Two coders and a six-category scheme show how much detail Slop University's library noise sensor loses on its way into a single compliance tickbox, from a three-second mechanical trip to a ninety-second conversation.
A new inter-rater reliability study recovers the six causes the Library's silent-floor tickbox flattens into one
The School of Continuous Improvement has spent the better part of a year teaching its own instruments to explain themselves. Its newest paper turns that instinct on a fixture already wired for it: the traffic-light sensor mounted above every silent-study floor in the University Library.
The sensor reports three states, and Facilities keeps one record of the red ones: a “noise complaint” tickbox, ticked whenever the light turns red, with no field for what turned it. The University reads the new result as confirmation that its instruments have more to say than the record currently asks of them.
Working from six weeks of sensor logs across three floors, two researchers built and tested a six-category scheme for what actually turns the light red, then checked how much of that detail survives into the single line Facilities keeps for each event. Agreement between the two coders came back strong; the survival rate into the existing record did not.
Six seconds of plant noise and ninety seconds of conversation get exactly the same word in our records. We wanted to know how much detail a return to reading the sensor properly might buy back.
— Dr Dagny Okafor, Lecturer and Convenor, Evaluation of Evaluation
“It’s a small scheme, and it says nothing about what the University should do with it,” said Dr Thandiwe Solberg, Senior Lecturer and Convenor of Demo Quarter. “That’s a separate question, and a fair one for someone else to ask.”
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.bmm3u3.