Alison Gopnik is a developmental psychologist and philosopher at the University of California, Berkeley. Her research focuses on how young children learn, and she has argued that the dominant framing of large language models — as intelligent agents, or as steps toward AGI — is a category error.
Background
Gopnik is known for work on children's causal learning, the "theory theory" of cognitive development, and Bayesian and probabilistic models of how children build models of the world. A throughline of this work is that children are active learners who experiment, form causal models, and generate new knowledge. She applies this developmental-psychology lens to questions about AI.
Positions on AI
Gopnik originated the framing of LLMs as cultural technologies (see AI as Cultural Technology). She argues that LLMs are effective at transmitting and recombining the existing corpus of human knowledge, as writing, print, libraries, and search engines do, but that they do not run the causal, experimental learning loop that produces new knowledge, and are not agents or minds. She states this argument in Yiu, Kosoy & Gopnik, Transmission versus truth, imitation versus innovation: What children can do that large language models cannot (2024).
She also collaborates with Henry Farrell, Cosma Shalizi, and James Evans on the related AI as Social Technology framing, which treats AI as continuous with human social institutions rather than as an emerging agent. Across both lines of work she describes herself as a skeptic of agentic-AI and existential-risk framings, arguing they misdescribe current systems and crowd out questions about access, control, and the quality of the cultural record. Her position contrasts with both AGI-imminent forecasting (see AGI Timelines) and existential-risk framings, supplying the developmental-psychology evidence base on which the "AI as cultural or social technology" arguments draw.
Relationships
- proposes: AI as Cultural Technology
- supports: AI as Social Technology
- collaborates-with: Henry Farrell, Cosma Rohilla Shalizi
- contradicts: AGI Timelines, AI Existential Risk — argues the agent/AGI framing misdescribes current systems