Patients Are Waiting: From Swarms to Synergy. Designing Complementary Collaborative Human–AI Agent Teams for Biomedical Discovery

Biomedical discovery begins with questions that matter to people and advances through the collective work of making evidence understandable and actionable. For patients awaiting effective treatments, the promise of AI lies in helping research communities turn more of those questions into validated progress. Human–AI agent teams could widen exploration, connect complementary perspectives, and sustain inquiry across experiments. Realizing this promise requires intentional design that keeps human agency, creativity, critical thinking, and meaning-making paramount. We propose a governed scientific record as the foundation for this collaboration: shared memory that distinguishes observations from inferences, preserves evidence and uncertainty through reuse, and makes challenge and correction effective across agent swarms. Oncology and rare-disease missions illustrate how patient priorities, human judgment, and agent exploration meet experimental validation. A dynamic evaluation harness combines sealed tasks, prospective discovery, and a matched-budget promotion-and-repair experiment. It tests synergy against human-only and agent-only workflows while assessing retained human capacities. This position paper offers an architecture and evaluation agenda for complementary collaboration, not demonstrated discovery gains.