Do language models learn typicality judgments from text?
This addresses a problem in cognitive science and AI by evaluating if language models can mimic human-like conceptual knowledge, but it is incremental as it builds on prior research on knowledge acquisition from text.
The study investigated whether language models can learn typicality judgments from text, finding only modest correspondence with human judgments, indicating text-based exposure alone is insufficient for acquiring typicality knowledge.
Building on research arguing for the possibility of conceptual and categorical knowledge acquisition through statistics contained in language, we evaluate predictive language models (LMs) -- informed solely by textual input -- on a prevalent phenomenon in cognitive science: typicality. Inspired by experiments that involve language processing and show robust typicality effects in humans, we propose two tests for LMs. Our first test targets whether typicality modulates LM probabilities in assigning taxonomic category memberships to items. The second test investigates sensitivities to typicality in LMs' probabilities when extending new information about items to their categories. Both tests show modest -- but not completely absent -- correspondence between LMs and humans, suggesting that text-based exposure alone is insufficient to acquire typicality knowledge.