A new paper from the School of Emergent Priorities audits Slop University's share-scooter corral occupancy sensor against a manual headcount across nine weeks, finding the reading behind next term's bike-rack capacity decisions tracks scooter dwell time as much as how many scooters are actually parked.
An unannounced headcount audit of 15 share-scooter corrals finds the sensor deciding next term's bike-rack capacity tracks scooter dwell time as much as it tracks scooters
Next term’s bike-rack capacity expansion is decided corral by corral, and the number deciding it comes from a sensor nobody had checked against a person standing there counting. A new paper from the School of Emergent Priorities does the counting.
Dr Lindiwe Achterberg and Associate Professor Casimir Beng set the University’s share-scooter corral occupancy sensor against an unannounced manual headcount across nine weeks and 15 of the University’s 21 registered corrals, timing 312 physical counts against the sensor’s own reading for the same minute. The two agreed only moderately, and the gap wasn’t random: corrals where scooters tend to sit longest read persistently busier to the sensor than a headcount would justify, while corrals with fast turnover read quieter than the queue at their kerb would suggest.
A sensor that has never been wrong yet isn’t the same as a sensor that has been checked. Ours hadn’t been, until now.
— Lindiwe Achterberg, Lecturer, School of Emergent Priorities
Associate Professor Casimir Beng, whose Adaptive Metrics Lab studies measurement systems and what happens when they stop tracking what they were built to track, called the dwell-time pattern “the more interesting number to have found,” noting that an instrument built to flag whether a bay is occupied at all will always have an opinion about how long, whether or not the capacity review asked it one. The School reads the finding as consistent with its wider interest in how a number gets trusted well past the point anyone last checked it.
Facilities has not revised the capacity-review process. The pair say a headcount-weighted supplement, even an occasional one, is what they would like to see trialled ahead of the network’s next allocation round.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.gdacz4.