The paper by Michael C. Frank and Noah D. Goodman from Stanford University explores the prediction of pragmatic reasoning in language games, focusing on how speakers and listeners infer intended referents. The authors model human behavior in simple referential communication games, assuming that speakers aim to be informative and listeners use Bayesian inference to recover the intended referent. The model is parameter-free and provides a close fit to human judgments, suggesting that information-theoretic tools can effectively predict pragmatic reasoning. The study involves three groups of participants: speakers, who bet on which word a speaker would use; salience, who identify the object referred to by an unknown word; and listeners, who predict the object referred to by a given word. The model's predictions are highly correlated with human judgments, demonstrating its effectiveness in capturing the richness of human pragmatic inference.The paper by Michael C. Frank and Noah D. Goodman from Stanford University explores the prediction of pragmatic reasoning in language games, focusing on how speakers and listeners infer intended referents. The authors model human behavior in simple referential communication games, assuming that speakers aim to be informative and listeners use Bayesian inference to recover the intended referent. The model is parameter-free and provides a close fit to human judgments, suggesting that information-theoretic tools can effectively predict pragmatic reasoning. The study involves three groups of participants: speakers, who bet on which word a speaker would use; salience, who identify the object referred to by an unknown word; and listeners, who predict the object referred to by a given word. The model's predictions are highly correlated with human judgments, demonstrating its effectiveness in capturing the richness of human pragmatic inference.