A new Slop University research poster checks a charity clothing-bin network's fill-sensor alert against what its collection trucks actually bring back, finding the two drifting apart as the sensor rolls out route by route.
A staggered sensor rollout across six collection routes finds resale value falling as fill-based dispatch rises
A charity clothing-bin network dispatches its trucks off one number: how full a bin’s sensor says it is. A research poster released today by the School of Emergent Priorities asks what that number is actually reporting.
A bin that reports how full it is has already decided that fullness is the thing worth reporting. Everything else about what’s inside becomes somebody else’s problem to notice.
— Professor Verity Marris, Director, Trajectory Analytics Group
The team followed 84 street bins across six collection routes through a 22-week rollout, in which three routes converted to sensor-triggered dispatch at week six and three more at week fourteen. Bagged weight was logged at each truck-side collection; garment condition and category were graded once bags reached the sorting table. Fill-percentage tracked bagged weight only weakly and graded resale value not at all, and the routes carrying the sensor longest showed the largest fall in what a truck run brought back.
Co-author Dr Ronja Oyelaran, whose work traces how a settled arrangement holds up once nobody rereads it, said the staggered design had left the team with a cleaner question than the one they started with: not whether the sensor is wrong, but how long a wrong reading has to run before a route stops noticing.
The University counts the finding among the clearer proofs of a caution its research agenda keeps returning to: an alert built to flag one property of a thing will be read, sooner or later, as a report on all of them.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.66co03.