A new Slop University paper interviews 58 dating-app users about when they stop trusting the app's compatibility percentage, and finds its highest-scoring matches keep resurfacing in the queue long after users have stopped opening them.
Fifty-eight interviews locate the message where trust in the percentage lapses, and find its top-scoring matches lingering in the queue long after
By the eleventh message, most of the daters interviewed for a new Slop University paper say the number that got them talking has already stopped mattering. The paper, from the School of Continuous Improvement, asked 58 users of a widely used dating application to walk back through one match in detail, message by message, and name the point at which its displayed compatibility percentage stopped shaping how they read the conversation.
Lecturer Marit Osayande and Senior Lecturer Runa Adegoke coded each account against the point it described, then checked the pattern against a smaller group’s own record of their match history — 41 participants, 1,842 conversations — to see whether a high compatibility score bought a conversation any more time than a middling one did. It didn’t: matches scored above ninety percent lasted about as long, message for message, as matches scored in the seventies.
A percentage only has to do its job once, at the very top of the funnel; nobody warned it that the chat was going to keep going.
— Lecturer Marit Osayande, Convenor of the Living Dashboard, School of Continuous Improvement
“What stayed with me wasn’t that people stopped trusting the number,” said Senior Lecturer Runa Adegoke. “It’s that none of them seemed to think it was worth mentioning to the app.”
For a School that has spent this run of studies on what a live measurement does once the people it was built to inform have stopped consulting it, a percentage that keeps resurfacing at the top of someone’s queue long after they’ve mentally filed the match away reads less like an anomaly than like the pattern’s cleanest version yet.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.rngdik.