Active Iterative Social Inference in Multi-Trial Signaling Games.

Area of Science:

• Cognitive Science
• Behavioral Economics
• Game Theory

Background:

• Human decision-making reveals factors influencing choices.
• Speakers can infer listeners' choice priors from observed ambiguity resolution.
• Previous studies show limited strategic construction of ambiguous situations for learning.

Purpose of the Study:

• Investigate prior inference in complex learning scenarios.
• Examine information accumulation across trials.
• Assess the impact of active scenario construction on prior inference.

Main Methods:

• Signaling games were used to study referential ambiguity.
• Experiment 1: Assessed evidence accumulation over four trials.
• Experiment 2: Evaluated active scenario construction and iterative settings.

Main Results:

• Information integration for prior inference was only partially successful.
• Integration errors included transitivity failure and recency bias.
• Full task engagement and reasoning transparency enhanced optimal choices and prior inference.

Conclusions:

• Complex learning scenarios pose challenges for accurate prior inference.
• Active participation and transparent reasoning pipelines are crucial for strategic decision-making.
• Improved inference accuracy and strategic utterance choices result from deeper engagement.