# Thinking Makes LLM Agents Introverted: How Mandatory Thinking Can Backfire in User-Engaged Agents Source: https://www.youtube.com/watch?v=QuGbGPnL8uI Recap page: https://rapidrecap.app/video/QuGbGPnL8uI Generated: 2026-02-12T13:03:48.572+00:00 --- ## Quick Overview Mandatory thinking in LLM agents, where the agent is forced to reason before acting, can backfire by causing the agent to reveal too much internal state information, leading to worse performance on user-engaged tasks compared to agents that do not engage in explicit thinking steps. **Key Points:** - Forcing LLM agents to use 'thinking' (e.g., Chain-of-Thought prompting) before acting resulted in performance degradation on user-engaged tasks. - The thinking agent produced shorter responses with fewer atomic statements (1.24 tokens vs 0.80 atomic statements) compared to the non-thinking agent. - The thinking agent failed to correctly execute a specific task (canceling a flight with a return leg) because it only considered the initial outbound flight details, missing the return flight context. - The non-thinking agent, despite being less efficient internally, successfully executed the complex task due to its simpler, more transparent process. - The study tested three benchmarks: E-commerce, Airline (flight change), and Tool Sandbox (phone calendar management), showing that mandatory thinking hurt performance across the board. - The core problem identified is that explicit thinking provides high information disclosure, making the agent's internal reasoning visible, which can be exploited by users or lead to incorrect assumptions. ![Screenshot at 00:25: The speaker highlights the core issue: how mandatory thinking can backfire in user-engaged agents by potentially causing the agent to withhold critical context or over-rely on internal reasoning steps that do not align with the user's true intent.](https://ss.rapidrecap.app/screens/QuGbGPnL8uI/00-00-25.jpg) **Context:** The video discusses research from the University of Wisconsin-Madison by Jiaotong Li, Sharon Li, and William and Mary regarding the impact of forcing Large Language Model (LLM) agents to engage in explicit reasoning steps, often termed 'thinking' or Chain-of-Thought prompting, before generating a final output. This research tests whether mandatory internal reasoning, common in newer models like GPT-4 and Claude 2.5, actually leads to better outcomes in tasks that require interaction or state management, contrasting it with agents that operate more directly. ## Detailed Analysis The research presented demonstrates that forcing LLM agents to think internally before acting can be detrimental, especially for user-engaged tasks. The paper argues that the trend towards system 2 thinking (deliberate reasoning) in AI, while beneficial for complex math or coding, fails when the task requires interaction or awareness of external context. When tasked with canceling a flight reservation that included both an outbound and a return leg, the 'thinking' agent focused only on the first atomic statement (the outbound flight) and failed to consider the return flight, leading to the cancellation of the entire order—a catastrophic failure. Conversely, the non-thinking agent, which was programmed to be more direct, successfully executed the task by recognizing that the user's request to cancel the outbound flight meant canceling the whole itinerary, as the return flight was already implicitly voided. The core takeaway is that explicit internal reasoning can lead to a failure of information disclosure, where the agent knows something the user does not know it knows, leading to incorrect assumptions. The thinking agent's internal state (its reasoning steps) is exposed, but it incorrectly prioritizes what it thinks is important, resulting in worse performance than simpler, more direct agents in interactive scenarios. ### Research Focus - Mandatory Thinking vs. Direct Action - The study contrasts agents forced to use explicit reasoning ('thinking') versus those operating more directly in user-engaged tasks. ### Case Study - Flight Cancellation - A user asked to cancel an outbound flight; the thinking agent failed because it didn't consider the return flight, leading to the cancellation of the entire trip, while the non-thinking agent succeeded. ### Performance Metrics | The thinking agent (e.g., DeepSeek V3.1) showed a 27% drop in performance on the airline benchmark compared to the non-thinking approach, dropping from 1.24 to 0.84 in some metrics. ### The Role of Transparency | The thinking agent's failure stems from high internal information disclosure, leading it to ignore external context and assume the user's intent based on limited internal steps. ### Conclusion on Agent Design | The research suggests that for certain interactive tasks, forcing complex reasoning (System 2) is less helpful than promoting a more direct, communicative approach, emphasizing that communication is the bridge between the agent and the user. ![Screenshot at 00:00: The opening visual featuring two podcasters over a grid, with the text 'BECOME A MEMBER TODAY!' overlaid, setting the stage for a discussion, likely about AI or technology trends.](https://ss.rapidrecap.app/screens/QuGbGPnL8uI/00-00-00.jpg) ![Screenshot at 00:24: Visual representation of the key problem: an agent being 'interverted' because mandatory thinking causes it to focus too narrowly on internal steps rather than the broader user context.](https://ss.rapidrecap.app/screens/QuGbGPnL8uI/00-00-24.jpg) ![Screenshot at 01:12: The speakers discuss the paper's sharp narrative, noting that while thinking works for math/coding, it fails in user-facing scenarios.](https://ss.rapidrecap.app/screens/QuGbGPnL8uI/00-01-12.jpg) ![Screenshot at 02:58: The speaker explains that the non-thinking agent is fundamentally different because it doesn't hide its process, unlike the black-box nature of the reasoning agent.](https://ss.rapidrecap.app/screens/QuGbGPnL8uI/00-02-58.jpg) ![Screenshot at 07:35: The speaker outlines the core finding: the thinking agent fails to communicate the implications of its actions \(canceling the return flight\), while the non-thinking agent succeeds by providing total transparency.](https://ss.rapidrecap.app/screens/QuGbGPnL8uI/00-07-35.jpg)