About the Bridge
An agent proposes to commit a week of beam time at a user facility. It selected the experiment from evidence distributed across three institutions, none of which may share its raw data. It has been fine-tuned twice since anyone examined its behavior. A staff scientist has ten minutes to decide whether to let it proceed, and the operation is irreversible.
Autonomous experimentation has crossed from demonstration into operation. Agents now select, schedule, execute, and interpret experiments at shared scientific facilities, and the question of what evidence justifies allowing them to act has become operational rather than aspirational. Yet the two communities who must answer it work almost entirely apart: AI researchers who develop assurance methods, and the experimental scientists, instrument staff, and facility operators who carry responsibility for irreversible physical decisions. This bridge brings them into the same room around one question — what must be true, and what must be recorded, before an autonomous agent is permitted to act on a scientific instrument, and how does that change as the agent, the facility, and the threat landscape keep changing?
Open Questions
Format
A one-day program combining education, collaboration, and outreach, held in person at AAAI-27. The morning introduces each community to the other through paired tutorials, an invited keynote, and lightning talks; the afternoon turns to contributed talks, a panel — “What would it take to let an agent run your instrument?” — and breakout groups mapped to the five open questions; the closing session synthesizes a draft community roadmap.
Call for Participation
We solicit non-archival contributions in AAAI-27 two-column format (references excluded from page limits):
- Full papers (up to 8 pages)
- Short papers (up to 4 pages) — work in progress, negative results
- Position and blue-sky papers (2–4 pages)
- Extended abstracts (up to 2 pages), including previously published work, for posters
- Challenge-problem proposals (1–2 pages)
- Experience reports (2–4 pages) from facility staff and practitioners
Submission site (OpenReview) and deadlines: to be announced.
Organizing Committee
- Olivera Kotevska Oak Ridge National Laboratory Contact
- Sumit Kumar Jha University of Florida
- Plamen Angelov Lancaster University
- Philippe Schwaller EPFL
- A. Gilad Kusne University of Maryland & NIST
Contact
For questions about the bridge, contact kotevskao@ornl.gov.