Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session C: Friday, August 15, 2:00 – 5:00 pm, de Brug & E‑Hall

Step-by-step analogical reasoning in humans and neural networks

Jacob Russin¹, Joonhwa Kim², Ellie Pavlick², Michael Frank²; ¹University of California, Davis, ²Brown University

Presenter: Joonhwa Kim

Both humans and large language models (LLMs) perform better on some reasoning tasks when encouraged to think step by step. However, it is unclear whether these performance gains are based on similar principles. Testing both humans and LLMs on a novel word analogy task, we find that interference caused by semantic similarity hurts performance in both and drives humans to engage in a sequential reasoning process. These findings pave the way for investigation into the mechanisms that underlie the benefit of chain-of-thought and the decision process behind sequential thinking.

Topic Area: Language & Communication

Extended Abstract: Full Text PDF