Lunch Seminar Series | Can We Crowdsource Fact Checking? Initial Findings from a New NYU-Stanford Collaboration
Lunch Seminar Series | Can We Crowdsource Fact Checking? Initial Findings from a New NYU-Stanford Collaboration
Tuesday, February 11, 20201:00 PM - 2:15 PM (Pacific)
Encina Hall, Second Floor, East Wing, E207
616 Jane Stanford Way, Stanford, CA 94305
The research on misinformation generally and fake news specifically is vast, as is coverage in media outlets. Two questions run throughout both the academic and public discourse: what explains the spread of fake news online, and what can be done about it? While there is substantial literature on who is likely to be exposed to and share fake news, these behaviors might not signal belief or effect. Conversely, there is far less work on who is able to differentiate between true and false stories and, as a result, who is most likely to believe fake news (or, conversely, not believe true news), a question that speaks directly to Facebook’s recent “community review” approach to combating the spread of fake news on its platform.
In his talk, Professor Tucker will report on initial findings from a new collaborative project between NYU’s Center for Social Media and Politics and Stanford’s Program on Democracy and the Internet designed to fill these gaps in the scholarly literature and inform the types of policy decisions being made by Facebook. The project has enlisted both professional fact checkers and random “crowds” of close to 100 people to fact check five “fresh” articles (that have appeared in the past 24 hours) per day, four days a week, for eights week using an innovative transparent and replicable algorithm for selecting the articles for fact checking. He will report on initial observations regarding (a) individual determinants of fact checking proficiency; (b) the viability using the “wisdom of the crowds” for fact checking, including examining the tradeoffs between crafting a more accurate crowd vs. a more representative crowd and (c) results from experiments designed to assess potential policy interventions to improve crowdsourcing accuracy.
About the Speaker: