Several years ago, MIT Professor of Marketing Drazen Prelec proposed a "mathematical truth serum" which he claimed could help government agencies predict future terrorist events. MIT and a local defense contractor used his work to win a $12 million grant from the intelligence community. The only justification provided for using the mathematical truth serum (a trivial application of KL-divergence) was that it would "promote honesty" among intelligence analysts who may hold divergent opinions.
Unfortunately, honesty on the part of MIT and its defense contracting partner was in short supply during the bid and proposal process. Federal prosecutors were tipped off shortly after the program was funded, prompting the government to swiftly cancel the contract.
I think this is a fairly inaccurate mischaracterization of Prelec's work and BTS. If you search "bayesian truth serum" on Google scholar, you'll find that there are hundreds of papers from credible researchers based or inspired by Prelec's work. Just because the math is simple does not mean Prelec shouldn't be credited with having introduced it to economics and mechanism design. Also the core BTS papers were published in Science and AAAI, so they are pretty mainstream.
I hadn't heard of the deal/fraud you're describing and am curious to learn more, but you should include some citations before trashing influential, interesting work.
Just because the math is simple does not mean Prelec shouldn't be credited with having introduced it to economics and mechanism design.
No one is saying that simple mathematics cannot have profound applications to the real world. One need look no further than Claude Shannon's fundamental work in information theory for an universally acknowledged example of this phenomenon.
Also the core BTS papers were published in Science and AAAI, so they are pretty mainstream.
That really doesn't prove anything either. Academia is full of brown-nosing "I'll cite your paper if you cite mine" types. After all, Prelec claims that one of his motivations for introducing a "truth serum" is that consensus opinion can't be trusted, because sometimes the experts have a reason to lie. Has he ever run his own research through the serum to see what comes out? And more importantly, would anyone else really care?
I hadn't heard of the deal/fraud you're describing and am curious to learn more ...
Don't really have that much more to add except that from what I understand, Prelec knowingly lent his name (and by extension, MIT's name) to a grant proposal that to a grant proposal that deliberately misled taxpayers.
The interesting question of course is why would Prelec do something this foolish after securing tenure at the MIT Sloan School of Management?
It is true that there is a lot of pressure on academics to raise money to help support their bloated university administrations, even after receiving tenure. Perhaps this was part of the reason why he allowed himself to become a willing tool for his partners in the defense contracting industry.
It is also possible, however, that Prelec was eager to move beyond the world of art critics and sexual promiscuity surveys into something with more prestige by associating his name, for example, with this country's national intelligence counter-terrorism program ... perhaps motivated in part by professional narcissism and/or delusions of grandeur.
Paper from Science, worth reading ...
Thanks for sharing the link, but including a list of the most significant applications of Prelec's theory over the past 15 years would be even better!
You should check out his more recent work. Your reply is pretty ironic given that his focus has been studying the prevalence of "questionable research practices" and how to detect and avoid them.
The first paper you mention (on measuring the prevalence of questionable research practices with incentives for truth telling) was especially timely since 2012 was one year after Drazen's program with the defense contracting partner was reported to federal prosecutors for misconduct.
By the way, not at all surprised that the paper found such a high rate of dishonesty in psychological research. After all, Drazen earned his PhD in the same department at Harvard University where now-disgraced evolutionary biologist Marc Hauser was forced to resign after admitting to fabricating data, manipulating experimental results, and publishing falsified findings.
The defense contractor considered hiring Hauser as a consultant on a different problem not long before it was reported in the news that his office had been raided by authorities:
The defense contractor unfortunately didn't act fast enough, because based on an investigation into that program, Hauser would have probably fit right in! Ironic too that Hauser was also the author of a popular best-seller entitled Moral Minds: The Nature of Right and Wrong :)
At the time of the raid, Hauser was already well known to students on the Harvard campus on account of his support for this salacious Harvard first:
That's why I included a reference to his job title here. When this country's intelligence agencies are looking to "professors of marketing" for new ideas to reduce terrorism, this is exactly what they should expect to land in their taxpayer-installed stainless steel toilets.
