Examining People's Assumptions When Reasoning from Consensus

Abstract

Reasoning on social media is complex. Individuals face lots of information of varying accuracy from different sources, often encountering misinformation. To guide their judgment about claims, people may often turn to the opinions of others, relying on consensus cues. However, this reliance adds complexity because the quality of a consensus varies greatly, like the independence of evidence and sources. Despite this important variation, people's assumptions about the value of consensus quality information are unclear. To explore whether people prefer consensus information supported by different authors or the same author, and whether those authors) provide different or the same reasons for their stance, individuals (N = 100) were recruited from Amazon's Mechanical Turk and exposed to claims on a mock social media platform. After rating their initial agreement with a claim, participants were shown diagrams summarising different combinations of diversity in authors and reasons in tweets responding to the claim. After selecting and reading their most preferred tweet, participants updated their agreement with the claim. Results showed systematic preferences across individuals. Consensus information involving different authors corroborating the same reason was preferred over all others. When this consensus type was unavailable, there were no systematic preferences overall. However, in these instances, there was systematicity within subgroups of individuals. Differences in consensus preferences did not lead to significant differences in agreement updating. These findings are important for understanding people's assumptions about consensus quality information and contribute to development of automated reasoning tools summarising consensus quality information for users to mitigate against misinformation.