Humans and LLMs Diverge on Probabilistic Inferences

TL;DR

This paper introduces ProbCOPA, a dataset of probabilistic inferences, and shows that current LLMs fail to replicate human-like probabilistic reasoning, highlighting differences in reasoning patterns.

Contribution

The paper presents ProbCOPA, a new dataset for probabilistic inference, and provides an analysis comparing human judgments with LLM responses, revealing significant gaps in reasoning capabilities.

Findings

Humans show graded, varied probabilistic judgments.

LLMs fail to produce human-like probabilistic distributions.

A common reasoning pattern is identified in LLMs.

Abstract

Human reasoning often involves working over limited information to arrive at probabilistic conclusions. In its simplest form, this involves making an inference that is not strictly entailed by a premise, but rather only likely given the premise. While reasoning LLMs have demonstrated strong performance on logical and mathematical tasks, their behavior on such open-ended, non-deterministic inferences remains largely unexplored. We introduce ProbCOPA, a dataset of 210 handcrafted probabilistic inferences in English, each annotated for inference likelihood by 25--30 human participants. We find that human responses are graded and varied, revealing probabilistic judgments of the inferences in our dataset. Comparing these judgments with responses from eight state-of-the-art reasoning LLMs, we show that models consistently fail to produce human-like distributions. Finally, analyzing LLM reasoning chains, we find evidence of a common reasoning pattern used to evaluate such inferences. Our findings reveal persistent differences between humans and LLMs, and underscore the need to evaluate reasoning beyond deterministic settings.