Apparently no one on the government side ever received the memo about the "replication crisis" that afflicts so much of what is passed off today as "research" in the field of psychology.
In which case, the gullible suckers on the government side provided the perfect audience for this snake-oil selling salesman. The fact that MIT and the defense contractor were only guilty of "marketeer-ing" here (rather than racketeering ... think Jeffrey Epstein, the underage girls that were trafficked to MIT professors, and the thank-you notes from the MIT administration) is probably why this "bad act" on the part of MIT and the defense contractor received far less media attention.
I'm an anesthesiologist so I frequently inject people with hypnotics and sedatives.
There's definitely a sweet spot where people will say things they probably otherwise wouldn't. Much like alcohol I guess but perhaps a bit more potent.
I had a spontaneous pneumothorax a couple times in highschool, so I ended up having a lung surgery to prevent them when I was 18. To this day my mom refuses to reveal the things I said to her while I was in recovery, and the following days when I was on Demerol. It drives me crazy because there is a lot I could have said, did I mention I was 18? I know she's probably doing me a favor, but it seems unfair.
I've since wondered if I have a big mouth when I'm dosed because that's who I really am underneath my impulse control, or if that's how everyone is.
Of the people I'm worried about hearing my confessions or fantasies, people with whom I exclusively have a doctor/patient relationship are the bottom of the list. It's the early visitors I receive in recovery that scare me :)
I came out of knee surgery and after waking up, my first sight was of two female nurses just staring at me hard. I have no idea what I might have said and I don't want to know.
They know the drill. Sounds like they were just being jerks.
It won't surprise me if we start seeing videos on Twitter of post-op people saying offensive things. Followed by the requisite mobbing, cancellation, etc.
There was a spy in the USSR during the cold war who had to have some procedure done. He refused anesthesia. His pretense was that he might reveal state secrets or something. He was given an award for his bravery or something. Interesting story. I wish I could remember more of the details.
Years ago at university I participated as a subject in a memory study that involved the injection of scopolamine. I was familiar with it's reputation as a truth serum and very curious to feel the effect.
But it's utter nonsense. There was little more than an awfully dry mouth and a short-term memory so impaired I felt stuporous.
In my opinion any euphoriant will elicit more truth telling than scopolamine.
This in itself is really beneficial as a truth serum. By enquiring multiple times the victim can't remember their previous answer, so therefore reveals which parts of their story are true and which are fictitious.
There we have to distinguish between dishonesty and a lie. A deliberate construction on the spot might not survive multiple angles, this is the basis of interrogation rooms after all, anyone will trip themselves up repeating versions over and over again.
What won't necessarily find the light of day is deception, a pattern of behavior. A drug dealer might get caught up in where he was or who he was with at a given time on a given date, but he'd have to be intellectually impaired to admit he was a professional criminal for a living.
>As with alcohol, the results of hypnosis were never completely trusted; Sigmund Freud, for example, used the technique to uncover patients’ repressed memories of sexual abuse but eventually admitted that most of the recollections were fantasies. “Such widespread perversion against children is scarcely probable,” he concluded.
I'd like to know the exact questions asked to determine whether someone was sexually abused or not, but I have a hard time finding the studies cited in your link. As I interpret the text from your link, a teenager voluntarily sending nudes to their partner would count as sexually abused because they participated in the creation of child pornography.
Although that might be part of it, I don’t find 28% to be a surprise.
One reason is that I’ve been told the age of consent in many places was 12 as recently as the Victorian era [0], and why would it be that low if there wasn’t a lot of people who wanted that? (It might just have been politicians who wanted that, but I fear that’s merely unjustified optimism).
The other main reason is the number of women who tell me about traumatic experiences is high enough that it’s plausible all of them have been (there’s no reason for me to know about everyone’s history, so of course I believe I know someone who hasn’t told me and never will).
