From Researcher to Orchestrator: Attention-Bounded Supervision of AI Agent Teams in Engineering Research
Authors/Creators
Description
Researchers in electrical and computer engineering increasingly delegate literature search,
coding, data analysis and drafting to AI agents, yet practices for supervising several agents at
once lag behind adoption, and the researcher remains accountable for every claim. Existing
multi-agent frameworks and levels-of-automation models do not tell a researcher how much
autonomy to grant a task, how work should pass between agents, or how many agents one person
can supervise. This paper presents an orchestration framework that answers these questions.
Each task’s autonomy level is set from three properties of its effects (reversibility, reach and
verifiability) by a rule that is complete and monotone. Work passes between agents as version-
controlled files that no agent may mark as accepted when it produced them, and every reported
number must name the computation that produced it. A review-load model bounds team size by
the researcher’s daily attention and shows that the ratio of planning to execution agents follows
from their review costs rather than being fixed. We apply the framework to an illustrative analog-
to-digital converter characterization task and to two AI-assisted publication projects, a 20-
chapter textbook and an arXiv paper, whose logs and repositories record 13 defects. Mechanical
checks caught every defect for which one existed, errors of meaning were caught only when
someone recomputed or read the work, and a quality score that no computation had produced
escaped every control and was published. The results suggest measuring review load and escaped
defects before enlarging an agent team.
Files
main.pdf
Additional details
Dates
- Available
-
2026-10-23Initial presentation at the University of Minnesota Workshop on AI in Research
Software
- Repository URL
- https://dmccreary.github.io/ai-for-ee-research/
- Development Status
- Active