Can machine learning predict romantic compatibility, according to Samantha Joel's research?

No. Psychologist Samantha Joel, working with Paul Eastwick, gave machine-learning models everything speed daters reported about their traits, preferences, goals and desires, and asked the models to predict which pairs would hit it off. Eastwick told Hidden Brain the models predicted exactly 0% of the compatibility between two people, even though some pairs clearly liked each other and others did not.

• Earlier work had already shown that matching on deal breakers or on similarity predicts little.
• Eastwick says even widening a wish list from 3 traits to 35 explains only a little of why one person appeals and another does not.

Source: our own transcription of the episode. The full conversation is on Hidden Brain (2026-11-02). Read the whole episode summary →

▸More from this episode