New research using machine learning suggests that relationship characteristics are more important than either an individual’s or their partner’s traits when it comes to predicting couples' relationship satisfaction. The findings are published this week in the Proceedings of the National Academy of Science.
An enormous team of 86 researchers shared datasets with lead authors Samantha Joel and Paul Eastwick. Joel and Eastwick then used machine learning to analyze more than 43 datasets containing more than 11,000 couples to find out which factors are most strongly associated with couples’ perceptions of relationship quality. Data were included in the machine learning analysis if they were collected longitudinally, if they contained responses from both members of a couple, and if they contained a measure of relationship satisfaction. The researchers investigated more than 2,000 variables including relationship-related variables such as sexual satisfaction, conflict, and investment and individual difference variables such as negative affect, attachment styles, and stress. The researchers also excluded some variables from consideration because they were potential alternate indicators of relationship quality such as love, trust, and intimacy.
The researchers used a machine learning model which “tests the strength of each available predictor one at a time…and [then] tests the…overall predictive power [of the model] on a subset of data…[to] reveal how much variance in the dependent measure was predictable and which predictors made the largest contributions to the model.”
According to the researchers, the five relationship-related variables which most strongly predicted relationship satisfaction and quality were perceived partner commitment, appreciation of one’s partner, sexual satisfaction, perceived partner satisfaction, and reduced conflict. Taken together, the relationship-related variables predicted as much as 45% of the variability in relationship quality. The individual difference variables with the strongest association with relationship quality were life satisfaction, negative feelings, depression, and both attachment anxiety and avoidance. Taken together, the individual difference variables explained as much as 21% of the variability in relationship quality. However, individual difference variables did not improve the model’s prediction once relationship-related variables were considered. Importantly, although relationship-related variables were strongly associated with relationship quality at the time of the baseline measurements for couples, these variables were not useful in predicting whether couples’ relationship quality would change in a positive or negative direction over time.
The researchers acknowledge that the couples all come from Western countries (the U.S., Canada, Switzerland, New Zealand, The Netherlands, and Israel) and that future research should explore the associations among these variables in other cultures. The authors conclude that “a relationship characterized by appreciation, sexual satisfaction, and a lack of conflict” matters more than individual characteristics of either member of the couple when predicting perceived relationship quality.
Facebook image: oneinchpunch/Shutterstock
Joel, S., Eastwick, P. W. et al. (2020). Machine learning uncovers the most robust self-report predictors of relationship quality across 43 longitudinal couples studies. Proceedings of the National Academy of Sciences Jul 2020, 201917036; DOI: 10.1073/pnas.1917036117