School of Emergent Priorities researchers fit a paired-comparison model to what leaves a community produce swap's trestle table, and find the ranking's month-to-month stability comes from the layout it sets rather than from what anyone at the table wants.
Eleven monthly swaps, 2,148 take-or-leave decisions, and a rank order that only holds where it decides the layout
A monthly community produce swap now sets out its trestle table in an order recalculated after every swap and lodged with the Indicator Commons as the swap’s desirability statement. Where that order comes from is what a new research poster from the School of Emergent Priorities set out to establish.
A Bradley–Terry Model of the Produce Swap Table works from 2,148 decisions to take or leave an item, logged by trestle-side observers across eleven monthly swaps, and fits to them the paired-comparison model that has ranked competitors against one another since the 1950s. The 34 catalogued item types were then split: half were set out in the order the model gave, half were set out at random, and the allocation was held for the series.
Under the ordered layout the model agreed with itself from one month to the next almost perfectly. Under the random layout it barely agreed with itself at all. A place at the near end of the table proved worth more, in the odds of an item being taken, than the entire spread of desirability the model was measuring.
A table has to be set out in some order, and there is no order that says nothing. We would like to know what the other orders say.
— Dr Petra Umbile, Postdoctoral Fellow, School of Emergent Priorities
The University is glad to see a question of this weight put to a setting where nothing whatever is at stake, and regards the work as characteristic of the School’s interest in measures that reshape the settings they are drawn from. Dr Ronja Oyelaran, Senior Lecturer, said the design should transfer to “any arrangement where the thing being measured and the thing being arranged are the same thing”.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.unlxck.