of academia and probably err towards supporting that narrative publicly.
While it's true that Profs may be under excessive scrutiny in some ways, which frankly might make them skittish - they are obviously not under 'the most important' kind of scrutiny which relates to the material legitimacy of their work in terms of 1) reproducability 2) fudging results and borrowing ideas and 3) misappropriation of credit 'up the chain of power' and 'from other peers'.
Hence this article, and some other issues of legitimacy within academia.
I think almost everything boils down to the fact that the low-hanging fruit have been had in science, and though there are 20x more scientists alive now than just a few generations ago, in many ways we're getting diminishing marginal returns - and even worse - it's incredibly hard to know which teams to back, and which not to.
In the fog of war for funding, it leaves more room for back-stabbing than in most other places, even in the corporate world where at least there is some degree of job security.
He can't blame other that himself for commiting suicide.
We should stop worshiping this kind of people. He has the option to join another team. One of the thousands of scientific teams not directed by a psycho and working on millions of interesting problems waiting to be solved. To leave and do other things was also an option. You can have a really fulfilling and happy live without being a scientist (Is more probable in fact).
People with suicidal tendences had deep inner problems that didn't started necessarily in the university. Some are attention suckers, manipulative professionals that need to assume the protagonic role, idiots that decide to jump by a window to avoid facing their first real conflict in their lives.
They dream about to punish papa, mama, the evil teacher and the cruel world that apparently owed them a career in science. They smile with the idea of everybody attending their funeral in a rainy day with sad faces in dread. The main motivation behind most (true) suicidal people is collective punishment.
Some people can choose to feel miserable for the rest of their lives when this happens. Other will be wiser, break the endless stream of bullshit and refuse the role of punished. Both groups will eventually keep with their lives instead to feel guilty and miserable for an unexpected act that was beyond their acts or wishes.
I partly agree, at the risk of sounding too callous. One of the reasons that reading about this suicide (and so many others) is painful, is because with our outside perspective we can see that he could have moved past it. He could have changed some aspects of his life, and very possibly had a happy future. He succumbed to short-time thinking.
But... that doesn't mean that the conditions that led him to suicide aren't worth examining and correcting. Even if he didn't kill himself, they would still be causing misery and undermining scientific progress.
Which is why several journals are pushing for reproducibility as a requirement and dedicated reproducibility tracks where there are opportunities to call bluff on established papers. As an academic from a small lab in an unknown university it is shocking for me when we are unable to reproduce performance numbers from papers by popular institutions.
> several journals are pushing for reproducibility as a requirement
The current requirement for academic publishing is peer-review, i.e.:
"Colleagues in my field looked over it briefly and said it's ok [because it doesn't challenge their work]"
This inevitably leads to whole fields of science being corrupted and biased in ways that we can't easily identify.
Science publishing standards could be improved in any number of ways: e.g. original data must be shared in an open format, publish your hypotheses before doing the research, publish the code, standardise the setting out of methods and statistics and results so that other scientists in and out of the field have a hope of following them.
Nobody in the industry is motivated to change this, and hides behind the romantic ideal of a scientist pushing the boundaries and breaking the rules to discover novel theories.
If such a scientist exists today, we can't identify them against the background of noise.
The reason it exists, is allowed to exist, and is no problem that it exists is that most science is infotainment, in that no one actually has any vested interest in it's accuracy.
It's meant to be interesting to read, but for most science no one is banking on it's accuracy to the point that they would surely notice if it weren't accurate because, say, a plane would crash or a train would derail or a rocket would explode.
Such does not happen with scientific research that forms the basis of the engineering on which many modern day applications are built. If the principles be wrong there, companies would loose substantial investment, so they are quite vested in verifying that it's accurate.
Most science is simply infotainment for other scientists: it exists for no other reason than to be entertaining and fascinating to read, and might as well be fiction.
Peer review works within a community, scientists review each other's work to decide who to trust.
But the public and funding agencies are not part of that review! Not even scientists in tangentially related fields (such as statisticians or methodology experts) are consulted.
Furthermore, "peer review" is limited to pre publication. Follow up research by others has mininal effect on the impact of a paper on a scientist's career.
So why to we believe or fund the results? Either we are not their peers and should not engage with them, or we should be reviewing their work as part of the publication process.
We also have "peers" in criminal prosecution. Would you allow a criminal trial where only the scientist's colleagues were in the jury?
Peer review works within a community, scientists review each other's work to decide who to trust. But the public and funding agencies are not part of that review! So why to we believe or fund the results? Either we are not their peers and should not engage with them, or we should be reviewing their work as part of the publication process.
We also have "peers" in criminal prosecution. Would you allow a criminal trial where only the scientist's colleagues were in the jury?
The problem is that laypeople or bureaucrats can't judge highly specialized technical contributions. It's already difficult to do if your subfield is a different niche.
> It's already difficult to do if your subfield is a different niche.
It's not that difficult, at least in science and science-adjacent fields. The scientific method still follows the same principles.
If you are used to reading papers, you can tell when someone has elided important detail, or not included the evidence for their claims - or, if someone [or a whole subfield] is using technical language to obfuscate their work.
There are many levels of analysis to a paper. Yours is part of it but not the full picture. One other thing is the novelty aspect. To assess that, you need to know the current state of the field. Also the best practices, known methodological pitfalls, judge what is a valid chain of reasoning and what is too speculative etc. It's far from simple and I often see outsiders online criticizing largely irrelevant things while missing core issues. There's no shortcut to expertise in a field. You need first hand experience and knowledge of the state of the art. Otherwise, if papers have to talk to laypeople or generic scientists, we'll see even more smokes and mirrors. The idea of peer review is that a work should be challenged by those best equipped to challenge it.
A journal could follow a set of statistical standards that could be examined and approved by experts outside of its narrow field.
I don't claim that paper reviews have to be or can be perfect - only that they could be held to a better standard than they are now.
We see plenty of avenues for publishing to a worse standard (self-publishing on an internet aggregator, for example). Let's see some avenues for a better standard.
Why isn't this a science in itself? We could trial some different standards and statistically compare them.
[It was not my claim that scientific papers must be understandable by laypeople, although I think this is an ideal case. If not, it should at least be understandable by people in an adjacent subfield - otherwise, it's just a small group of people talking amongst themselves]
Thank goodness I am retired and don't have to publish stuff in the current environment. Thirty years ago when machine learning was a sleepy backwater, life was a lot easier for us.
As a present-day researcher (or at least someone who aspires to be), I'm eagerly waiting for the hype to die down, and for ML to become a sleepy backwater again. You were lucky to play around with ML during those days.
