Language models do not yet achieve explanatory adequacy because language is more than incremental prediction
- 1 July 2026
- journal article
- Published by Cambridge University Press (CUP) in Behavioral and Brain Sciences
- Vol. 49, e205
Abstract
In this commentary, we challenge the idea that Language Models (LMs) provide explanatorily adequate models of human language. Findings from language evolution show that both humans and LMs fail to produce human-like language via inductive biases alone; the communicative function of language is crucial. More generally, both experimental and computational work underscore the fact that language is more than just incremental prediction.Keywords
This publication has 12 references indexed in Scilit:
- What enables human language? A biocultural frameworkScience, 2025
- Communicative pressures shape language during communication (not learning): Evidence from case-marking in artificial languagesCognition, 2025
- AI models collapse when trained on recursively generated dataNature, 2024
- Artificial language learning2023
- Communicating artificial neural networks develop efficient color-naming systemsProceedings of the National Academy of Sciences, 2021
- Evolving artificial sign languages in the lab: From improvised gesture to systematic signCognition, 2019
- Structured Sequence Learning: Animal Abilities, Cognitive Operations, and Language EvolutionTopics in Cognitive Science, 2019
- Compression and communication in the cultural evolution of linguistic structureCognition, 2015
- The Oxford Handbook of Language Evolution2011
- Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human languageProceedings of the National Academy of Sciences, 2008