A new Slop University paper surveys 487 gig food-delivery riders, finding that distrust in a bundled efficiency score predicts a higher rate of self-reported on-road shortcuts, a pattern that holds after adjusting for tenure, weekly hours, and weather.
A survey of 487 riders finds distrust in the score, not its arithmetic, best predicts who cuts corners on shift
Four weeks of QR codes at eleven delivery staging hubs, 487 completed surveys, and one four-item distrust scale later, the School of Continuous Improvement has a number of its own to report: riders who trust a delivery app’s bundled efficiency score the least are the same riders most likely to say they cut a corner on shift.
A score that folds speed, acceptance, and lateness into one figure is very good at settling an argument and very bad at explaining itself, and riders seem to notice the difference well before we did.
— Dr Torun Ezeigwe, Senior Lecturer and Director, Master of Applied Measurement, School of Continuous Improvement
The paper surveyed 487 active riders recruited at eleven metropolitan staging hubs, pairing a short instrument measuring distrust in the app’s efficiency score against six commonly reported on-road shortcuts, from an amber light run to an unclipped helmet strap. Distrust tracked a higher rate of reported shortcuts even after the University’s usual covariates, tenure, weekly hours, and a weather-adjusted-shift term, were added to the model, and the specific shortcut riders leaned on shifted with experience, moving from an on-screen habit among newer riders to an on-road one among veterans.
“The most useful thing the score could do for anyone is stop pretending it’s one thing,” said Dr Marit Osayande, Lecturer and Convenor of the Living Dashboard. “We’d welcome a version of this work that follows a rider for longer than a single shift and asks what happens once they stop trusting the number at all.”
The work reflects the School’s ongoing interest in what a bundled indicator quietly asks of the people it scores. The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.xoebw1.