[2509.12626] DoubleAgents: Human-Agent Alignment in a Socially Embedded Workflow

Computer Science > Human-Computer Interaction arXiv:2509.12626 (cs) [Submitted on 16 Sep 2025 (v1), last revised 6 Apr 2026 (this version, v3)] Title:DoubleAgents: Human-Agent Alignment in a Socially Embedded Workflow Authors:Tao Long, Xuanming Zhang, Sitong Wang, Zhou Yu, Lydia B Chilton View a PDF of the paper titled DoubleAgents: Human-Agent Alignment in a Socially Embedded Workflow, by Tao Long and 4 other authors View PDF HTML (experimental) Abstract:Aligning agentic AI with user intent is critical for delegating complex, socially embedded tasks, yet user preferences are often implicit, evolving, and difficult to specify upfront. We present DoubleAgents, a system for human-agent alignment in coordination tasks, grounded in distributed cognition. DoubleAgents integrates three components: (1) a coordination agent that maintains state and proposes plans and actions, (2) a dashboard visualization that makes the agent's reasoning legible for user evaluation, and (3) a policy module that transforms user edits into reusable alignment artifacts, including coordination policies, email templates, and stop hooks, which improve system behavior over time. We evaluate DoubleAgents through a two-day lab study (n=10), three real-world deployments, and a technical evaluation. Participants' comfort in offloading tasks and reliance on DoubleAgents both increased over time, correlating with the three distributed cognition components. Participants still required control at points of uncer...