Problem
Research questions and friction points this paper is trying to address.
Cognitive Delegation
Large Language Models
Human-AI Interaction
Reasoning Steps
AI-initiated Decisions
What Did the AI Take On? Characterizing Cognitive Delegation in LLM Reasoning
🤖 AI Summary
This study addresses the blurred boundary between user intent and AI autonomous cognition during LLM reasoning, which risks either uncontrolled critical decisions or excessive oversight burdens. Drawing on qualitative interviews, 992 retrospectively annotated data points, and cognitive science theories, this work constructs the first step-level multidimensional taxonomy for reasoning, encompassing cognitive labor, delegation execution, and expectation protocols. Our analysis reveals that 48.6% of reasoning steps are proactively initiated by the AI, uncovering implicit cognitive delegation mechanisms in human–AI collaboration. Based on these findings, we propose design implications supporting flexible protocols alongside inspectable and revisable AI decisions. This research provides both theoretical and empirical foundations for achieving controllable human–AI cognitive delegation in LLM-based reasoning systems.
📝 Abstract
Large language models (LLMs) often perform intermediate cognitive work while carrying out users'requests, yet it remains unclear which parts users intended to delegate and how they wanted to remain involved. This matters because consequential choices may go unnoticed, limiting users'ability to steer the process, while reviewing every step would make delegation burdensome. We examined this with 24 LLM users across three knowledge-work tasks, collecting 992 retrospective annotations of reasoning steps. From this, we developed taxonomies of LLM cognitive work, delegation enactment, and desired delegation protocols at the reasoning-step level. Our analysis revealed that participants viewed about half of all steps (48.6%) as AI-initiated, meaning the AI took on work they had not requested. Desired involvement varied with cognitive work and delegation enactment, even when contributions matched participants'intent. We propose design implications and sketches for supporting more deliberate cognitive delegation through flexible protocols and inspectable, revisable AI-initiated decisions.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Cognitive Delegation
LLM Reasoning
Taxonomy Development
AI-initiated Decisions
Human-AI Interaction
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
