Who benefits most from couple relationship education: A machine learning approach

Area of Science:

• Psychology
• Relationship Science
• Machine Learning Applications

Background:

• Couple relationship education (CRE) aims to improve relationship functioning but shows mixed results.
• Intervention effectiveness may depend on couple characteristics, necessitating better targeting.
• Previous research overlooked complex pretreatment risk factors, leading to inconclusive findings.

Purpose of the Study:

• To identify pretreatment characteristics predicting differential treatment effects in CRE.
• To leverage machine learning for personalized intervention strategies in relationship education.

Main Methods:

• Applied causal forest, a machine learning technique, to analyze data from two randomized controlled trials of CRE.
• Trained models on data from 6,298 couples and validated on 1,595 couples.

Main Results:

• Pretreatment psychological distress and low relationship happiness predicted greater improvements in relationship quality.
• High psychological distress and perceived stress predicted reductions in negative relationship emotions and behaviors.
• Findings were robust and validated on an independent dataset.

Conclusions:

• CRE treatment effects are heterogeneous, varying by individual characteristics.
• Machine learning effectively identifies couples most likely to benefit from CRE.
• This approach can inform more targeted and effective relationship interventions.