A computational model of language functions in flexible goal-directed behaviour

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psycholinguistics

Background:

• Inner speech, a form of self-directed language, is crucial for higher-order cognitive functions.
• Existing research links inner speech to working memory, perception, categorization, and executive functions.
• The precise role of inner speech in feedback processing and cognitive flexibility requires further elucidation.

Purpose of the Study:

• To investigate the relationship between inner speech and feedback processing/cognitive flexibility.
• To develop a computational model simulating an agent using inner speech for internal representation manipulation.
• To assess the model's ability to replicate human behavioral data in cognitive flexibility tasks.

Main Methods:

• Development of a computational model controlling an artificial agent.
• The agent utilizes inner speech for internal manipulation of representations.
• Model performance evaluated against human behavioral data from the Wisconsin Card Sorting Test (WCST) under varied conditions, including verbal shadowing.

Main Results:

• The computational model successfully reproduced human behavioral data from the WCST.
• Lesioning model components revealed inner speech significantly improves internal representation manipulation efficiency.
• Inner speech disentangles representations linked to visual features, enhancing attentional control following feedback.

Conclusions:

• Inner speech enhances goal-directed internal manipulation of representations.
• The model demonstrates how inner speech can improve behavioral flexibility.
• Findings suggest inner speech plays a key role in adaptive cognitive processing and decision-making.