The neat thing is, the next AI winter, we'll still have massive hardware rigs. That's very different from 30 years ago. It's a lot easier to discover new things when you can just test every possibility overnight, rather than carefully planning. So if "winter is coming," it will be a lot less harsh this time.
“The hope is that the progress in hardware will cure all software ills. However, a critical observer may observe that software manages to outgrow hardware in size and sluggishness."
Note that all the advances that eventually led to today's successes and applications of deep learning happened 20 or more years ago (e.g. LSTMs was in the '90s, the neocognitron in the '80s and so on).
Then the hardware and the data came online and the field boomed, but no progress has been made in terms of new approaches and the overall state of research has stagnated with the inevitable results discussed in the article above (I can quote at least one Turing Award winner on that if that sounds just like me bloviating). In a sense, because it's now easy enough to get good results by throwing a bunch of data at a large computer, everyone's at it and nobody is looking for a way to obtain good results without a lot of data and a big computer. Which of course is not sustainable, not least because it means only large corporations can obtain state of the art results (and to be honest, I don't think any such have been obtained by "motivated hackers" using colab).
No, my hope is that things will continue as they are now, until academics realise they've been kicked out of the game by the Big Players, and then go look for a way to compete that doesn't require gigantic amounts of compute. After all, academia can still motivate people to do actual research and find actually new things. The same motivation is much harder in industry, that only cares about one thing and only knows how to hand out one kind of reward.
The motivation behind it is very simple: eye-watering salaries in the field.
Moneyed people are ready to pay these salaries because a lot of them have succumbed to hype equating "AI" (that phrase alone is complete bullshit) to something bordering on "magic," and bringing equally magical money making opportunities.
The prospect for business owners to get rid of white collar workers in a manner not unlike how machines swept blue collar workers is just this much irresistible.
>> The motivation behind it is very simple: eye-watering salaries in the field.
For a very broad meaning of "in the field". Academic salaries are kind of meh. The eye-watering salaries you mean are paid by large technology companies and I don't think those hire you on the merit of your publications, only. For example, allegedly google and facebook et al hire drop-outs from PhD programmes in highly rated institutions etc. so people who haven't yet had the chance to change the face of AI with their research, if I may be so bold.
So, in other words, there really is not any "publish or perish" in academia. It's simply that to outcompete your fellow graduate students in getting a top job in industry (or a tenured position in academia) you need to have the most publications with the most "novelty". And that can be justified at any cost.
So, is this, at the end of the day, that the ratio of students willing to study graduate level studies, and jobs that require graduate level studies is dramatically imbalanced?
Sometimes, we as humans like to avoid the obvious discussion because it is the most uncomfortable discussion to have.
The truth is that we need more restrictions and an even higher bar on education that artificially pipes the best of the best in corresponding numbers to estimated availability of jobs in 4-5 years.
I am pretty sure that the medical world already does this.
Yes, the would be the outside solution if you were an all seeing God. But in the real world we have to make do with local incentives. And that is leading to a ballooning of academia, professors want more grad students for more papers and more prestige, some of whom also turn into profs and it snowballs. It's great while the music is playing, but then when winter comes back, what will all these people do then? Things are already insanely competitive to the degree that many profs even say they couldn't make it into a graduate program now, you already need top conference papers just to get accepted at grad school (in better groups).
In a way, I think a lot of the internal squabbles regarding diversity etc. in the AI community come from the animosity due to a shrinking cake and more hungry mouths.
As long as there's a possibility of making more money by lying and cheating, people will lie and cheat. Especially when it's a lot more money. But I don't think there's any unemployment danger among CS PhD's to worry about. Unless they somehow managed to avoid learning how to use any tools or technologies of value to anyone. PhD's do not have to be at academia or a "top job" to be employed in a research or at least R&D capacity.
By the way, tenured positions do not pay eye-watering salaries. It's still the same vicious low stakes.
> But I don't think there's any unemployment danger among CS PhD's to worry about
This is why my comment specifically addressed top positions and tenured positions i.e., a position with a specific world-renown research group. Competition for the positions that putting all that time and effort into a PhD would be for.
> employed in a research or at least R&D capacity.
Sure, jobs exist. I am not seeing where I denied this fact anywhere.
> tenured positions do not pay eye-watering salaries
Tenure is what I am specifically referring too. It's a form of intellectual and economic freedom.
So you want to restrict people from getting a phd unless they can get a top job out of it? What about people who want to do research but aren't worth the pay/responsibility of top jobs? Is your goal to increase the value of a phd by limiting it to highest quality people possible (like with MD's)?
Tenure is a benefit that is given instead of money to keep you from leaving for more money. It may be nice to have late in your career if you want to keep working forever or bizarrely want to stay at your research-focused school but focus on teaching. But money adds up nicely and has more flexibility. And like with job offers and promotions everywhere, it's something you generally can only get if they need you more than you need them.
Would there be enough people willing to go into Medical School debt without a high salary on the other end?
The entire point of game theory is to align incentives. It is not to simply paint pictures of ideals and deny that humans are human.
I understand that the AMA has incentives that are not nearly as noble as simply ensuring the world has a steady stream of doctors. Nonetheless, the answer is not to simply say "the AMA has some impure incentives" as an actual dismissal of a different argument.
The same issues exist in a ton of other academic fields though, most of which do not get crazy salaries. IMO academics are more concerned with recognition and other ego-stroking than they are with salary.
I experienced parts of CS academia and it was off-putting to say the least. Full of mediocre people who understood "the system" - not that much different from a direct-marketing operation, or a mafia - all while pestering CS professionals working in the weeds with their crazy-talk. Creating papers out of things, my colleagues and I talk while having a coffee. I'm not joking, intelligent guys get high on their own supply and will sell you ideas, that never gonna fly in the real-world with a straight face - and they will try it over and over again. It's embarassing.
Personally, I'm trying out the route of getting the resources myself to do my research, much more work - but at least sane work with a level of independence that I feel is necessary to see things through.
Not to try and dissuade this internet mob...but instead of doing the whole internet-mob-get-eyeballs-to-my-blog-by-rehashing-someone-elses-blog thing, get to basics and be precise like a scientist would be. Examples. Forensics. Details.
I blame the incentives. Academia is not about selling facts, it is about selling papers. This is very similar to newspapers; they are about selling newspapers.
I graduated in 2014, and was astonished at how much cheating there was in university. From thefts and bribery, to groups sitting in the back of the class whispering in another language, integrity at the university level has plummeted. When students openly cheat and advance, it’s no wonder that behavior becomes permissible among faculty
The level of academic unhappiness with publishing and its surrounds is actually surprising to me.
