Comment by lalos | Hacker News Reader

Comment by lalos | Hacker News Reader

Parent Full thread

There is a lot to it.

You can start here: https://en.wikipedia.org/wiki/Nuremberg_Code#The_ten_points_...

Specially point number 1. Also read on the history of Nuremberg Code and why it was triggered after WW2.

Also you could read this explanation by a professor of law at U of Maryland http://laboratorium.net/archive/2014/06/28/as_flies_to_wanto... which talks about Facebook research which is similar to this one.

In general: you are tricking the end user, this is not a simple A/B test this is more of here is a company that does not exist providing nothing but let's see how you react to it. Today is fake pizza, tomorrow is a fake HIV med that cures it in one day (you click on it, there is a strong inference that you have it), etc the possibility is endless. Who is deciding this? That's why I mention ethic panels or regulation or whatever way of putting limits on it.

The only good thing here is now we know that we are test subjects when using Youtube/Google.

Well if it's just the slippery slope argument then I get that but I don't agree it's a problem until we actually do hit fake HIV med ads.

Furthermore, it's been pretty normal for startups to try gauging customers interest by setting up pages with products that don't exist yet. Maybe it's different because of the size of google but I am personally waiting until the slippery slope hits things like fake HIV cure ads.

Don't forget the time they replaced videos on the YouTube homepage with rick astley and deceived users into clicking!

Those inhumane monsters

It is not that, it is mostly informed test subjects. The slippery slope argument is the most extreme case let's say. The thing is this is not black or white, there are grays in between, who decides what shade of gray? Do we wait like we did with social media and until we see another CEO in front of Congress saying 'I'm sorry we helped misguide people'? Or should we be more proactive? Moving fast and breaking stuff works when you are small company with no liability not when you are responsible for a bunch of data of the whole world.

Why is this "not a simple A/B test"? It seems to be functionally the same (maybe A/B/C/etc. and not A/B), just with no actual product.

Is it the nonexistence of a real product which bothers you? Is it the comparison testing?

More generally, what do you feel constitutes an experiment that requires informed test subjects? I agree with you that there are shades of gray, but I feel the distinction likely lies somewhere between "red vs green button on landing page" and "new cancer drug," while you seem to be arguing that the former should still require informed consent. (Which could be a reasonable argument! But I'm trying to get a better gauge on what it is that you're saying.)

Again, informed test subjects. Were the users informed? How do we know when is it a real ad and when are we test subjects? Are they going to publish every test and result they do or cherry pick the funny pizza ones only? Maybe read this other comment who explains it way better and organized https://news.ycombinator.com/item?id=17815404

This is almost literally a Godwin argument. Your argument basically unpacks into, "there's a dangerously slippery slope between fake door tests and war crimes", and it's most often used as a self-serving justification by bureaucrats who want you to do hours of busy work before giving people a questionnaire: http://slatestarcodex.com/2017/08/29/my-irb-nightmare/

Look, marketing researchers are not going to just accidentally vivisect concentration camp inmates to death just because you leave them unregulated. It's not like people just accidentally go around committing crimes against humanity because there aren't any regulations stopping them from doing so. In fact, at Nuremberg, it turned out that most of them committed crimes against humanity specifically because there were regulations requiring them to do so!

This isn't to completely pooh-pooh the necessity of research ethics and regulation. But those regulations need to be based on realistic risk models.

It is not my fault that the rules that govern research and interaction between researchers and tests subjects occurred because of WW2 and that Godwin's law exists. I mentioned that as a historic fact. So nobody can talk about how this started because it is literally Godwin's law? Great idea.

I'm not saying it's your fault; I'm saying that AB testing pizza advertisements for a fictitious pizza brand on YouTube is so far away from the risk of committing war crimes that it's a big stretch to invoke the Nuremberg Code as a criticism of it.

The code is more about a simple set of rules for research involving human test subjects. That is the baseline. It started because of that and got adopted by the scientific community as the most basic rules. I'm not implying anything about war crimes because I would've still referenced if they were created randomly by any other random event.

If they would first create a Doctor Fork pizza in some small town ("we won't deliver to you, but wherever you are, you can come here and try it") would it be OK then? What is your argument based on in that scenario?

Because if you think that it would be OK, then I would argue it makes zero difference for the end user experience. And if it doesn't make any difference to users, then the only issue seems to be that you don't like something about it.

There are loopholes everywhere, yes that is one. No, it's not about finding the minimum set of changes to make it airtight. It is about informing test subjects that they are being tested on.

Reply on news.ycombinator.com