Crowdsourcing strategies to fact-check politicians are benefited from heterogeneous social influence

NAVAJAS, JOAQUIN

Conferencia; The Cognitive Science Society Conference (CogSci) 2022; 2022

The spread of misinformation has substantially sped up in recent years, posing tangible risks to public health, democratic life, and the fight against climate change. For this reason, finding novel solutions to scale and reduce the time-intensive work of fact-checking agencies has become an urgent issue in social science. In this work, we evaluated crowd sourcing as a strategy to fact-check statements made by politicians, a type of misinformation that has been understudied in comparison to fake news. We performed a behavioural experiment (N=180) where the structure of social influence was manipulated, with the objective of identifying which communication network designs enhanced or undermined detection of 20 statements made by political leaders from a left-wing and a right-wing political party in Argentina. We found that crowds of laypeople are generally able to classify political discourse in a similar way as a fact-checking agency, but they are subject to partisan biases at the individual level. We also observed that social influence in heterogeneous dyadic interactions works as an antidote to these biases. Altogether, our results support recent efforts on developing crowdsourcing tools for aiding fact-checking over a variety of misinformation formats, adding a precautionary note on how partisanship may limit its performance, at least in the context of factchecking politicians