First, I never hear the counterargument.. that current publishing is OK. Does it exist, or is everyone unhappy while nothing changes?
Second, I'm surprised at how much of a single institution academia seems to be. "Publish or Perish," for example, has to be supported by tenure committees, grant makers and such. Are they all the same?
Michael Stonebraker^ suggests that tenure committees accept a limited number of papers on a resume. This is to encourage lower volume, higher quality publication. What stops some tenure committees from implementing such changes.
I just don't understand what's locking everything into place.
Well, this trend is more or less everywhere but it's not uniform, some countries have much more extreme "publish or perish" cultures than others.
For example, in Spain practically everything related to the academic career (promotions, salary decisions, tenure, etc.) is measured by pure bean counting. Actual quality doesn't matter at all, it's all about having X papers in Y journal quartile during Z period. Evaluations are conducted by scientists but they might as well be clerks, because there is usually a grading scale on everything saying "each paper in this quartile gives X points" so the weight of the actual opinions of the evaluators is zero or close to zero (even if they read one of those papers and it's total bullshit, the regulations still say that they have to give it X points). As a result, the kind of subtle fraud described in the post is widespread, with most people gaming metrics and trying to publish in the crappiest possible journals that happen to have a high impact factor.
In contrast, for example in France things seem to be more relaxed. I don't know the exact details as I'm not from there, but I have many colleagues from there and I do know that they can actually spend years thinking about very difficult problems, and then publishing a really good paper (even if it's in a journal or conference without such good metrics) and their career seems to be more or less fine. As a result, research coming from France in my field tends to generally be really high quality (IMHO) even though in Spain those people's CVs would be considered bad due to not optimizing metrics.
The US seems to be somewhere in between, China looks similar to Spain, Northern European countries seem to be somewhere between the US and France in the spectrum, and so on.
It’s not the papers per se that get you tenure — it’s the grants you bring in, and you can only get those by going through funding agencies, who are the ones that care about your publications. The grant process is like this: you write a proposal that says “I want to do this. Here’s how it will work. I’ve published these papers to support my idea. Now give me $x million dollars”. That pitch only works if there is a significant publication record to back it up.
That’s why there’s a monoculture — everyone is going through the same agencies: nsf, darpa, and a handful of other government/institutional funding sources. If you are bringing in 10s of millions of dollars worth of funding into the university with only a handful of papers, you’re going to still have a lot of support for your tenure. The flip side is that you can publish all day every day, but if you’re not bringing in grants you’re not getting tenure.
Ding-ding-ding! This is indeed the root cause of much of the pressure that drives academic researchers to publish All Of The Things. Funding agencies care very, very, very much that you have a published track record of having done something similar to whatever it is that you are proposing to do, and regularly will reject otherwise solid applications due to the PI not having sufficient relevant publications. Not saying it’s right or wrong (that’s too big of a question for an Internet comment) but it’s one of the main motivating forces behind anything that a researcher does.
> If you are bringing in 10s of millions of dollars
What's even more fun is what happens when you do bring in those funds. It doesn't go directly to your research and you will have restrictions on how you can spend the money. Some organisations take between 0% and 70% of off the top. Then, they require you to spend your money on specific areas. We wanted disk space for storing some data. They wanted to charge us >$1/GB + cost of backups to store this data and we couldn't just go to AWS or any generic cloud vendor.
The fundamental problem is that there isn't enough money in the system and there aren't enough professorships for the legions of PhD students that exist. We either need to cap the number of PhD students much more aggressively or we need to spool up many more institutes and increase funding dramatically. If you set the bar for tenure at people who are publishing in Nature or Science or prestigious journal X then people will just come up with ways to game that too.
The other thing that goes unsaid is that huge swaths of researchers in the US are immigrants on visas. If they cannot find that next postdoc they will be kicked out of the country. This creates a massive incentive to do whatever you need to do to get your next opportunity.
I am a student at University of the People. It's an online-only school which relies on volunteer instructors who are paid a small honorarium for each class. Instructor involvement varies from instructor-to-instructor, but most act mostly as moderators rather than instructors. This is due to the "peer learning" environment of the school.
In 2015, the university either hired or hosted (or had some relationship with) an instructor who was fired a month later. He was fired because he'd lied about his credentials. He claimed to have a PhD from either Stanford or MIT, but had none.
Prior to volunteering at the school, he'd worked on building up an online profile for himself. Yes, astroturfing.
He wrote a few "academic" articles on Second Life. In one, he claims one of the founders of the game as his co-author or a contributor. He "published" this article to a couple of websites and then he posted it on Wikimedia Commons. It has even been cited in real academic works.
He wrote a self-published book on Second Life which he submitted to the Library of Congress. He leverages this quite often.
He claimed to hold a "world record" on an ACM ICPC challenge. And he managed to work this into the ACM ICPC article on Italian Wikipedia which has since been removed.
He used Freebase and Wikidata to create "info boxes" about himself on search engines.
He even spent time writing fake articles promoting himself on a websites including Blasting News and IMDb. On the latter he claims to have created a commercial with Julia Roberts.
The best part is that while he was at University of the People he was also running a service to help students cheat services like Turnitin. This was something he prompted on LinkedIn, Reddit, and even Wikipedia. Yes, the traces are still there on Wikipedia.
To this day, the guy holds a grudge against the university. He posts fake reviews on TrustPilot. He posts comments on the Reddit sub such as posing as a fake recruiter or going on about the university's subsidiary in Israel. He uses a plethora of socks and IPs to edit the university's Wikipedia pages and engage in edit wars. He has also tried to scrub some of his past actions by getting articles deleted.
And this doesn't even get into his attacks on other institutions, attempt to run an Italian NGO/political party, and more. My mind continues to be blown as I dig into his activities.
> Undermining the credibility of computer science research is the best possible outcome for the field, since the institution in its current form does not deserve the credibility that it has.
This is obviously somewhat tongue in cheek, but I’m not sure the author appreciates the full implications of how this would play out. Research funding for AI exists in competition with funding for all other fields. Revealing deep-seated BS would serve mostly to decrease funding commitments to AI projects in favor of other fields, or away from R&D entirely. It would be bad for all researchers (but if the true level of fraud is that high, and I have no doubt that it is, this may simply be an inevitable market correction).
Other comments in this thread seem to miss the point as well: “why don’t universities select for quality not quantity”? Well, the change has to start with the funding sources for the unis to care.
The amount of people pouring into AI, or more specifically to Neural Network version of AI is unnatural, and it resembles a bubble in the classic economic sense.
