Artificial Intelligence Modeling of Risk-Taking Behavior: Contributions of Sensation Seeking, Delay Discounting, Emotional Dysregulation, and Peer Influence

Objective: The present study aimed to develop and evaluate an artificial intelligence-based model for predicting risk-taking behavior based on sensation seeking, delay discounting, emotional dysregulation, and peer influence.Methods and Materials: This cross-sectional predictive study was conducted on ۵۱۲ young adults aged ۱۸ to ۳۰ years in Canada, selected through stratified convenience sampling. Data were collected using validated psychometric instruments, including the Domain-Specific Risk-Taking Scale (DOSPERT), Brief Sensation Seeking Scale (BSSS), Delay Discounting Task, Difficulties in Emotion Regulation Scale (DERS), and Resistance to Peer Influence Scale (RPI). After preprocessing procedures such as normalization and missing data imputation, both traditional statistical analysis and machine learning approaches were applied. Multiple regression analysis was used to examine linear relationships, while machine learning models including Random Forest, Support Vector Machine, and XGBoost were implemented using ۱۰-fold cross-validation. Model performance was evaluated using accuracy, precision, recall, F۱-score, and AUC-ROC, and SHAP analysis was employed to interpret feature importance.Findings (inferentials only): The regression model was statistically significant (F(۴, ۵۰۷) = ۱۲۸.۶۴, p < ۰.۰۰۱), explaining ۵۰.۳۸% of the variance in risk-taking behavior. Sensation seeking (β = ۰.۴۱, p < ۰.۰۰۱), peer influence (β = ۰.۳۴, p < ۰.۰۰۱), emotional dysregulation (β = ۰.۲۷, p < ۰.۰۰۱), and delay discounting (β = ۰.۲۲, p < ۰.۰۰۱) were all significant predictors. Among machine learning models, XGBoost demonstrated the highest performance (accuracy = ۰.۸۷, AUC-ROC = ۰.۹۲), followed by Random Forest and Support Vector Machine. SHAP analysis confirmed sensation seeking as the most influential predictor, followed by peer influence, emotional dysregulation, and delay discounting.Conclusion: The findings indicate that risk-taking behavior can be effectively predicted using an integrative artificial intelligence framework that captures the combined effects of dispositional, cognitive, emotional, and social factors, with sensation seeking and peer influence were the most influential determinants. Objective: The present study aimed to develop and evaluate an artificial intelligence-based model for predicting risk-taking behavior based on sensation seeking, delay discounting, emotional dysregulation, and peer influence. Methods and Materials: This cross-sectional predictive study was conducted on ۵۱۲ young adults aged ۱۸ to ۳۰ years in Canada, selected through stratified convenience sampling. Data were collected using validated psychometric instruments, including the Domain-Specific Risk-Taking Scale (DOSPERT), Brief Sensation Seeking Scale (BSSS), Delay Discounting Task, Difficulties in Emotion Regulation Scale (DERS), and Resistance to Peer Influence Scale (RPI). After preprocessing procedures such as normalization and missing data imputation, both traditional statistical analysis and machine learning approaches were applied. Multiple regression analysis was used to examine linear relationships, while machine learning models including Random Forest, Support Vector Machine, and XGBoost were implemented using ۱۰-fold cross-validation. Model performance was evaluated using accuracy, precision, recall, F۱-score, and AUC-ROC, and SHAP analysis was employed to interpret feature importance. Findings (inferentials only): The regression model was statistically significant (F(۴, ۵۰۷) = ۱۲۸.۶۴, p < ۰.۰۰۱), explaining ۵۰.۳۸% of the variance in risk-taking behavior. Sensation seeking (β = ۰.۴۱, p < ۰.۰۰۱), peer influence (β = ۰.۳۴, p < ۰.۰۰۱), emotional dysregulation (β = ۰.۲۷, p < ۰.۰۰۱), and delay discounting (β = ۰.۲۲, p < ۰.۰۰۱) were all significant predictors. Among machine learning models, XGBoost demonstrated the highest performance (accuracy = ۰.۸۷, AUC-ROC = ۰.۹۲), followed by Random Forest and Support Vector Machine. SHAP analysis confirmed sensation seeking as the most influential predictor, followed by peer influence, emotional dysregulation, and delay discounting. Conclusion: The findings indicate that risk-taking behavior can be effectively predicted using an integrative artificial intelligence framework that captures the combined effects of dispositional, cognitive, emotional, and social factors, with sensation seeking and peer influence were the most influential determinants.

کلیدواژه ها:

risk-taking behavior ، sensation seeking ، delay discounting ، emotional dysregulation ، peer influence