Sunday, June 29, 2014

Something is rotten in the state of Facebook?

Facebook must have decided that people don't find their social media siteadvertising platform creepy enough, I know I didn't. Then Adam Kramer of Facebook's Core Data Science Team along with collaborators from UCSF and Cornell published the results of a social experiment performed on their users. The authors modified the contents of ~680,000 user's New Feed to see if they were able to manipulate the emotional state of Facebook users.

Federally funded research experiments, such as this, are required to be overseen by an institutional review board (IRB). The main job of an IRB is to protect subjects from physical or psychological harm. The events leading to the formation of IRBs are dark, very dark. A key principle used is that of informed consent. The editor at the journal is quoted as saying that the study was IRB approved (it is not clear which IRB, UCSF or Cornell) based on Facebook's terms of service agreement. The Facebook data use policy does mention research:
we may use the information we receive about you: ... for internal operations, including troubleshooting, data analysis, testing, research and service improvement.
The data use policy is about the privacy, the URL even contains the word. Facebook's response to the storm of criticism unleashed by this paper has focused on the privacy issue. That would be the issue if the paper was an observational study but it is not. I'm not a lawyer but the only information I could find on how Facebook manipulates their user's relates to the pairing of advertisements with social context. There is nothing within a stone's throw of informed consent covering active experimentation on their users.

I would have felt better if the study had somehow dodged getting IRB approval. That a university IRB and Facebook's own "strong" internal review process would rubber stamp this experiment is depressing. Maybe this is just a brief ethical lapse and not the sign of a troubling disregard of ethical issues in tech, but it is worrying.

1 comment:

  1. The Laboratorium has a very nice write up with more details about what informed consent means and how it was totally absent.

    ReplyDelete