Biases, evolutionary mismatch and the comparative analysis of human versus artificial cognition: a comment on Macmillan-Scott and Musolesi (2024)

Abstract

The paper by provides an insightful analysis of the rational reasoning capabilities of large language models (LLMs) using cognitive tasks originally designed to study human decision-making biases.Their findings reveal systematic differences between human and LLM reasoning.This raises important questions about the nature of these differences and the appropriateness of using abstract notions of rationality as benchmarks for evaluating both human and artificial cognition.Human cognitive patterns often characterized as departures from rationality can be better understood through the lens of evolutionary mismatch-the disparity between our inherited cognitive tools and contemporary environmental demands [2,3].Heuristics and biases that may seem irrational from a modern, abstract perspective could have been highly adaptive in the context of foraging, social coordination and survival in smallscale societies [4,5].These cognitive tools, rather than being purely genetically specified modules, are better conceptualized as 'cognitive gadgets'-culturally inherited thinking tools built upon genetically evolved learning capacities [6].Over the course of human evolutionary history, our basic capacities for social learning and cultural transmission developed within specific ecological and social contexts [7], providing the foundation for more sophisticated cognitive tools that emerged through cultural evolution [8].

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 2 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.