Social meta-inference and the evidentiary value of consensus.

Reasoning beyond available data is a ubiquitous feature of humancognition. But while the availability of first-hand data typically diminishesas the concepts we reason about become more complex, our ability to drawinferences seems not to. We may offset the sparsity of direct evidence byobserving the statements of others, but such social meta-inference comes withchallenges of its own. The strength of socially-provided evidence depends onmultiple factors which themselves must be inferred, like the knowledge, socialgoals, and independence of the people providing the data. Here, we present theresults of an experiment aimed at examining how people draw conclusions frominformation provided by others in the context of social media posts. Bysystematically varying the degree of consensus along with the number of peopleand distinct arguments involved we are able to assess how much each factoraffects the conclusions reasoners draw. Across a range of topics we find thatwhile people are influenced by the number of people on each side of anargument, the number of posts is the dominant factor driving beliefrevision. In contrast to well established findings in simpler domains, we findthat people are largely insensitive to the diversity of the arguments made.