# AI reasoning effort predicts human decision time in content moderation

## TL;DR

This paper demonstrates that the length of reasoning chains in large language models correlates with human decision times in content moderation, highlighting parallels in reasoning effort and decision difficulty.

## Contribution

It reveals that model reasoning effort, measured by chain-of-thought length, predicts human decision times, establishing a link between AI and human reasoning processes in practical tasks.

## Key findings

- CoT length predicts human decision time across models
- Longer CoTs occur when variables are held constant, indicating increased difficulty
- Models reference contextual factors more when making complex decisions

## Abstract

Large language models can now generate intermediate reasoning steps before producing answers, improving performance on difficult problems by interactively developing solutions. This study uses a content moderation task to examine parallels between human decision times and model reasoning effort, measured using the length of the chain-of-thought (CoT). Across three frontier models, CoT length consistently predicts human decision time. Moreover, humans took longer and models produced longer CoTs when important variables were held constant, suggesting similar sensitivity to task difficulty. Analyses of the CoT content shows that models reference various contextual factors more frequently when making such decisions. These findings show parallels between human and AI reasoning on practical tasks and underscore the potential of reasoning traces for enhancing interpretability and decision-making.