Are Narrow Focus Exhaustivity Inferences Bayesian Inferences?

Area of Science:

• Cognitive Science
• Linguistics
• Computational Linguistics

Background:

• Successful communication relies on pragmatic inferences beyond literal meaning.
• The Rational Speech Act (RSA) framework models pragmatic reasoning using Bayesian goal recognition.
• Interpreting question-answer sequences with narrow focus involves complex pragmatic inferences.

Purpose of the Study:

• To model the interpretation of narrow-focus question-answer sequences within the RSA framework.
• To investigate the influence of domain size and prior probabilities on pragmatic inferences in question-answer contexts.
• To test whether exhaustivity inferences are sensitive to domain size and prior probabilities as predicted by Bayesian models.

Main Methods:

• Utilized the Rational Speech Act (RSA) framework to model pragmatic interpretation.
• Designed experiments to collect data on question-answer sequence interpretation with narrow focus.
• Analyzed experimental data to assess the relationship between domain size, prior probabilities, and focus exhaustivity.

Main Results:

• Experimental data indicate that interlocutors' focus exhaustivity is not dependent on domain size.
• Exhaustivity of interpretation remained constant across varying prior probabilities of states.
• Observed results contradict the predictions of RSA models regarding the influence of priors and domain size.

Conclusions:

• The study's findings challenge Bayesian accounts of pragmatic inferences in language understanding.
• Focus exhaustivity in question-answer sequences appears independent of domain size and prior probabilities.
• Further theoretical development is needed to explain pragmatic reasoning beyond current Bayesian models.