Metropolis-Hastings algorithm in joint-attention naming game: experimental semiotics study

Area of Science:

• Cognitive Science
• Computational Linguistics
• Semiotics

Background:

• Previous research explored symbol system organization via artificial experiments.
• Human communication and symbol emergence remain areas of active investigation.

Purpose of the Study:

• To experimentally test if human behavior in a joint-attention-naming game (JA-NG) aligns with Metropolis-Hastings Naming Game (MHNG) theory.
• To investigate the emergence of symbols through human interaction and communication.

Main Methods:

• Conducted an experimental semiotic study using a joint-attention-naming game (JA-NG).
• Compared human acceptance decisions of partner's naming with acceptance probabilities computed via the Metropolis-Hastings (MH) algorithm within the MHNG theory.
• Evaluated the predictive accuracy of the MH-based model against Constant, Numerator, Subtraction, and Binary models.

Main Results:

• Rejected the null hypothesis that human acceptance judgments are constant.
• The MH-based model significantly outperformed models with constant acceptance probability in predicting human behavior.
• The MH-based model demonstrated superior accuracy in predicting human acceptance/rejection decisions compared to four other models.

Conclusions:

• Human behavior in the JA-NG is consistent with the MHNG theory, suggesting symbol emergence can be explained by decentralized Bayesian inference.
• The MH algorithm provides a robust mathematical framework for understanding symbol emergence in human communication.
• The study validates the MHNG theory as a predictive model for symbol emergence in interactive settings.