Cognitive scientist and Dileep George clash over whether LLMs can be said to have minds

On October 11, cognitive science researcher Andrew Lampinen and Dileep George engaged in a multi-round public debate on X over whether psychological terms can be used as analogies for LLMs. The core disagreement: is talking about LLM "mental states" a scientifically grounded metaphor, or unmeasured anthropomorphism?

What's confirmed

• The debate was sparked by George pressing on "where do mental states come from": Lampinen clarified he is not a dualist, but stressed a core tenet of cognitive science—through proper manipulation experiments, one can infer mental states and processes from external behavior, and use these to understand and predict behavior in other contexts, so metaphors at this level are useful.
• George countered: "Where exactly have you ever measured a mental state? Can there be a mind without a brain?" He also criticized the other side for merely declaring their own anthropomorphism correct while dismissing others'; he further asked, if an LLM exhibited a survival drive, what exactly would prevent it from being analogized to a living being.
• George then issued a "metaphor consistency" challenge: since people accept that LLMs can "metaphorically feel pain" and be "metaphorically conscious," why can't they "metaphorically be a person."
• Lampinen responded that "person" refers to a physical living being, whereas intentions and beliefs are mental states—the two are different kinds of metaphors, and the latter may be more defensible; using the analogy "some scientists study ESP, so all science is the same," he pushed back against George's attempt to erase the distinction between the two types of metaphors by arguing mental states just are physical states.
• As of Lampinen's final full statement of his position, he said he had not seen George present an argument that would overturn his view, and the two sides reached no consensus.

Why it matters

• The debate touches on one of the thorniest boundary questions in current AI discourse: whether describing LLMs with mental vocabulary like "understanding," "intention," or "feeling" is a useful scientific analogy or anthropomorphism that encourages misattribution.
• The two positions represent the major schools of thought on this question: Lampinen stands in the cognitive science tradition of behavioral inference, arguing that the usefulness of such metaphors can be tested through manipulation experiments; George demands a stricter ontological standard and reveals the logical consequence that accepting a "metaphorical mind" may slide into "metaphorical personhood."

2026-10-11 ~ 2026-10-11 · 6 related posts

• [source] Dileep George pushes Lampinen: if LLMs feel metaphorical pain, why aren't they metaphorical persons? — dileeplearning · 2026-10-11
• Researchers debate: if an LLM feels metaphorical pain, why isn't it metaphorically a person? — dileeplearning · 2026-10-11
• Lampinen hits back: calling mental states 'just physical states' is like arguing science via ESP — AndrewLampinen · 2026-10-11
• [source] Dileep George: if an LLM has a survival drive, why isn't it analogous to a living thing? — dileeplearning · 2026-10-11
• Cognitive scientist Andrew Lampinen defends inferring mental states from behavior in LLM debate — AndrewLampinen · 2026-10-11
• [source] Andrew Lampinen defends behavior-to-mental-state inference as core cognitive science — AndrewLampinen · 2026-10-11