Recreational Activities and Youth Outcomes: An Explainable Machine Learning Study

Area of Science:

• Neuroscience
• Developmental Psychology
• Behavioral Science

Background:

• Recreational activities are crucial for adolescent emotional and cognitive development.
• Limited empirical research exists on the specific links between youth recreational activities and behavioral/cognitive outcomes.

Purpose of the Study:

• To quantify the contribution of various recreational activities to behavioral and cognitive outcomes in youth.
• To identify data-driven groups of activities associated with distinct outcomes.

Main Methods:

• Employed interpretable machine learning (random forest and SHAP analysis) on Adolescent Brain Cognitive Development (ABCD) Study data (ages 9-10).
• Defined activity groups based on SHAP contribution patterns.
• Utilized linear mixed-effects models to assess associations in a hold-out sample (cross-sectional and longitudinal).

Main Results:

• Identified 14 activities with favorable and 8 with less favorable behavioral patterns; 13 with favorable and 2 with less favorable cognitive patterns.
• Favorable behavioral activities (e.g., skiing) correlated with fewer problems; less favorable (e.g., martial arts) correlated with more problems.
• Less favorable cognitive activities (e.g., football) linked to lower scores; favorable (e.g., musical instruments) linked to higher scores, moderated by family income.

Conclusions:

• Recreational activities show distinct associations with youth cognitive and mental health.
• Findings underscore the importance of individual differences and context in youth recreational activity engagement.