Behind the EMNLP keynote: LMs as simulated learners to generate child language acquisition hypotheses
najoungkim · x · 2026-10-26
After her EMNLP2026 keynote, Najoung Kim pointed to her paper with Kanishka Misra: a systematic framework that uses LMs to generate novel experimental hypotheses.
Core idea
- Instead of only evaluating human-machine alignment, the work uses LMs to simulate outcomes of experiments that don't yet exist in the literature, extracting new insights about human cognition.
- Instantiated on dative verb acquisition and cross-structural generalization in child language: simulated learners yield the testable hypothesis that alignment between argument ordering and discourse prominence in exposure contexts modulates how children generalize new verbs to unseen structures.
- The authors also design lab experiments with children to test these hypotheses, contributing both a domain-general framework and domain-specific, lab-testable hypotheses.
More from AGI Musings
- Dev warns abliterated models could wreck software supply chains with 1000+ transitive dependencies — nitarshan · 2026-10-27
- Delete your priors: why every SOTA model drop should reset how you think the world works — zack_overflow · 2026-10-27
- Replit predicts every Fortune 500 will be a 'Self-Driving Company' by end of 2030 — amasad · 2026-10-27
- Anthropic says insurers can't cover its own catastrophic risk, X mocks it — GarrisonLovely · 2026-10-27
- MIT's Ziming Liu: AutoResearch doesn't generalize — blind training tricks cause 'forking' disasters — ZimingLiu11 · 2026-10-26
- AI accusations shake literary prizes from Caine to Goncourt: who decides what's human? — TuhinChakr · 2026-10-26