Neural Networks Track the Logical Complexity of Boolean Concepts

Area of Science:

• Cognitive Science
• Psychology
• Artificial Intelligence

Background:

• Two prominent theories, the language of thought hypothesis and connectionism, explain category acquisition.
• The degree of agreement between these theoretical frameworks remains largely unexplored.

Purpose of the Study:

• To compare the predictions of the language of thought hypothesis and connectionism regarding category acquisition effort.
• To identify areas of convergence and divergence between the two models.

Main Methods:

• A comparative analysis of theoretical predictions.
• Focus on the cognitive effort required for acquiring categories.
• Examination within the domain of Boolean categorization.

Main Results:

• The language of thought hypothesis and connectionism yield similar predictions for Boolean categorization tasks.
• Substantial variations in predictions emerge contingent upon the specific logical operators employed within the language of thought framework.

Conclusions:

• The two major accounts of category acquisition show significant overlap in their predictions for simple Boolean concepts.
• The complexity of the underlying representational operators critically influences the divergence between these theoretical perspectives.