Analogical Reinforcement Learning: Computational Models and Human Experiments

James M. Foster's thesis presents a computational and empirical framework for learning abstract concepts. The work integrates principles of analogy and reinforcement learning, with models and simulations in the domain of tic-tac-toe. A second section describes human experiments testing model predictions about feedback and concept reinforcement.

Use Cases

• Developing computational models of analogical reasoning based on the described theoretical framework.
• Comparing AI agent performance with human learning patterns using the tic-tac-toe domain mentioned.
• Studying the effect of reward feedback on abstract concept formation as explored in the human experiments.
• Benchmarking reinforcement learning algorithms against cognitive models that incorporate relational structures.

Strengths

• Integrates computational modeling with human experiments, providing a multi-method approach.
• Focuses on a well-defined domain (tic-tac-toe) for testing abstract relational concepts.
• The work is presented as a thesis, suggesting a structured and documented research project.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

University of Colorado Boulder

Collection Method

Theoretical framework development, computational simulation, and human subject experiments.