One big problem in our society is that there is a Darwinian selection process occurring which selects for psychopathy. It's only going to get worse.
Psycopaths are attracted to power and are willing to do anything for it so they are more likely to get it. Power comes with money. Money allow psychopaths to have more children. The children are more likely to be psychopaths too so they will also end up in positions of power with more money. As automation increases, the system can afford to support an increasing number of psychopaths and they quickly take up all available positions of power via ruthless means. Psychopaths end up occupying all the important political and judicial positions. Altruistic people who follow the rules can't afford to compete in the market place, they have fewer children, they are discriminated against by psychopaths who recognize that they are different from themselves. Altruists become extinct as they are slowly replaced by machines which make altruism a redundant character trait; not necessary for a 'functioning' society. Now that we have machines to automate everything, the economy will no longer rely on 'exploitees' to do the hard value-adding work.
Then a whole new class of ultra-psychopaths (e.g. serial killers) will become more common and proceed to wipe out the regular psycopaths through violent means. Humans will get progressively worse and end up driving themselves to extinction. There's not going to be any robot uprising; humans will make themselves extinct before that happens.
Your thesis relies on the selective pressure for psychopathy being larger now than it was previously. Is that the case? It's harder nowadays to be an outright criminal or murderer and get away with it due to strength of the rule of law.
I do think that in a globalized environment, there is much stronger selective pressure for psychopathy due to higher competitive pressure and more ways to avoid consequences for misbehavior. Opportunities are fewer and bigger - Those people who are more ruthless and manipulative have a higher chance of qualifying for the few opportunities that are available because they're willing to step over ethical boundaries to get those positions. This makes them stand out as candidates.
Another point is that our modern financial system is designed to limit the liability of individuals. Corporate executives repeatedly get away with financial misconduct and their companies merely end up paying relatively small fines. Also, the government keeps bailing out big corporations; ensuring that the same people who created the problems stay in power. There is also the problem of international money laundering; wealthy people fleeing one country which considers them to be criminals to move to another country which doesn't consider them to be criminals. Similar story with bankruptcies; people changing countries to escape consequences.
I see opportunities tending towards the most competent foremost, with psychopathy potentially helping but in many cases hurting. I've seen many people fired for being assholes, as an example of cases where these traits have hurt. All the psychopaths I've met have had pretty pathetic and lonely lives. People figure them out and they get sidelined.
Politicians, quantum healers, fraudster psychics, snake oil salesmen, etc, are rare niches where psychopaths genuinely thrive over regular people.
5000 years ago you pretty much had to be willing to murder someone to get what you wanted and survive. Nowadays such ruthlessness is punished in many contexts more than rewarded. The corporate ruthlessness that we see today is such a tamed beast in comparison.
So I'm not convinced that the selection for psychopathy has increased.
Only the least intelligent psychopaths get caught. The really smart ones are masters of PR; they can make themselves look more altruistic than actual altruists. They only show their true face once they no longer have a use for you.
They are excellent actors. If the best manipulator in your local community is talented, think of how talented the best manipulator in your country is... Then compare that to how talented the best manipulator in world is. That is the danger of globalization.
World-class psychopaths are as intelligent as other world-class individuals (very intelligent) but the difference is that almost all of their intelligence goes into presentation and manipulation - Very little goes into the actual craft that they are supposedly experts in. But from the perspective of non-experts, they come across as smarter than the actual experts.
It's true that smart psychopaths can do way outsized damage in today's world. The flip side is also true though, the genuinely good and competent can also have an outsized positive impact.
I think this speaks more to the scale people can access through modern society (technology, capital, large orgs, and so on).
Well, the unfortunate part is that he's at least somewhat correct, and it's been found that psychopaths are much less likely to seek/adhere to mental health treatment such as SSRIs.
> By partaking in a form of fraud that has left the Overton window of acceptability, the researchers in the collusion ring have finally succeeded in forcing the community to acknowledge its blind spot. For the first time, researchers reading conference proceedings will be forced to wonder: does this work truly merit my attention? Or is its publication simply the result of fraud?
But I don't see how this follows. If I follow the link to the description of the actual fraud ( https://cacm.acm.org/magazines/2021/6/252840-collusion-rings... ), it says essentially the opposite: the "fraudulent" papers are no different from papers published by ordinary means.
> the review process is notoriously random. In a well-publicized case in 2014, organizers of the Neural Information Processing Systems Conference formed two independent program committees and had 10% of submissions reviewed by both. The result was that almost 60% of papers accepted by one program committee were rejected by the other, suggesting that the fate of many papers is determined by the specifics of the reviewers selected
> In response, some authors have adopted paper-quality-independent interventions to increase their odds of getting papers accepted. That is, they are cheating.
> Here is an account of one type of cheating that I am aware of: a collusion ring.
> A group of colluding authors writes and submits papers to the conference.
> The colluders share, amongst themselves, the titles of each other's papers, violating the tenet of blind reviewing
> The colluders hide conflicts of interest, then bid to review these papers, sometimes from duplicate accounts, in an attempt to be assigned to these papers as reviewers.
> The colluders write very positive reviews of these papers
So the system is: conferences already can't tell the difference between a good paper and a bad paper. Researchers respond by adopting strategies for passing review that are irrelevant to paper quality (since paper quality doesn't count). But those strategies aren't bad for paper quality. If I'm reading conference papers, why would I worry about whether one of them is the product of review collusion?
> If I'm reading conference papers, why would I worry about whether one of them is the product of review collusion?
Because the one you are reading may have crowded out a better one. Even if the current review system is essentially random, replacing it with something that is essentially a contest of well-connectedness is worse. Young researchers with good ideas but fewer connections, or people from less well-known institutions would have their ideas suppressed.
So you should be worrying about stagnation, and about not reading what might actually be new and exciting.
None of that reflects negatively on the paper. There is no additional caution warranted when reading papers. That's just a question of "are you happy with the state of the world?". You can think about that question any time.
I think it does. Good papers are written by people who care about doing good research, whose primary motivation is to do good research. Such people will not accept to collude or commit academic fraud to get their papers published, because they are idealistic and are in research because they want to do useful work. To such people, committing fraud is anathema, for personal reasons that have nothing to do with economic or other incentives.
There are such people in academia but they are also crowded out, to borrow tom_mellior's turn of phrase, from others, who don't hesitate to commit academic fraud to get published and who don't give two flying figs about the quality of their own work. This is obviously a concerning state of affairs that can only be detrimental to the overall quality of research.
So I'm sorry but you're dismissing the issue out of hand without having thought about all the consequences. Academic fraud is like, I don't know, broken windows? It just perverts everything around it and creates a black hole of bullshit that sucks everything in it. Good research cannot thrive in such conditions.
