A new paper from the School of Continuous Improvement pilots a timeout-release agent for booked-but-empty study rooms, and finds its near-miss eviction rate tracks an unrevisited installation setting more closely than any deliberate design choice.
A grace-window release agent lifts study-room use in three buildings, and quietly evicts a small, predictable slice of latecomers
A booked study room with nobody in it is, to the room, indistinguishable from an empty one. Three teaching buildings spent a term acting on that distinction anyway, continuing the School of Continuous Improvement’s habit of turning a quiet inefficiency into a measured one.
Associate Professor Casimir Beng and Dr Joost Nwosu fitted the three buildings’ study-room panels with a release agent, cancelling any booking with no check-in or motion signal within a grace window, against three matched buildings left on the University’s ordinary rules. Room use crept upward where the agent ran and stayed flat where it didn’t; a follow-up check against access logs found a smaller, steadier population of near-misses — bookers who arrived only minutes after their room went back to the waitlist.
A control system is only as good as the number nobody remembers choosing. We built the release rule; somebody else, years before us, built the number that decides who it lets through.
— Associate Professor Casimir Beng, Lead, Adaptive Metrics Lab
That number mattered more than the pilot’s design did. Two buildings run the panel’s factory-set eight-minute window; the third runs fifteen, set by an installer years ago and never revisited. The shorter window recovers marginally more room time each week and produces markedly more near-misses; dropping the motion sensor from the release decision changed almost nothing about either.
“Every deployed threshold is a decision someone made once and never revisited,” said Dr Nwosu, Postdoctoral Fellow in the School of Continuous Improvement. “We’d rather this raise that question for every timer on campus than settle it for the one we studied.”
The University has not decided whether every building should adopt the rule. For now, the pilot’s own numbers — eight-minute default included — are what it stands behind.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.dzfu3a.