Searching for Synergy in Shared Workspace Human-AI Collaboration

[Submitted on 16 Jun 2026 (v1), last revised 19 Jun 2026 (this version, v3)]

View PDF HTML (experimental)

Abstract:Automated AI agents are increasingly capable, yet many scientific and professional tasks require human judgment and contextual expertise. We use simulated shared-workspace human-AI teams as a controlled testbed for studying how collaboration structure shapes team behavior. Using the Collaborative Gym environment with tasks from DiscoveryBench, we vary team compositions and collaboration structures across 1,482 sessions. We find that adding additional collaborators can lower performance when coordination structure is absent. We then evaluate collaboration scaffolding that combines shared group memory with simulated human-in-the-loop (HITL) gates, where selected actions require approval from a designated simulated participant. This scaffolding improves performance, most clearly in three-person teams, with clearer responsibility signals and stronger routing of expertise to team actions. Overall, our results suggest that coordination structure is central to whether available capability improves team outcomes.

Submission history

From: Nachiket Kotalwar [view email]
[v1] Tue, 16 Jun 2026 19:08:43 UTC (608 KB)
[v2] Thu, 18 Jun 2026 02:11:35 UTC (607 KB)
[v3] Fri, 19 Jun 2026 23:47:07 UTC (607 KB)