Probabilistic Learning and Psychological Similarity.

Area of Science:

• Cognitive Science
• Machine Learning
• Computational Psychology
• Developmental Psychology

Background:

• Psychological similarity and probabilistic learning are foundational concepts in multiple scientific fields.
• The relationship between these concepts is often not explicitly defined across disciplines.
• Computational cognitive science seeks to integrate these ideas for a deeper understanding of cognition.

Purpose of the Study:

• To critically evaluate the mutual contributions of psychological similarity and probabilistic learning within computational cognitive science.
• To explore how notions of psychological similarity can inform probabilistic models of cognition.
• To provide a framework for interpreting representational primitives in cognitive models.

Main Methods:

• Review and critical evaluation of existing literature.
• Case study using probabilistic models of concept learning.
• Analysis of how similarity representations impact the interpretation of probabilistic models.

Main Results:

• Two distinct notions of psychological similarity offer normative constraints for interpreting representational primitives.
• Similarity representations imbue probabilistic models with meaningful cognitive content, moving beyond purely mathematical interpretations.
• Subjective probabilities in models gain clearer interpretations regarding their connection to learning experiences.

Conclusions:

• Psychological similarity is essential for grounding probabilistic models in cognitive reality.
• Integrating similarity notions enhances the explanatory power of computational models of cognition.
• This integration bridges the gap between psychological theory and machine learning approaches.