Wow. That really hit home. I work for a large enough company that's constantly running experimentation, usually as A/B testing (not Facebook). It honestly never even dawned on me to consider how borderline ethical that is.
Deliberately infecting someone with syphilis is clearly unethical, regardless of whether or not there is a control group. However, if showing them different sorts of social media posts is ethical when done randomly, why is it unethical when done deliberately?
As far as deliberate vs. random goes, it's an important distinction because intent and consent are arguably paramount when dealing with questions of morality and ethics - as reflected by our legal system, where intent and consent are often a source of debate. In this case, experimentation is clearly done with intent and often without explicit consent (as opposed to the implied consent given by the Terms of Use).
It's why giving someone HIV (or Syphillis) unknowingly isn't illegal or really all that immoral (because the infector is unaware), but it is illegal to infect people intentionally.
Recently I went to a conference and acted like a sales guy. I gave different (truthful) pitches to different people and observed how enthusiastic they seemed about the product afterwards. Was that also unethical? Would it become unethical if I did a hypothesis test afterwards, rather than merely going with my gut?
Right, hence the whole rest of my message concerning intent, consent, and their importance with regards to morality and ethics.
> Was that also unethical?
Maybe? I wasn't there, I can't tell you. There are plenty of ways to act VERY unethically when it comes to sales and advertising, even while being truthful (e.g. you could omit some very important information). Deceptive advertising is a whole class of illegal actions in many developed nations, and not all of those actions involve the strictest definition of lying.
Not sure what hypothesis testing has to do with anything, experimentation certainly doesn't need to go hand-in-hand with statistical analysis.
I almost added a bit about taking the whole thing to its logical conclusion that the entire field of advertising and sales is arguably unethical. I haven't really thought that entirely through so I omitted it, but it's food for thought nonetheless.
I would argue it may not be, but this would depend on the exact situation. Did the user ask for random posts? Or is the selection of posts being imposed on them by a service? Do they even know that they are being show a certain subset of posts and how that subset is chosen?
Any time there is misrepresentation or dishonesty about what is really going on - including lies of omission - then could easily be ethical issues.
> Why is experimentation unethical.
Yes, experimenting on people - even things like Facebook's experiment - is shockingly unethical in all cases if it is done without an external watchdog such as an IRB[1]. Usually, it will also require the explicit informed consent of tho people involved, among other requirements (e.g. an opportunty to remove themselves from the experiment at any time, even after the experiment is over). In some cases, if the experiment is not risky, some or all of those additional requirements could be adjusted or waived.
We require this (by law under the Common Rule if federal money is involved), because human experimentation has a long, disturbing history of unethical behavior. You probably believe that an experiment involving blog posts is nothing like the problems we've had in the past... and I would agree. Facebook's experiment seems incredibly low risks when compared to the experiments performed by a pharmaceutical company or research hospital.
So why the outrage at Facebook? While there is some concern over their failure to properly debrief those involved (and other problems relating toi informed consent), most of the outrage is about their lack of proper IRB approval. Running an experiment may be ethical. The point is that YOU, THE EXPERIMENTER are NOT do not get to make that decision on your own! The point isn't that these experiments were unethical in some way.
The entire point of an IRB is to review the experimental methodology to verify that it meets all ethical requirements. This should be easy for most experiments conducted by a social media business or other software/internet business. All silicon valley has to do is setup their own IRB (they could probably partner for a while with a university IRB to gain legitimacy faster).