Google ran ads under a false brand name to test how viewers reacted to them. From what I can tell, the marketing experiment was fairly similar to standard A/B t

Google ran ads under a false brand name to test how viewers reacted to them. From what I can tell, the marketing experiment was fairly similar to standard A/B tests advertisers on ad design and copy or the kind of tests used by designers and developers for new site functionality. The biggest difference being that--and this is the possible ethical concern from the parent comment--the the brand wasn't real. Users had no way of knowing who the ad was produced by, or who was attempting to analyze their behavior. They aren't totally analogous, but there are circumstances where obscured identity is tested: focus groups market research tests that ask consumers opinions about possible product lines and/or changes, political polling asking how potential voters would perceive an unnamed politician (such as might done when developing a response plan prior to a major scandal breaking...if they have time), etc. In those examples, however, the subject at least knows who is doing the research if not who commissioned it. And most importantly, they give consent. The Facebook paper was a much more significant ethical concern, largely because it was (1) clearly a psychological study and was, in fact, published as such[0]; (2) had the explicit goal of eliciting changes in subject emotional responses and behavior in order to test the hypothesis; and (3) made no effort to comply with standard ethical practices for informed consent in human research. The paper got around the lack of informed consent because Facebook was a private company and was able to point to their Data Use Policy. That was enough of a fig leaf for Cornell's Institutional Review Board to accept because the Cornell professor was only given access to completed data after the experiment was independently run by Facebook. They accepted it in those circumstances, but I doubt anyone was very happy about it. Google's experiment is quite tame in comparison, and there are obvious parallels to the sort of benign A/B testing marketers use. There's also--arguably--a pretty major difference in the type of experimentation going on, and it's not intended for academic publication. Also, YouTube terms of use presumably contain language that gives them enough cover to avoid any legal consequences. Ethicists and university IRBs put a lot of thought into where the boundaries lie and how proposed experiments can cross them unexpectedly. Some of those boundaries have been learned about after some truly horrendous experiments in our past that nobody ever wants to even consider repeating. Others have been much less destructive, but nevertheless with their own flaws and problems. Personally, I think Google ought to have at least labeled it as a research ad and put their name on it at the end or on the page users clicked through to. While this example was fairly benign with only relatively minor ethical concerns, I understand the parent comment's concerns. There's a lot of interesting questions that can be researched, and I'd imagine some pretty abusive ways in relevant data could be discovered. In that context, the discussion's a good one to have for both the public and the companies themselves. Not because they're likely to jump into those abusive methods or because they'll repeat something along the lines of the Milgram experiment (or far, far worse). It's simply useful to understand the pitfalls beforehand. I'm not normally one for slippery slopes, but there are instances where considering them beforehand helps avoid the problematic outcomes altogether. 0. http://www.pnas.org/content/pnas/111/24/8788.full.pdf http://www.pnas.org/content/pnas/111/24/8788.full.pdf