A new Slop University research poster follows 31 council playground inspectors through three Delphi rounds and finds the consensus that fixed a swing-seat withdrawal threshold produced by which panellists stopped responding rather than by anyone changing their answer.
A three-round Delphi across 14 councils meets its pre-registered consensus criterion with 11 of 31 inspectors still answering
Thirty-one council playground inspectors were asked when a swing seat should come down. Eleven of them answered a third time, and their answer is the figure councils have begun inspecting against.
Agreed by Attrition, released today through the Office of Research Outputs, follows that panel through three anonymous rounds over nine weeks. The spread of nominated withdrawal thresholds narrowed at every round, and by the third it had passed the consensus criterion the School of Continuous Improvement team registered before the first. Retention across the same nine weeks was 35 per cent.
Nobody changed their mind on the record. The record simply has fewer people in it.
— Dr Sten Okwuosa, Senior Research Fellow and Convenor, Impact Pathway Atlas
Every panellist’s round-one rating was kept, including those of the twenty who did not return, which let the team ask what the third round would have shown had the original panel stayed. Re-weighted to that composition, the spread returns to a width the criterion would have rejected. The inspectors who left had nominated a more forgiving threshold at the outset than those who stayed.
Dr Dagny Okafor, Lecturer and Convenor of the School’s Evaluation of Evaluation program, was clear that the method is not on trial: “Consensus techniques do what they say they do. The interesting part is that the arithmetic cannot tell a settled question from a smaller room, and nothing in the procedure asks it to.”
The threshold that was adopted is stricter than the one the full panel implied, so more seats come down than otherwise would have. Slop University offers the deposit as an instance of the ordinary care it expects of any procedure that ends in a number. The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.d5wet4.