MILAN, Italy — Are you ready for an AI system to predict your next breakup? It may sound farfetched, but Italian scientists have used a complex machine learning process to name the top two “most important predictors of union dissolution” among over 2,000 married or cohabiting couples.
Those factors are: The life satisfaction of both partners and the female partner’s percentage of housework.
This research, conducted at Bocconi University, comes from data originally collected on 2,038 couples for the German Socio-Economic Panel Survey. On average, researchers tracked those couples for 12 years, making a total of 18,613 observations. Over that period, 914 couples (45%) broke up.
Study authors used a machine learning technique called Random Survival Forests (RSF) to sift through all the data and come to some meaningful conclusions while still accounting for numerous independent variables.
“A clear-cut example of the potential difficulties of considering all variables and their possible interactions concerns the ‘big five’ personality traits,” says Professor Letizia Mencarini in a media release. “To account for both partners’ traits (10 variables) and all their two-way interactions (25 variables), one would need to include 35 independent variables, which would be very problematic in a regression model.”
Happy husband, happy life?
Machine learning is advantageous due to its superior predictive power in comparison to more conventional models, which tend to focus on explanations over predictions. When the research team divided their sample into two parts and analyzed the first portion to predict the outcomes of the second half, it quickly became apparent that the predictive accuracy of RSF was much better than conventional models.
However, RSF’s predictive accuracy still had limitations despite study author using all the “union dissolution predictors” they considered especially important. These predictors were factored in as “input variables” during calculations.
Variables with the greatest predictive ability included life satisfaction of both partners, the woman’s percentage of housework, general marital status, the woman’s working hours, the woman’s level of openness, and the male’s extraversion levels.
Plenty of variables also appear to interact with one another in complex ways. For example, if the man in the relationship is feeling good and highly satisfied with his life, increases in his female partner’s life satisfaction always increased their union’s chances of surviving. Unfortunately, if the man’s life satisfaction is low, his female partner’s life satisfaction stops protecting the relationship past a certain threshold.
Interestingly, personality traits among relationship partners don’t appear to interact in terms of breakup risk. Both a woman’s openness and a man’s extraversion make union dissolution more likely to occur, regardless of their significant other’s personality.
The study is published in the journal Demography.