A new Slop University research poster validates a community choir's automated seating-balance score against an independent panel's blend ratings, and finds the two don't move together.
A season-long validation across 18 community choirs finds an automated seating-balance score untethered from how a blind panel rates the choir's own blend
Eighteen community choirs spent a season being reseated by an algorithm that had never heard them sing. Slop University’s newest research poster asks whether the tidier roster it produced tracks anything about how the choir actually sounded, and reports back with a scatterplot that never quite finds a line.
The School of Continuous Improvement’s Adaptive Metrics Lab keeps returning to instruments that satisfy themselves before anyone checks whether they satisfy anything else; a seating-balance score run against an independent ear is its most literal test yet.
A score that climbs all season while the sound doesn’t move isn’t a contradiction — it’s an instrument telling you exactly what it’s actually attached to.
— Lecturer Marit Osayande, Convenor of the Living Dashboard, School of Continuous Improvement
Eighteen suburban choirs took part across a fourteen-week season, half of them handing weekly seating decisions to the tool and half sticking with whatever arrangement their singers settled into on their own. A panel of six listeners, blind to which arrangement produced which recording, rated a sample of the season’s rehearsals on how well the parts blended.
Senior Lecturer Renke Sabel, Convenor of the Indicator Commons, placed the finding inside a pattern the School keeps rediscovering: “the fastest way to find out what a balance score is really tracking is to put an ear that’s never seen it in the room.”
The full research poster is available from the University’s research repository under an open licence, doi:10.5555/slop.bx5gb5.