How human-like is LLM cognition?

2026

Open publication workspace · Sign in to read the full PDF.

AI-generated summary

1) This paper synthesizes the growing body of research on Large Language Models (LLMs) and human cognition, arguing that LLM cognition is "jagged" and varies across dimensions rather than being uniformly human-like.
2) * Examines LLM cognition through design specifications, internal states, and behavior.
* Organizes empirical findings into patterns of alignment and divergence between humans and LLMs.
* Suggests that comparing LLMs to humans can refine our understanding of cognitive science.
3) LLMs, human cognition, artificial intelligence, cognitive science, representational alignment

Check the original publication for accuracy and context.