A Cognitively Grounded Bayesian Framework for Misinformation Susceptibility
Researchers have developed a new Bayesian framework called the Bounded Pragmatic Listener (BPL) to model how people are susceptible to misinformation. This framework incorporates cognitive limitations such as working memory, information bottlenecks, and sampling biases. The BPL framework was tested on the LIAR and MultiFC benchmarks, demonstrating its effectiveness in classifying veracity and supporting the concept of a "depth-mismatch paradox" in susceptibility to misinformation. AI
IMPACT Introduces a novel framework for understanding and potentially mitigating the spread of misinformation.