8 Policy Brief No. 44 Toda Peace Institute otherwise be possible. Ironically, Muller and Schwartz’s paper debuted during a cultural moment when media coverage of Facebook was very heated. As a result, the paper was massively overhyped (especially considering that itis not peer-reviewed). Well-intentioned media outlets promoted wild claims, which eventually caused many readers to dismiss the paper’s findings. Still, notwithstanding the hype and attendant backlash, the methodology from Muller and Schwartz’s paper has potential, especially given that showing causation in this field is so hard. In fact, the main problem with the paper was that the researchers were forced to work with limited data, because Facebook was choosing not to share internal data with academics at the time. Many criticisms of the study focused on its data analysis, including some of Facebook’s criticisms. Yet as tech blogger Ben Thompson pointed out at the time (in a paywalled article on his website Stratechery), “Facebook itself can overcome some of the biggest deficiencies in this article: the company actually knows exactly how much people use Facebook on an individual level, they have location data for the vast majority of people, and these researchers have helpfully compiled all of the data of violence against refugees that Facebook would need to replicate this study.” Some experiments that are conducted off Facebook could potentially apply to Facebook. For example, in 2013, Stroud et al. published a study looking at affordances: They learned that a “Respect” button “yielded less polarized comment section behavior” when compared to the more common “Like” button. However, Stroud’s study was conducted using comment sections for news site articles rather than Facebook comments. Facebook — like other tech companies — routinely experiments with its buttons and other affordances. (Google infamously tested 41 shades of blue for its links, as blogged in 2009 by Douglas Bowman, a designer who worked on the test and then quit.) But the vast majority of those experiments are not public. In 2016, Facebook officially switched from the “Like” button to six different “emoji reactions” (though there is not a reaction that seems to denote “respect”). However, the data that led to this affordance change has never been released, and neither has the data they gained from changing it, so we do not know if this change was influenced by Stroud’s study, and we also do not know if Facebook’s results are similar to Stroud’s. 8. How Digital Platforms Can Change Voting Behaviour Some of Facebook’s fear of sharing data is related to its ongoing public relations crisis. Facebook was more willing to share data with researchers and publicise its internal experiments several years ago. Back then, some interesting research got published that shows causal mechanisms. One 2013 study by Bond et al., conducted in partnership with Facebook, found that Facebook has the ability to directly affect the voting behavior of millions of people. In other words, a small change in Facebook’s interface can cause many Facebook users to vote when they otherwise would not.