The results are striking: the model explains most of the variation in who people choose and how fast they decide. When single traits were presented, the theory held up. When multiple traits were combined, people used a biased averaging process where the single most positive feature influences decisions more than other traits. That pattern suggests partner choice draws on a broad decision system that evaluates and integrates value signals rather than a specialized mate-selection module.
For anyone curious about human potential, these findings matter because they link everyday social decisions to general cognitive machinery. Understanding the algorithms our brains rely on could improve how we design social platforms, relationship education, or interventions that promote inclusive pairings. Follow the link to see how a computational lens reframes attraction and what that implies for real-world relationships.
Abstract
Mate value is theorized to be a key driver of romantic partner choice, yet the cognitive mechanism underlying romantic partner choice remains poorly understood. Here, we assess whether initial romantic partner choice can be predicted by a general-purpose, computational cognitive model of value and choice. To do so, we enlist Psychological Value Theory (PVT), which predicts both choice and decision reaction time (RT) simultaneously. To establish the scientific controls necessary to test PVT’s strong a priori predictions, we use discrete choice experiments. Experiment 1 tested choices between partners described by single features, while Experiment 2 investigated how the values of multiple features are integrated. PVT’s predictions were highly accurate across both experiments, accounting for over 85% of the variance in choice and RT for groups and individuals. Critically, Experiment 2 revealed that people integrate multiple features via a Biased Average algorithm, where the most positive feature holds disproportionate influence. These findings indicate that initial romantic partner choice recruits a general-purpose, value-based decision mechanism, providing a computational framework that can be extended to model partner choice in more complex, real-world contexts.
