911±¬ÁÏÍø

Title: Tackling the Misinformation Problem with Inductive Learning and Receiver Operating Characteristic (ROC) Analysis

Abstract: Failure to tell apart true and false news can have devastating consequences. Therefore, we investigated how people discriminate between true and false news, and what can be done to improve this discrimination. First, we scrutinised the effectiveness of two popular misinformation interventions: Bad News and Go Viral!. Specifically, we used ROC analysis to reanalyse data from five papers (k = 13; n = 17,867). In contrast to what was reported in these studies, Bad News and Go Viral! did not improve true and false news discrimination, but rather elicited conservative responding (a tendency to rate all news as false). Second, we examined the type of knowledge people use to discriminate between true and false news. Accordingly, in a preregistered study (N = 327), participants rated the veracity of news headlines and indicated what decision strategy they used to make each rating. Participants discriminated between true and false news well despite choosing guess and intuition 63% of the time and only choosing rule and prior knowledge 21% of the time. Third, since news discrimination may predominantly be a tacit (rather than explicit) process, we reasoned that providing explicit guidance to improve it (a key feature of Bad News and Go Viral!) might have limited success. Therefore, we created an inductive learning intervention that involves observing true and false news headlines and classifying them as either true or false with immediate feedback (but no explicit guidance). Overall, across three preregistered experiments (N = 1,135), the intervention improved true and false news discrimination.

Host: David Shanks