Of course, this percentage does mean that the “think of the children” meme ought to put all children under 24-7 surveillance so that nobody can ever touch them. Side effects haven’t slowed that meme down before, why would it now?
[0] I never cared to check, so I don’t know how many or when any given jurisdiction changed
> and why would it be that low if there wasn’t a lot of people who wanted that?
Not to detract from your other points, which I agree with, but ... according to e.g. [0], and I've found similar data in other places, life expectancy for a girl in the victorian era was ~40. So you can't judge the number 12 by today's standards.
Life expectancy includes infant and child mortality and thus is commonly misunderstood. Given that you’re already alive and healthy at age 12, your personal life expectancy is going to be higher than the population mean.
And even later on it turned out, that perversion against children does happen un-exceptionally in authoritarian environment with no checks and balances and no one to investigate the claims. Which pretty much describes environment his patients were subject to.
So while alcohol is not truth serum and Sigmund Freud was not working on peer reviewed modern studies, some of perversions those people talked about might really happened. lLcohol or not, some of his patients might as well be suffering from what we call ptsd from child abuse.
In a book /Against Therapy/, Jeffrey Masson, Freud scholar, explains how Freud deluded himself into disbelieving accounts that he originally had believed. The thought of them being true simply became unbearable to him.
Freud used to talk about wish-fulfillment fantasies. Apparently he fell to his own wish-fulfillment fantasy. He wished to live in a world in which child sexual abuse was not a thing.
Not quite truth serum, but if you feel more comfortable with someone, you are more like to be more open to them. If you've watched some police interrogations you might see that one of the things they try and do is build rapport with the person they are questioning for this reason.
So decreasing inhibitions so that you don't care about this, or increasing empathy so that you feel more trust in general can let people be much more truthful. For instance, alcohol or MDMA, or talking to a close friend compared to a stranger.
Even if we ignore the strong ethical, humanitarian, and strategic objections, using torture as a tool to extract information always seemed like the wrong tool for the job. Causing someone distress/pain obviously decreases the chance that they will tell you the information willingly and incentivizes simply making something up to make the torture stop. Instead, modern successful[1] methods of interrogation like the Reid technique[2][3] are designed around incentivizing the desired outcome. Very low (sub-psychadelic) doses of MDMA in a comfortable "social" or "party" setting might be very effective.
I don't know how true the story is, but I once heard an anecdote about this method working during WW2. The Allied soldiers took a German POW out for beers and just chatted with without trying to push any particular subject. Then they made a boasting claim about how German aircraft couldn't keep up with the latest American engineering, causing the POW to respond with an excited objection explaining their new airplane's capabilities (the information they wanted).
[1] Often dangerously successful - the Reid technique in particular is know for a very high rate of false positives. "Of the three hundred and eleven people exonerated through post-conviction DNA testing, more than a quarter had given false confessions"[2]
> Often dangerously successful - the Reid technique in particular is know for a very high rate of false positives.
I would argue that makes it unsuccessful — and worse, that it’s a case of Goodhart's law, optimising for the measure (convictions) rather than the much harder task of actually finding the guilty people.
> The Babylon 5 episode "Intersections in Real Time" (S4E18) is almost entirely Sheridan being interrogated with the Reid technique.
It doesn't seem to agree with the Wikipedia description of the Reid technique.
I've watched this episode quite recently, and from what I remember, it was all about trying to break Sheridan's grip on reality, through techniques such as food and sleep deprivation, making it impossible to track the passage of time, and calling into question his reasoning skills. On top of that, the interrogator was seeking guilt admission, not the truth.
Entirely speculation, but governments will soon have (if they don't already) the ability to extract the answers to yes-no questions from subjects via fMRI or something similar.
It'll be interesting to see whether this capability moves out into the law enforcement community.
Unfortunately, honesty on the part of MIT and its defense contracting partner was in short supply during the bid and proposal process. Federal prosecutors were tipped off shortly after the program was funded, prompting the government to swiftly cancel the contract.