Predicting Public Intention to Seek AI-Based Psychological Help Using Random Forest: A Survey-Based Study
Abstract
Artificial intelligence offers innovative approaches for mental health support; however, public acceptance of such technologies remains insufficiently investigated. This study employs the Random Forest algorithm to predict public intention to seek AI-based psychological assistance and to identify the key determinants influencing such behavioral intentions. Data were collected from 660 valid questionnaire responses obtained from participants across multiple regions in China, systematically measuring social-psychological characteristics, trust mechanisms, and affective experiences related to AI mental health services. The Random Forest classifier demonstrated robust predictive performance, achieving 85.3% accuracy, 91.8% F1 score, and 0.828 AUC-ROC in distinguishing between high and low usage intention groups. Feature importance analysis revealed that age, gender, and social anxiety constituted the strongest predictors of usage intention, followed by loneliness and trust tendency. While object trust and affective attraction showed significant bivariate correlations with intention, their multivariate predictive contribution was relatively limited. These findings substantiate the effectiveness of ensemble machine learning methods in elucidating the adoption patterns of AI mental health services and underscore the significance of demographic and mental health factors in technology acceptance processes. The results offer practical guidance for the development of human-centered AI mental health services that prioritize targeted outreach strategies across demographic groups.