Evaluating Computational Analogues of Human Intuition

This exploratory Position Paper from Artificial Intuition Research (AIR) examines whether and how aspects of human intuition can be evaluated through computational analogues. It reviews six scientific accounts of intuition and considers what each account can and cannot support in a computational setting. The paper argues that behavioural resemblance, task performance, or the implementation of a model does not by itself establish that a computational system uses the same process as a human decision-maker. It distinguishes between implementing a theoretical construct, producing similar outputs on a task, and demonstrating correspondence with a human decision process. The paper also outlines the kinds of evidence needed for stronger evaluation, including clearly specified claims, appropriate baselines, response and error patterns, confidence and calibration, and process-sensitive measures where available. This is an exploratory research output developed through AIR; it has not undergone journal peer review.