Hacker News new | past | comments | ask | show | jobs | submit login

People have some well-founded respect for the procedures that medical trials have to go through, including informed consent. That’s spilling over/being misdirected to the use of the trial mechanism in other fields, even though it makes absolutely no sense.

The alternative to A/B testing in UX design is not some more rigorous trial with fully-informed consent. Instead, it‘s the older and still predominant mechanism of “let’s just change it”.

Consent in medical tests is important because it involves a ‘B’ that is a substance or procedure that has not (yet) been shown to be safe & effective. The trial is less restrictive than the alternative scenario, where it would be illegal to just treat you with ‘B’.

None of that applies: both the A and B in A/B tests are software versions, for which there are zero legal requirements. Neither version will have an impact on your health or kill you, which cannot always be said for untested pharmaceuticals. A tailor is absolutely free to suggest dark navy to one customer and anthracite to the next, even if they flip a coin to decide and/or mentally take note of what customers like.

The possible harm from A/B testing is the maximum of the harms of A and B. If you wouldn’t complain about either alternative, there is no legitimate reason to complain about the combined A/B harm.




A/B tests suck if you are in the test group and suddenly, completely unannounced, a functionality you have been relying on stops working or gets changed and you are now stuck with the problem.

> None of that applies: both the A and B in A/B tests are software versions, for which there are zero legal requirements.

There should be, that is my point.

All that A/B testing and the constant other changes even in minor patch versions only lead to update hesitancy in people, which in turn leads to security issues because people don't update their software because they don't want to be disrupted in their workflows.

Not to mention that most A/B testing is essentially pseudoscience at best if not outright fraud that's being peddled by people wanting to sell A/B testing "solutions", or the implications on user privacy because of the mandatory phone-home.

FWIW this is what made Microsoft so big in the corporate world: for decades you could expect that nothing major would change, you would not need to re-train your employees or buy new software. That changed with Windows XP (although it was easy to roll back at least the optical changes) and got exponentially worse with the utter crap that is Windows 11.


The alternative to A/B testing is paying people to test your software and learning from their direct feedback. Alternatively, you can provide people with a choice to test the changes you're considering.

If I pay for a product, I expect to be treated with respect. Being turned into a guinea pig to save on feedback cost isn't being respectful. Doing so without notice or consent is even worse.

There's no law against it, of course, but that doesn't mean it's good or even okay.


Surely if it's not informed consent then it's not authorised access to your computer. Unauthorised computer access is illegal in USA (CFAA) and UK (CMA), at least.

If it in anyway restricts bandwidth or causes other harms, eg uses up the last of your HDD space causing an issue, then that's an additional crime according to UK law.

These laws shouldn't just be applied to individuals.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: