Taming the Centaur(s) with LAPITHS: a framework for a theoretically grounded interpretation of AI performances

2026-04-30

AI Summary

This study addresses the prevalent misinterpretation in contemporary AI research that equates human-level performance of large language models with human-like cognitive capabilities, often without theoretical grounding. The authors propose the LAPITHS framework, which uniquely integrates a minimal cognitive grid with behavioral contrastive analysis to rigorously distinguish superficial performance from genuine cognitive plausibility. Applying this framework, they successfully reproduce the high-performance results of models such as CENTAUR and demonstrate that such performance can be achieved by systems lacking cognitive plausibility, thereby challenging claims of “cognitive unity” attributed to these models. This work curbs the overextension of behaviorist interpretations and establishes a theory-driven paradigm for evaluating the cognitive capacities of artificial intelligence systems.

Abstract

We introduce a framework called LAPITHS (Language model Analysis through Paradigm grounded Interpretations of Theses about Human likenesS) and use it to show that several major claims advanced by models such as CENTAUR, proposed as an artificial Unified Model of Cognition, are not theoretically or empirically justified. LAPITHS provides a principled reference point for counteracting the current behaviouristic tendency in AI research to interpret the human level performances of transformer based language models as evidence of human like underlying computation and, by extension, as signs of cognitive abilities. The novelty of LAPITHS lies in making explicit the arguments grounded in two quantitative assessments: (i) the Minimal Cognitive Grid, a theoretically motivated method for estimating the cognitive plausibility of artificial systems, and (ii) a behavioural comparison showing that results similar to those reported for CENTAUR like models can be reproduced by other systems that do not satisfy the structural constraints typically associated with cognitive plausibility, and whose outputs do not provide independent explanatory insight into human cognition.

Problem

• AI interpretation
• cognitive plausibility
• behavioristic tendency
• human-like computation
• language models

Innovation

• LAPITHS
• Minimal Cognitive Grid
• cognitive plausibility
• behavioral comparison
• theoretically grounded interpretation