Thinking Makes LLM Agents Introverted: How Mandatory Thinking Can Backfire in User-Engaged Agents

TL;DR

This study reveals that mandatory reasoning in user-engaged LLM agents can backfire by reducing information sharing, leading to performance drops, but explicit prompts for transparency can improve outcomes.

Contribution

The paper provides a comprehensive analysis showing that explicit thinking can hinder user-agent interactions, and highlights the importance of proactive transparency for optimizing LLM agents.

Findings

Mandatory thinking often degrades agent performance in user scenarios.

Thinking causes agents to be more introverted, reducing information disclosure.

Explicit prompts for transparency improve agent performance.

Abstract

Eliciting reasoning has emerged as a powerful technique for improving the performance of large language models (LLMs) on complex tasks by inducing thinking. However, their effectiveness in realistic user-engaged agent scenarios remains unclear. In this paper, we conduct a comprehensive study on the effect of explicit thinking in user-engaged LLM agents. Our experiments span across seven models, three benchmarks, and two thinking instantiations, and we evaluate them through both a quantitative response taxonomy analysis and qualitative failure propagation case studies. Contrary to expectations, we find that mandatory thinking often backfires on agents in user-engaged settings, causing anomalous performance degradation across various LLMs. Our key finding reveals that thinking makes agents more ``introverted'' by shortening responses and reducing information disclosure to users, which weakens agent-user information exchange and leads to downstream task failures. Furthermore, we demonstrate that explicitly prompting for information disclosure reliably improves performance across diverse model families, suggesting that proactive transparency is a vital lever for agent optimization. Overall, our study suggests that information transparency awareness is a crucial yet underexplored perspective for the future design of reasoning agents in real-world scenarios. Our code is available at https://github.com/deeplearning-wisc/Thinking-Agent.