Emotional Contagion: Your Chance to Opt-In

Emotional Contagion: Your Chance to Opt-In

This post is the last of a three part series covering the Facebook study on emotional contagion. As we noted in our first and second posts in this series, the original study (Kramer, Guillory and Hancock 2014) falls into an ethical gray zone. Allowing Facebook users to opt-in to (or opt-out of) such experiments would certainly address some of the ethical concerns, though limiting participants to those who opt-in (or -out) could affect the study’s conclusions since the optionality would directly impact the subject pool. That potential bias might then be factored into the researchers’ analysis to mitigate some of those impacts, since Facebook holds the data and user agreements, but it could diminish the reach of similar experiments.

Only Facebook can perform experiments like Kramer’s at a statistically significant scale, but there are ways to simulate the same alterations to your personal news feed. To this end we have designed an opt-in version of the study as a browser plugin that evaluates each item in your Facebook feed by the same LIWC lexicon used in the original study. Once you add the extension, the next time you visit Facebook our server will flip a coin and associate your Facebook username with one of the two groups: ‘omit posts with positive words’ or ‘omit post with negative words,’ as in the original experiment. Thereafter, all stories on your Facebook feed will be sent off to our server, which will respond with which stories to show and which to hide. Seven days after first using the extension, we will email you a survey to the email account associated with your Facebook username, and ask you to guess whether positive or negative stories have been omitted from your feed, upon which we will show which setting you actually got and the text content of the posts that you missed during that period. All data will be deleted after a month, whether or not you’ve answered the survey or seen your posts.

The extension can be installed here, and the source code can be downloaded from my Github page. At the moment, the extension is Chrome-only, but it could easily be translated to other extensible browsers, e.g., Firefox. If you want to reproduce the extension, you will need the full commercial LIWC lexicon, which is available here.

Note that this extension is experimental, and may not behave correctly if, for example, Facebook changes the styling or structure of the feed. Likewise, the server that is performing the evaluation should handle a reasonable number of users, but certainly nothing at Facebook scale. If your feed seems to stall, please just refresh the page and it should recognize the new posts. To uninstall from Chrome, enter chrome://extensions into the address bar, and click the trashcan icon next to the “Facebook Repression” extension.

We encourage you to test the system on your own! In the meantime, feel free to share your thoughts and comments here or by connecting with the People Pattern team at Hello@PeoplePattern.com, or requesting a demo of the platform. We look forward to publishing results of this trial in the coming weeks!