Cognitive Plausibility Ranking of Analogy and Metaphor Models
A new study applies the Minimal Cognitive Grid (MCG) to systematically evaluate computational models of analogy and metaphor, including SME, CogSketch, METCL, and LLMs. The MCG framework assesses cognitive plausibility across three dimensions: Functional/Structural Ratio, Generality, and Performance Match. The research provides a formal, quantitative operationalization of the MCG, enabling consistent comparison of models against standard cognitive theories. The paper is available on arXiv.
Key facts
- The study uses the Minimal Cognitive Grid (MCG) to evaluate computational models of analogy and metaphor.
- Models assessed include SME, CogSketch, METCL, and Large Language Models (LLMs).
- The MCG framework has three main dimensions: Functional/Structural Ratio, Generality, and Performance Match.
- The evaluation offers a formal and quantitative operationalization of the MCG.
- The research allows comparison of models based on cognitive plausibility using consistent mathematical criteria.
- The paper is categorized under Computer Science > Artificial Intelligence.
- The study aligns each system with standard cognitive theories of analogy and metaphor.
- The paper was submitted to arXiv.
Entities
Institutions
- arXiv