Explainable AI Identification of Protective Family Factors Against Adolescent Substance Abuse
Applied Family Therapy Journal · 2025 · 6 (6) · 1-10
DOI: 10.61838/kman.aftj.4957 ↗Kurz gefasst
Fachartikel in der Zeitschrift Applied Family Therapy Journal, 2025. Verfasst von Rebecca Collins und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Objective
The objective of this study was to employ explainable artificial intelligence to identify and quantify the most influential protective family factors associated with reduced risk of adolescent substance abuse among adolescents. Methods and Materials: This study adopted a cross-sectional predictive-analytic design involving 684 adolescent–caregiver dyads recruited from public schools, community youth centers, and family health clinics across Michigan. Adolescents aged 13–18 years and their primary caregivers completed a comprehensive battery of validated psychosocial assessments measuring substance use behaviors, parental monitoring, emotional warmth, family cohesion, quality of communication, parental norms against substance use, conflict resolution skills, and household stability. Data were analyzed using advanced supervised machine learning algorithms, with gradient boosting selected as the optimal model based on performance indices. Explainable artificial intelligence techniques, including SHAP analysis, were applied to interpret model outputs and identify the relative contribution and interaction of protective family factors.
Findings
The final model demonstrated high predictive accuracy (AUC = 0.93; F1-score = 0.87), indicating strong discriminative ability in identifying adolescents at reduced risk of substance abuse. Parental monitoring emerged as the most influential protective factor, followed by emotional warmth, quality of parent–adolescent communication, family cohesion, and parental norms against substance use. Significant interaction effects were observed, particularly between parental monitoring and emotional warmth, yielding a 34.6% reduction in predicted substance abuse risk. Nonlinear patterns revealed threshold effects whereby moderate improvements in core family processes produced substantial decreases in risk probability.
Conclusion
The findings demonstrate that explainable artificial intelligence provides powerful and interpretable insights into the complex family mechanisms protecting adolescents from substance abuse.
Zitieren
APA
Collins, R., & Anderson, M. (2025). Explainable AI Identification of Protective Family Factors Against Adolescent Substance Abuse. Applied Family Therapy Journal, 6(6), 1-10. https://doi.org/10.61838/kman.aftj.4957
Vancouver
Collins R, Anderson M. Explainable AI Identification of Protective Family Factors Against Adolescent Substance Abuse. Applied Family Therapy Journal. 2025;6(6):1-10. doi:10.61838/kman.aftj.4957
BibTeX
@article{Collins2025,
author = {Collins, Rebecca and Anderson, Michael},
title = {Explainable AI Identification of Protective Family Factors Against Adolescent Substance Abuse},
journal = {Applied Family Therapy Journal},
year = {2025},
volume = {6},
number = {6},
pages = {1--10},
doi = {10.61838/kman.aftj.4957}
}
EUVIMED