Actually, today an article was posted to HN (not by me) that points to Basquiat's parable of the broken window:
The parable of the broken window was introduced by French economist Frédéric Bastiat in his 1850 essay "Ce qu'on voit et ce qu'on ne voit pas" ("That Which We See and That Which We Do Not See") to illustrate why destruction, and the money spent to recover from destruction, is not actually a net benefit to society.
In microeconomic theory, opportunity cost is the loss of the benefit that could have been enjoyed if the best alternative choice was chosen instead.[1]
The author is saying that all of the papers are bullshit, garbage, and a waste of time. He is saying that everyone is turning a blind eye to this. He is saying that now that the fraud has got this bad, it is out in the open, and people will be forced to deal with the fact that the entire field is filled with garbage.
That's right. Btw, I think the author is wrong. This is not "out in the open" now. It's been "out in the open" since for ever. And those who act all surprised about it are probably the ones who knew it better than most.
I don't mean the author, of course, but he is pointing to some of his own papers as bullshit. Why did he publish bullshit papers if he knew they were bullshit? Or doese he mean those papers are bullshit with hindsight? That, I can understand- "I wrote this paper ten years ago and reading it now, I cringe". Sure, that happens and it's only evidence of the person's progress. But to say that one's research was bullshit in the context of an article that describes academic fraud that leads to the publication of bullshit? That I don't get. Surely the author is not confessing to committing academic fraud himself! I didn't get that from the footnote anyway.
When I published those papers, I was new to the field. I was guided by the standards set by the community, by my mentors, by my peers. I was proud of each of those papers. At publication time, I believed I was doing good science, and the belief was re-affirmed by acceptance at top conferences.
My thinking has evolved since then. The community norms have not.
Oh, er, sorry for the harsh words, anyway I hope they didn't come across as too harsh.
So I think you're saying the papers you linked to where "bullshit with hindsight", not that they were unethical at the time. Or at least you didn't think they were.
>> If I'm reading conference papers, why would I worry about whether one of them is the product of review collusion?
Well, because if there's no incentive to write a good paper, very few good papers will ever get written and you'll waste your time reading crap papers.
I get what you say: if the process is random anyway, then what's the problem? But the process is not random: authors maximise their chances to publish their paper if they partake in a collusion ring. Partaking in a collusion ring diverts resources from writing a good paper, therefore the quality of published papers goes down the drain. The most important resource necessary to write a good paper of course is the motivation to not publish crap. If that goes, everything else follows.
The lies in science and academia are out of control. Whether its this example, the replication crisis, the funding crisis - the upshot is that we cannot be confident about the 'reality' that is being presented to us.
The model of academia works well when some part of society is willing to fund researchers spending time on open-ended basic research, even if it produces no results. Saying "Don't engage in unethical behavior" is well and good, but if the incentives are towards unethical behavior, you shouldn't expect very much.
Academic funding and employment is very closely linked to the number of published papers. The article says that a grad student who publishes three papers a year is a professor's dream - but on reflection, it should be clear that this is not because three papers a year is necessarily good science, it's just a good metric. The professor's lab looks good for publishing so much, the professor can ask for more funding from grant agencies, and the student is likely to get a good faculty job which reflects well on the professor.
So it's extraordinarily difficult to stay in the field, have a well-funded lab, hire enough grad students, and get tenure if you're optimizing for the quality of your science over the number of publications. The "publish or perish" culture (literally, in at least one case, it seems) isn't driven by the practitioners; it's driven by the requirements of getting grants and getting academic jobs in an increasingly competitive market.
Now in some fields you do need the facilities from a university to do your research, but for CS in general - and especially for AI - you'll be just fine in industry. All you need is a pile of cloud VMs, and industry can get you that. Perhaps you also need interesting data and real-world problems; industry can get you that too.
Society today underfunds academia and overfunds industry. You, the individual researcher, are not going to be able to fix this. Go where the incentives are better aligned for you.
(And if you're a Ph.D. student who sees no future in your life after being complicit in fraud, please, please reach out to industry. Your prior publications are nowhere near as high-stakes as in an academic job; you can contact the journal, get it retracted, drop out of the program, and have a great life ahead of you. What saddens me is that the student who allegedly took his own life was probably so deeply surrounded by academia that he didn't know there's a world who won't judge him for having one fewer paper on his CV and will even look on him positively for reporting fraud and getting his advisor into career trouble.)
---
That said, my argument is undermined a bit because the papers OP denounces have coauthors in industry. So I don't have an answer for why those coauthors went along with it.
I find it quite ironic that every time some unethical academic behaviour is published on HN, there's several posts about "move to industry this never happens here". I mean this is HN, a significant portion of the industry is working on optimizing for clicks and keeping people engaged (addicted?) to some platform or the other. Let's not even talk about the blatantly unethical (and often times illegal) things that happen in the startup world to increase evaluations, secure funding etc.
Now academia has lots of issues, but to say industry is better... It's sort of like telling an amateur athlete who is upset by some competitors using caffeine (or amphetamines) to instead to professional sport, because they are more ethical. The stakes (and money) are much higher so people are more willing to cheat to gain an edge.
Sure, but as an individual researcher, the integrity of your scientific work is not compromised by the unethical goals of your employer or the unethical ways in which it raises funding. That's what I'm claiming. It's rather different from amateur vs. professional athletics, where both have the same goal of "run as fast as possible but within some nebulous concept of natural human ability": the goals are different between academia and industry.
Or put another way: Of the many deep moral questions raised by the Manhattan Project, not a single one was "Did they commit academic fraud and claim that an atomic bomb was scientifically possible when it wasn't?" They were employed to actually get the job done, not to act as if a job were getting done.
And at the end of the day, funding for university AI labs is largely driven by the existence of those same unethical industry goals - governments fund the work because it's good for the economy, students pay to take AI classes in their undergraduate degrees because it's an investment in their future careers, the industry donates everything from fellowships to entire buildings to academia, etc. I don't think you can cleanly wash your hands of industry's ethical concerns by staying in academia but working on the problems that industry finds interesting. If your goal is employment without any ethical concerns, you're going to have a very hard time.
(In the case of this particular post, the co-authorship of papers with industry makes it clear that the research directly benefits industry.)
And how is an individual researcher funded, if I may ask? We already have many homeless teachers living in their cars, so I don't think the power of public opinion will rebalance financial incentives to make an even more useless profession (in the eyes of the general public) attractive. Joe-Schmoe-the-janitor won't see financially supporting lone nerds as a first priority.
