Effective Human-AI Teams via Learned Natural Language Rules and Onboarding
This addresses the challenge of effective human-AI teaming for users interacting with AI assistants, though it appears incremental as it builds on existing collaboration frameworks.
The paper tackles the problem of improving human-AI collaboration by learning natural language rules that guide humans on when to rely on or ignore AI suggestions, using a novel region discovery algorithm and LLM-based descriptions. Results from user studies on object detection and question-answering tasks show it leads to more accurate human-AI teams.
People are relying on AI agents to assist them with various tasks. The human must know when to rely on the agent, collaborate with the agent, or ignore its suggestions. In this work, we propose to learn rules, grounded in data regions and described in natural language, that illustrate how the human should collaborate with the AI. Our novel region discovery algorithm finds local regions in the data as neighborhoods in an embedding space where prior human behavior should be corrected. Each region is then described using a large language model in an iterative and contrastive procedure. We then teach these rules to the human via an onboarding stage. Through user studies on object detection and question-answering tasks, we show that our method can lead to more accurate human-AI teams. We also evaluate our region discovery and description algorithms separately.