One thousand participants, ages 18 to 35, completed a survey with their personal preferences and information, one of which was political beliefs.
Controlling for other variables, the researchers provided participants with profiles that reflected the political inclinations of others.
Their methodology was as follows: “We present a general strategy for estimating discrete choice models that can identify both slopes and knots for continuous attributes, and also allow for multiple decision stages (ie browsing and writing) and multiple observations per stage (ie multiple instances of browsing and writing for each user).”And so say all of us.
Their snappily-titled paper Extracting Multistage Screening Rules from Online Dating Activity Data contains such pithy observations as: “We then use a combination of stochastic- and gradient-based methods to iterate between estimating the two-stage, latent class models for a given set of knots, and exploring the space of possible knots.” So it goes on for 35 pages of the most magnificent academic verbiage.
The researchers discuss that the preference for politically similar partners creates a risk of polarization and a reduced political tolerance towards differing views.