Huh, I thought I was pretty clear that I meant an individual researcher employed by industry to do their research ("for CS in general - and especially for AI - you'll be just fine in industry.... Go where the incentives are better aligned for you.") - was it not?
So, limiting research to projects small enough for a single individual, then? Because similar situations will and do arise in industrial research groups.
Ah, I see the confusion. By "individual researcher" I do not mean "independent researcher" - I mean the researcher, considering their motivations as an individual person. They can be part of a group of researchers, and they're generally working for some employer. After all, the "individual researcher" in academia is generally working with a group - e.g., the example given about the grad student who is "every professor's dream" only makes sense in the context of that relationship - and that produces the negative pressures described.
In particular, I mean that a researcher in academia, as a person (an "individual researcher"), is motivated by the demands of academia to get grants and fill their CV and is therefore incentivized to conduct dishonest science to make that happen, and a researcher in industry, as a person (an "individual researcher"), is not directly incentivized to conduct dishonest science - perhaps there's dishonesty in how their employer gained funding or what they do with the research, but that doesn't compromise the accuracy of their research, motivate them to game the peer review system, etc. The researcher as an individual has the choice about whether to be in academia or industry.
So, I don't think I follow how similar situations will arise in industrial research groups. (Though, as mentioned in my original comment, I'm probably missing something, because there were researchers from industry who coauthored these papers.) Even among a group of researchers in industry, the incentives should be to produce things of value to the employer, not to play the part of productive-looking researchers.
I'm specifically not claiming that independence will solve anything; the fundamental problem is funding, and (as you say) nobody is going to want to live out of their car to do good research. And you need access to facilities/tools of some sort to do your research; my claim is that industry can provide those at least as well as academia can, not that they are unneeded. Admit that you're constrained to work at a place that can fund your research and that no place exists that will pay you well and leave you to your own devices, and then find the place whose incentives to fund you are least likely to compromise your research integrity and most likely to reward you for actual good work. At the moment, at least in the society where I live, that happens to be industry.
> What saddens me is that the student who allegedly took his own life was probably so deeply surrounded by academia that he didn't know there's a world who won't judge him for having one fewer paper on his CV and will even look on him positively for reporting fraud and getting his advisor into career trouble.
Could be that, or could just be that the student couldn't imagine himself not doing academic research despite knowing of the rest of the world.
It's sad nevertheless, of course, and I'm not trying to say you're wrong in any way. It's just that people's motives and the reasons they feel trapped can be difficult to guess.
>> All you need is a pile of cloud VMs, and industry can get you that.
That is true if by "AI research" you mean "applications", as opposed to contributing anything new. To contribute new knowledge what is needed is time and the resoruces to study the work of those who have contributed new knowledge before you, and to become an expert in that knowledge. In theory, that's half the job of a PhD student. The other half is creating new knowledge. Beating benchmarks is not the job of a researcher, despite what's the norm in machine learning research these days.
>> So I don't have an answer for why those coauthors went along with it.
Because the motivation of researchers in industry is money, not knowledge. Academia is capable of motivating researchers to create new knowledge. It is also capable of motivating them to commit academic fraud. But industry only understands one motive and can only offer one reward.
Basically, if your university pushes you to publish or perish, you might publish bullshit, or perish, or find yourself a niche where you can publish something that isn't bullshit. Many researchers do this, but of course you don't hear about them because they're not in the news. In fact, you probably hadn't heard about Geoff Hinton, Yoshua Bengio, Yan Le Cunn and You-Again Schmidhuber before the big boom of deep learning, because at the time they were just such small-fry researchers (compared to what they are today- I once had a look at Hinton's publication record and he was anything but small fry academically, even in the 1980's, when the lore says he was roaming the academic wilderness isolated from the mainstream; it's all bollocks). For sure, no Google was throwing millions at them at the time and the machine learning community as a whole had more or less given up on neural networks, or rather the majority of machine learning researchers where happy to leave Hinton et al and Schmidhuber to work on neural nets, while everyone else was working on Bayes nets, then decision trees, then SVMs, etc.
I digress, but what I'm trying to say is that "publish or perish" is the norm for researchers who are not motivated enough, for their own personal reasons, to make meaningful contributions. They are not the only kind of researcher and when progress happens, it comes from the other kind. But researchers in industry only have one motive, because industry offers only one incentive and is only interested in results that satisfy that incentive- regardless of how those results are brought about.
If the noble of heart can get good work done in academia despite the pressures on them, why can they not get good work done in industry despite the pressures on them?
Because the "pressures" in industry is that you get fired if you don't do the work you're hired to do. There's still a modicum of "academic freedom" in academia, the freedom to pursue your research interests, whatever those may be. In industry, you pursue your employer's interests and if they are not also your interestes, tough.
Speaking in this with the experience of working in the industry for a few years, then going into academia, precisely because I got bored doing other peoples' work and I wanted to do what I'm interested in.
Depends on what industry you're talking about. Those that publish are working according to the same practices. Google, Facebook et al. need to pass peer review too and they care about papers as a dick measuring contest.
The author mentions a collusion ring in AI, but I'm pretty sure the article refers to a situation that arose in the computer architecture community (which itself happened quite a while back).
There's a lot of discussion here about changing the incentive structure of academia to prevent this kind of fraud, but based on the discussion I've read so far I don't think a lot of people here understand the exact incentives that are involved, so I'd like to try to explain how academics actually operate.
We've all head of "publish or perish", and there is a general understanding that you have to publish papers to get tenure, so that's where the incentive is. But it's a little more complicated than that.
When you are hired as a tenure-track assistant professor at a university, you're given what's called a "startup" package. Assistant professors are essentially startup founders where their company is a lab. Your job as an assistant professor is to use that money to purchase equipment and to recruit PhD students (aka employees) so that you can launch your research agenda. This money pays for:
- rent taken off the top (essentially paying for your lab space and other costs the University incurs)
- a portion of your own salary
- your students' tuition
- your students' stipends
- conference travel and fees
- lab equipment (computers, desks, chairs, machinery, scientific apparatuses). To note here, for every dollar you spend on equipment, the University takes a percentage of that. So if I buy a chair for $100, the University will charge for example $10. This is a sort of tax for being affiliated with the University and getting free use of their resources (internet, libraries, subscriptions, etc.).
- As well as other things.
Crucially, this money is not free. The expectation is that you will in the next 6 or so years use that money to jumpstart a successful lab operation. A "successful" lab is one that is able to procure significant funding from government and institutional sources. Your first goal is to procure enough funding to recoup the initial startup package the University laid out. That's your bare minimum. Ideally, you want to procure funding far in excess of this by the time you're up for tenure, to prove to the University that your lab will bring in more dollars. Because you've got to earn your keep.
