Open Conference Systems, CLADAG2023

Font Size: 
Bayesian aggregation of crowd judgments for quantitative fact checking
Michele Lambardi di San Miniato, Michela Battauz, Ruggero Bellio, Paolo Vidoni

Last modified: 2023-05-31

Abstract


Political fact-checking can be carried out by crowd workers, providedthey are supervised by experts. We propose a Bayesian latent variable ordinal probitmodel for truthfulness rating data, to estimate workers’ reliability, weigh in their con-tributions, and surrogate expert judgments. This is a notable example of aggregationfunction of an implicit type. This method may be used to dynamically assign workersto new tasks, as illustrated with an analysis of PolitiFact data.