A new poster from the School of Continuous Improvement audits eighteen months of reflection-room visits across nine halls of residence and finds visit length carries no reliable signal about whether a resident reappears in next month's incident log.
An 18-month audit across nine halls finds reflection-room visit length carries no signal about next month's incidents
The School of Continuous Improvement keeps a standing interest in numbers that outlive the question they were built to answer. Its newest poster turns that interest on a fixture of residential life: the reflection room every hall of residence keeps for de-escalating a conflict, and the minutes logged there every time it’s used.
Dr Mirela Hanke and Dr Joost Nwosu matched eighteen months of reflection-room visit records against the incident register each hall keeps for the following month, testing whether time spent in the room said anything about what happened to a resident afterward. Across 214 visits, residents who reappeared in next month’s incident log spent about as long in the room as residents who didn’t — a median of eighteen minutes against seventeen, a gap the pair’s analysis could not distinguish from none at all.
The School has spent a decade asking what an indicator is quietly measuring instead of its stated target. This one, it turns out, is mostly measuring itself.
— Dr Mirela Hanke, Postdoctoral Fellow, School of Continuous Improvement
Dr Nwosu, whose own work follows a performance target once it has been met, said the finding fits a pattern the School keeps rediscovering in its own record: “A number can keep climbing the reporting chain long after anyone has checked what it’s supposed to be climbing toward.” The School counts the null result among its more useful of the term, confirmation that its own instruments are still worth doubting. Residential life continues to log reflection-room minutes under the existing protocol; no replacement measure has been proposed.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.6yvb0r.