Following Delegation Traces to Study Human Oversight in Interaction with Multimodal AI
As multimodal AI systems are becoming our everyday thought partners, the shift towards studying how users transfer control to these systems becomes especially urgent.
ProofPaper ↗
Key points
- In this paper, we report an overview of findings from an online study with n=56 AI-literate participants and 392 Video Question Answering trials.
- After inspecting each multimodal AI response, participants rated task-specific trust and selected one of three intended policies for an analogous future task: reject AI assistance, use it with verification, or delegate fully to the AI.
- These findings motivate a shift from intended delegation to a non-intrusive observation of how delegation is enacted in collaborative human-AI tasks.
- We therefore conclude by proposing enacted delegation traces as a methodological framework for future studies of multimodal and agentic AI use.
Sources (1)
- [1]Following Delegation Traces to Study Human Oversight in Interaction with Multimodal AIarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 15, 02:18 PM
As multimodal AI systems are becoming our everyday thought partners, the shift towards studying how users transfer control to these systems becomes especially urgent.
In this paper, we report an overview of findings from an online study with n=56 AI-literate participants and 392 Video Question Answering trials.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 15, 2026A team of Gemini 4 Argon agents beats Minecraft
- Oct 15, 2026From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research
- Oct 14, 2026Analyzing technical documents with Deep Review
- Oct 14, 2026Nvidia's Jensen Huang pledges $2 billion for Europe, $250 million in compute for UK safety labs
- Oct 14, 2026Amp puts AI agents to work beyond code with OpenAI models
- Oct 13, 2026Building agents that transact with Anchor Browser and AgentCore payments