Most readers of the Infernal Machine (though you, like us, may have been taking a break from blogging) have probably read something about the controversial Facebook “emotional contagion” study. It seems that a couple of years ago a researcher at Facebook decided to see what would happen if he tweaked the words in Facebook’s “News Feed” of close to 700,000 users so as to manipulate the “emotional content” of the News Feed. Would users respond on their Facebook pages in step with the manipulations? That is, could Facebook make people feel better if they tweaked their secret algorithms to prioritize “positive” words in the News Feed? (People might spend more time on Facebook if they could!)
The researcher, Adam Kramer, then brought in some university researchers to look at the massive data set. They did some “big data” statistical analyses (of a relatively straightforward type) and added some psychological theory and found this:
When positive expressions were reduced, people produced fewer positive posts and more negative posts; when negative expressions were reduced, the opposite pattern occurred. These results indicate that emotions expressed by others on Facebook influence our own emotions, constituting experimental evidence for massive-scale contagion via social networks.
The paper was peer reviewed and published in the prestigious Proceedings of the National Academy of Sciences (PNAS).
The merits of the study itself are highly questionable. My social science colleagues tell me that with such a massive sample size, you are almost always going to arrive at “statistically significant” findings, whether you are measuring “emotional contagion” or compulsive belching. The fact that the statistically significant “effect” was minimal throws more doubt onto the validity of the study. Furthermore, my social science colleagues tell me that it is doubtful that the study is really measuring “emotional contagion” at all — there are other theories (other than emotional contagion) available that would explain, why when Joe posts a “negative” statement, Jane is reluctant to follow with a “positive” one.
But the main controversy surrounds the ethics of the study: Participants were never made aware that they were part of a massive “experimental” study, and Facebook seems to have fudged on the timing of the “data use policy,” inserting the current bit about “data analysis” and “research” after the data was collected. This is out of keeping with common practice in the social sciences, to say the least.
Reviewing responses to the study within the academic community, I’ve noticed quite a lot of squirming. The fact is that university researchers were a major part of this study; Cornell University’s Institutional Review Board (IRB) board approved the study; and a prestigious academic journal has published it.
Now everyone, including Facebook’s researcher Adam Kramer, is squirming. Cornell University issued a bland statement clearly intended to wiggle its way out of the issue. PNAS has issued an “editorial expression of concern.” And the new media researchers that work in the same circles as the the university-based Facebook researchers have turned to such cliches as don’t “throw the baby out with the bath water” (don’t over-react to university researchers using corporate-owned data on how they attempt to manipulate their users!). They say people should stop “pointing the finger or pontificating” and instead “sit down together and talk.” “We need dialogue,” writes Mary Gray of Indiana University and Microsoft Research, “a thoughtful, compassionate conversation.”
Okay. But perhaps before we have that compassionate sit-down about the ethics of big data social media research, we might take a moment to think about the ethics of squirming. For, in this case, there has been way too much of it. I, for one, think TIME’s television critic James Poniewozik has been clear enough about the study: “Facebook can put whatever it wants in the fine print. That shouldn’t keep us from saying that this kind of grossness is wrong, in bold letters.” I don’t think much more needs to be said about the ethics of this particular study.
But something more does need to be said about the broader ethics of research, which sometimes puts us in uncertain ethical situations. There is something about the will to know, and more about the professionalization of knowledge production, that leaves us more frequently than we would like in tricky ethical territory. Rather than simply relying on an IRB “stamp of approval” university researchers might instead simply stop squirming and take responsibility for their work and even say they regret it.
Here’s what an ethics of squirming might look like:
(a) Start with full disclosure. Here are the conflicts I am dealing with (or have dealt with); these are the messy issues; here is why I really want to squirm, but I won’t.
(b) Here’s where I (or we) may have been, or indeed were, wrong. Here are our moral regrets. (By the way, Mr. Kramer, making people “uncomfortable” is not a a moral regret.)
(c) Here’s why I would or would not do it again.
All in all, the ethics of squirming entails less squirming and more speaking directly to the issues at hand. It means taking responsibility, either by repenting of wrongs and righting them if possible, or justifying one’s actions in public.
All together now: Mea culpa.
. . . . . . . .