Notice there hasn't been much discussion of papers so far. The initial startup package can be quite significant -- on the order of millions of dollars depending on the research agenda. As an assistant professor, you need to bring in millions of dollars from funding agencies like the NSF to make that back in multiples. But the NSF budget was only $8 billion dollars in 2020, and they need to spread that money around to new assistant professors across the entire country. Therefore, funding rates usually range from 10-20% for a proposal, which is quite low when your entire future is depending on it.
This is where "publish or perish" comes in. When you are up for tenure, yes they'll be looking at the quantity and quality of your publications. However, you could have 1000 high quality publications, but if you don't bring in significant grant money, you will not be getting tenure. The way you get that grant money is by writing a good research grant proposal, and the way you back up and bolster those proposals is through publications.
This is why there is so much fraud in publications. Not getting tenure at this stage means your research career is essentially over. You are as good as fired from your current role, and few Universities will take you on as a tenure track professor when you've already failed at your current institution. At this point you're in your mid to late 30s, and your best years (in terms of research ideas) are already behind you. It's time to enter industry or become a lecturer.
So how to fix this?
1. More research funding for government agencies. Higher funding rates means less incentive to commit fraud.
2. A feedback loop of profits from industry to universities. Right now when technological advancements happen at universities, those flow out to the general public, then to companies, who figure out how to monetize them. The profits generated by that monetization are captured entirely by private corporations, even though the foundational innovations on which they are based are funded by Universities and by extension the public. If some of this profit were diverted back into Universities (maybe through increased corporate taxes), then this would help.
3. Lessen the incentive for Universities to essentially become hedge funds with schools attached. My University is currently sitting on a billion dollars that is locked up in investments. Harvard's endowment is currently 40 billion dollars. They could fund the entire country's research agenda 5x over with that. Yet professors are paying their own salaries with public money? And the University gets a cut? This is insane.
4. Offer an alternative off-ramp for academics who don't get tenure. If not getting tenure won't end your career you've spent your entire life building, then the incentive to commit fraud would be lessened as well.
Anyway, tldr; there's not enough money to go around, so fraud exists.
I agree with everything you said. At the risk of being pedantic:
> A feedback loop of profits from industry to universities. Right now when technological advancements happen at universities, those flow out to the general public, then to companies, who figure out how to monetize them. The profits generated by that monetization are captured entirely by private corporations, even though the foundational innovations on which they are based are funded by Universities and by extension the public. If some of this profit were diverted back into Universities (maybe through increased corporate taxes), then this would help.
This isn't ENTIRELY true. Universities "own" research IP and they do have tech transfer offices which at least theoretically help them monetize their largest breakthroughs via researcher startups. Many large, well established labs will also perform collaborative industry research, effectively acting as a CRO, which they can charge a lot of money for (and the Universities can subsequently take overhead from). The huge, field leading labs with well known PIs will have multiple R01s and various grants but the vast majority of their funding will come from industry.
Yes, this is true. But I was thinking more along the lines of fundamental research. For example, DARPA spent a lot of money investing in transistor research in the 50s, which begat microcontrollers, which begat companies like MS, Apple, and Google. Those companies together are worth multiple trillions of dollars, yet wouldn't exist but for the publicly funded sector research that preceded them. Harvard's $40 billion looks piddly in comparison. And what kind of research does MS/Google/Facebook fund in the University? They'd like us to figure out how to optimize ad placements on their search engines and social media sites. I've seen far too many promising researchers swallowed by that kind of work, only to quit after realizing the direct application of their research is simply to make those companies more profitable while they are paid $25k per year and eating ramen 3x a day.
We see the same thing happening right now in robotics. DARPA spent a lot of money jumpstarting driverless cars with the DARPA grand challenge in 2005. The researchers I know who were involved with that have left academia and are in the private sector now because that's where all the money is. There's no money flowing back in the academic direction even though there's a direct line from what they were doing in 2005 to what Toyota, GM, Google, Tesla, etc. are doing today.
I'm not saying they shouldn't make profits and be worth a lot of money. They provide a lot of value. But the outcome definitely seems lopsided to me.
Yes, as someone in the system, it's all true and an open secret, everyone knows it and we complain about it over some beers all the time.
But isn't something like this true everywhere?
And blaming the individual is futile. I like Scott Alexander's Moloch [0] concept more useful. If you apply a high standard against your own research regarding thoroughness, not overclaiming etc. never looking at the test set etc., you won't beat the benchmark, which contains entries you know are too good to be true. But if you don't beat them, you'll get rejected. The pace is extreme, lots of groups work on very similar things and it's a race to publish very similar stuff first. As a PhD student you need publications. You are in your late twenties, early thirties, you can't afford to lose several years.
The big lie is that there is so much discovery happening as suggested by the thousand upon thousands of papers at big conferences, most of which is never really engaged later on. We have to pretend that all grad students in every group can contributing some actual valuable, bulletproof novel thing to science every few months and pretend that you can have multiple papers at conferences while remaining painfully critical towards your own work. I mean how could it be that people happen to always find something that works out in the end? How can people sit into a project at the start of the PhD and immediately crank out a valuable novel contribution? It isn't realistic, but we have to pretend because this is the background assumption behind a PhD. But the current state of affairs is inflating away the value and prestige of a PhD. It means less and less because everyone knows that it's not just a measure of scientific research skill and knowledge.
However, what people are suggesting, namely making it less based on peer review and more on some social media like system is also misguided. It's like the instagrammization of the field: the sexiness, the dopamine hit, the celebrities.
But this isn't specific to AI or CS or academia. This is how humans are. It's how status hierarchies are always gamed. Think office politics, actual politics, dating.
Atvthe same time actual AI methods are really getting better, it's not a lie and not just hype. It's actually quite paradoxical how well the system works despite the flaws. Vision algorithms work really damn well, its very obvious to anyone who remembers the pre-2010 times. Perhaps there is somewhat of a stagnation in the last few years but also a lot is happening in smaller niches, as the big deep leaning ideas are adapted and applied specifically to particular tasks. The methods work, but the literature is distorted. You need to be an insider to know what to actually believe.
> everyone knows it and we complain about it over some beers all the time.
Then why not speak out publicly against it?
Obviously because you know that there are still a significant number of individuals that want the system to work this way. Your fear of those individuals is why you don't speak out against it.
Well, I'm not talking about outright fraud but the more subtler forms mentioned in the article. I think it's probably harder for outsiders to understand the nuance.
