Dimensions of the Dating Mind

OKCupid used to let users answer thousands of short match questions ("Do you smoke?", "Could you date someone who is really messy?", "Is it wrong to burn your country's flag?") and matched people on the answers. This test takes 200 of them, chosen from a public dataset of about 68,000 OKCupid profiles collected between 2009 and 2015 (Kirkegaard & Bjerrekær, 2016), and scores you on the ten dimensions that ran through those answers.

The dimensions were extracted with a multidimensional item response model fitted to the daters' answers. Since users chose which questions to answer, the source data is mostly blank — a typical dater answered only a fraction of these items — so the model was fitted from partial records and your score is a posterior estimate that accounts for which items you answered.

Percentiles are relative to the OKCupid daters, model-implied: where you would stand among them had they all answered every question. The dataset holds about 68,000 profiles, but most of them answered only a handful of these questions: a typical item was answered by around 30,000 users, and about 32,000 answered at least 100 of the 200. Those are the daters the norms effectively rest on: mostly American and British, median age 30, about 60% men, so "average" here means average for that crowd, not for the general population. Since users chose which questions to answer, the ones who answered most of these items differ a little from the rest (they report somewhat more sexual experience and less interest in fitness, by about a third of a standard deviation), and the norms inherit that. Several of the dimensions were relabelled after inspecting which items actually load on them; the pole descriptions under each bar list the items that define it.