A School of Emergent Priorities research poster benchmarks a twelve-member route ensemble against 31 observed ice-cream van rounds, finding forecast skill collapses to guesswork 8.4 minutes after the first chime while the residents' association publishes a nightly outlook regardless.
A verandah listening network gives one suburban street eight minutes of warning, and finds more microphones cannot buy a ninth
For fourteen weekends, forty-four verandahs on one suburban street listened for the same chime. A research poster released today by the School of Emergent Priorities reports what all that listening can and cannot buy: a forecast of the ice-cream van worth having at two minutes’ lead, and worth nothing at nine.
How Far Ahead Can You Hear the Ice-Cream Van?, by Dr Fenna Okoro, Dr Ingrid Vasseur and Dr Osei Vandermeer, benchmarks a twelve-member route ensemble — fed by chime bearings from a residents’ association’s verandah-mounted acoustic network — against the simple strategies any household already runs. Inside five minutes the ensemble is clearly ahead; at a mean lead of 8.4 minutes its skill reaches zero, and doubling the number of listening verandahs moved that horizon by eleven seconds.
The work harmonises household-scale sensing with the School’s anticipatory-capability agenda, carrying a question usually reserved for national forecasting infrastructure — how far ahead the future can be bought — to the kerb. For the University, the poster affirms a conviction that anticipation is a civic capability before it is a technical one.
There is a horizon past which more microphones only buy you a better record of being surprised. We think the sector has something to learn from sitting with that.
— Dr Fenna Okoro, Senior Lecturer and Convenor, Horizon Register
“What we heard, in the end, was the neighbourhood listening to itself,” said Dr Osei Vandermeer, Convenor of the Review Cadence Observatory.
The association’s nightly Van Outlook remains on the noticeboard. The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.s7ul28.