The impact of engagement and partisan influence campaigns in an isolated social media environment

Despite growing concerns about the effect of social media engagement on people’s beliefs and behavior, estimating the actual impact is difficult. Here we present preliminary results from our own isolated social media platform named Magpie Social. In it, participants could interact with each other like typical social media, but we had control over the platform and measured people’s beliefs and behavior before and after using it. This allowed us to more closely approximate the ecological validity of naturally occurring social-media data, while retaining the ability to measure variables and infer causation. Our week-long task had three between-subject conditions (total N= 311): a CONTROL in which people engaged on Magpie with no external influence, and two (LEFT and RIGHT) in which a small number of posts were secretly made by us, sharing typical talking points from one political side. We found small but statistically reliable effects suggesting that, relative to the CONTROL, the presence of right-wing trolls resulted in a higher level of right-wing belief and a greater perception of political division in the US. Conversely, the left-wing troll campaign did not appear to have any statistically reliable effect on these measures. We also found considerably more overall engagement in both troll conditions, probably because content with a clear political stance tended to receive more activity. However, participants (especially those on the left) disliked the RIGHT condition more than the others.