Don’t leave Big Data to computer scientists, Facebook, you need marketing and psychology

Did you notice your Facebook Newsfeed got a lot more positive or negative? Or did you get the message 2 years ago you were being locked out of Facebook unless you proved to be real, i.e. not a robot? Then you probably were a subject in a Facebook Data Science experiment. The news broke this week: Facebook has been experimenting on millions of users without their informed consent. The company claims you gave your consent by agreeing to its Terms of Service, which state that data may be used to improve Facebook’s products. Who is Facebook kidding? First, you and I did not agree to manipulation of our social media environment – instead we agreed to Facebook observing our natural interactions and analyzing its patterns. Second, how exactly does manipulating the mood of our newsfeed help Facebook “improve its products”? Instead, this is social science research, where getting answers to very interesting questions (how does your friends’ mood affect you?) should be balanced with the rights of human subjects whose feelings are being manipulated. But how can we achieve such balance?

We know the answer, and so should Facebook. Academics in biomedicine, marketing and psychology have wrestled with the issue of human subjects ethics for decades. Every researcher in these fields needs to get explicit prior permission from the university’s Ethics Board. Besides not (physically or emotionally) harming human subjects, informed consent is considered crucial in getting such permission. Instead, at Facebook’s data science team, “there’s no review process, per se, and anyone on that team could run a test” according to Andrew Ledvina, a Facebook data scientist in 2012-2013 (http://online.wsj.com/articles/facebook-experiments-had-few-limits-1404344378?ru=yahoo%3Fmod%3Dyahoo_itp#). Facebook says it has now adopted stricter guidelines and a panel of internal experts in fields such as privacy and data security. However, I suspect Facebook would be even better off by hiring PhDs that have learned human subjects ethics in their fields, i.e. biomedicine, marketing and psychology. Instead, the recruitment ads for their Product Science team ask for PhDs in ‘computer science, sociology, information, statistics, physics or related fields” (https://www.facebook.com/careers/department?req=a0IA000000CzAdDMAV). With all due respect to these fields, they typically don’t train their students in the ethical balancing act of manipulating people’s emotions. We need to channel less Saruman and more Gandalf in our care for the “subjects” we study.

Advertisement