It's not "who are they", its everyone. The incentives are set up wrong. Blaming individuals is insufficient. It's like dreaming if only politicians stopped lying. Academia is a hierarchy and a power system. People have interests, reputation, prestige and it often directly translates to money.
Again, I'm not talking about blatant outright lies or fudging numbers.
Thing is, after some rejections or sabotaging yourself a few times by being too critical with your own research and not having publications you will imitate what you see around you or you drop out due to a lack of publications. It's rewarded to be less critical with yourself. And ultimately you want to graduate after all that sunk cost. Even switching to industry isn't so easy without publications in prestigious venues to show for those years.
> I think it's probably harder for outsiders to understand the nuance.
No, it's simple. It's called game theory. So, who benefits from the publish or perish system that drives this profoundly fraudulent behavior and outcomes in science?
You are avoiding the "who" question. Why not maybe, perhaps, identify the who and get them removed from their position?
It may or may not even be a fellow scientist.
> lack of publications
uh huh, science is not and never was a stationary process.
It's absolutely disgusting that modern academia is filled with so many cowards and buffoons that the state of modern science has reached such a low.
> Even switching to industry isn't so easy without publications
Well, this is stupid. The only thing useful to corporations is your training and not your novelty.
Any novelty is a cherry on top, cream-of-the-crop pick and has nothing to do with being "hirable". But, that is also just an unchangeable aspect of our competitive reality and is thus entirely irrelevant to this discussion.
I disagree that this is for the lizardmen, the leftists, the woke or whoever. No, it's us all. We need publications because otherwise research is hard to evaluate for outsiders. If it's evaluated by peers, it's politics again.
There's a generic problem in academia and there is a specific one in the explosively growing and lucrative ones like AI. We can't think clearly if we blur all sorts of complaints against academia into a generic complaint against elites or something. Publish or perish also comes from the desire to measure, quantify things objectively etc. It's a generic trend in the management community. It's easier to hide behind mechanistic procedures, we live the age of the bureaucrat and not of the strong leader, for various historic reasons. Of the doctrine of equality, not of Great Men theories. But it all leads too far.
I recommend the SSC articles on Moloch for thinking deeper about this.
Why do parents chase extracurriculars and test drilling for their kid? Because it's competitive. With lower stakes, backwater academia can have lower pressure amd perhaps more attention and care.
We cannot abstract away the human. This is not a mere technical problem of a problem of morals in a few individuals. Read some novels, drama etc. We in the technical fields would benefit a lot from learning more from the humanities, literature, even classics from Antiquity. Man has not changed, our nature is still rich and complex and the same lesson that power corrupts.
Re training: outside the US, PhD programs have no explicit training. Training stops at the master level. From day one you just work on methods in the hope to publish the results. Advisors usually don't have a lot of time for each individual PhD student. Maybe let's call it experience, not training.
One surprising discovery about AI: There's nothing inherently wrong with cherrypicking. I used to think it was this dirty thing. Oh, you're only showing your best results?
But the best results are what matters. If you can create an amazing song from scratch by telling a computer what to do, you only need to do it once. The song is still good.
And it's easy to automate. If you have a way of detecting a good song, or at least filtering out rubbish, then you can generate thousands of attempts to get the good result.
I like to say "It's hard to pick cherries from a rotten tree."
The flipside of this post, though, really hits home. I recently was super excited about a paper called FNet, which showed that "fourier transforms can be competitive with transformers", i.e. just replace multihead attention with a fourier transform. 7x faster on a GPU! Woo!
In other words, the paper was ... advertised in a misleading way, to put it kindly. Most of the transformers they compared with aren't the traditional multihead GPT style transformer that people think of when they hear "transformer". And the one that was (BERT) totally annihilated them.
But it's also easy to jump to conclusions too quickly. I've been a hothead in the past, and called out a paper when it turned out that I just didn't understand enough. It's tricky to know for sure.
Not the kind of cherry-picking condemned here, which is more like your drug killing 90% of the animals it was tested on, and making a small difference in 2% of them, and you present it as only that 2% ever happened (and thus represent 100% of the cases).
Trying that shiny new algorithm out on a couple dozen seeds, and then only reporting the best few. Running a big hyperparameter sweep on your proposed approach but using the defaults for the baseline. Cherry-picking examples where your model looks good, or cherry-picking whole datasets to test on, where you’ve confirmed your model’s advantage. Making up new problem settings, new datasets, new objectives in order to claim victory on an empty playing field. Proclaiming that your work is a “promising first step” in your introduction, despite being fully aware that nobody will ever build on it.
Yeah, I misread; sorry. Testing your model on a small dataset and presenting it as a general solution is all too common.
That's why generative models are cooler than classifiers, to me at least. You can show the outputs visually, whether it's text or image or sound. But with classifiers, you're chasing an accuracy rating. I forgot that people often test on CIFAR-10 (smol data) without verifying on Imagenet (big-ish data) and then present their paper as very general.
But, I do have something to say about that kind of dataset cherrypicking: Researchers often find it hard to test on large datasets because of limited compute resources. One of the pioneers of DDPM wasn't able to train on imagenet due to lack of GPUs. CIFAR-10 was the best they could do. So it's sometimes hard to tell whether there's intentional deceit, or just a shoestring budget. (I was surprised how much research a lot of people get done in spite of limitations.)
> . If you have a way of detecting a good song, or at least filtering out rubbish, then you can generate thousands of attempts to get the good result.
So often the filtering technique is a human in the loop, manually going through the output. Humans become the hidden, undifferentiatable and undisclosed loss function.
Sure is! And you'd often be amazed at just how much data you can churn through, if you put your mind to it. I once watched Gwern classify thousands of poetry outputs over the course of a day or two, with nothing more than dedication and a bash script to mark "a" or "b".
It took a long time to accept that this is both normal and effective. It's not some hidden secret to be ashamed of. The headline images in https://openai.com/blog/dall-e/ are probably the best and most interesting images they could find; they were picked out by humans. It's a positive statement to say that dall-e can do that some of the time, rather than "it fails X% of the time."
Scientific community tainted itself by producing fake medical cannabis "research" as they could only get funding if their "studies" could prove negative effects.
The thing that can happen: fields gradually split into rigorous and non-rigorous camps. Like with evidence-based medicine, or chemistry/alchemy. Depending on the field, either might prevail in the market. Medical research and bridge-building are mostly rigorous, programming is mostly non-rigorous.
AI/ML has a range of rigor levels, from fairly good to total crap. I think people in the field have a reasonable idea which is which. It's frustrating to people outside the field that think they can just take the technique with highest reported performance numbers and expect good results.