Optimizing Behavioral Strategy Selection during Cognitive Behavior Therapy for Procrastination using Machine Learning
Abstract
Objective: To determine whether machine-learning models using baseline self-report data could personalize selection between two behavioral strategies embedded within cognitive-behavioral therapy (CBT) for procrastination.
Methods: Participants were 256 Polish-speaking university students aged 18–26 years (M=22.4; 73.0% women) who completed active treatment in two preregistered randomized trials (development sample, N=173; external validation sample, N=83). Participants received five weekly group CBT sessions with identical psychoeducation and cognitive restructuring but different behavioral modules: Realistic Planning and Timely Beginning (RPT) or Working Time Restriction (WTR). Separate partial least-squares regression models used 483 baseline demographic and self-report items to predict posttreatment Aitken Procrastination Inventory–Revised (API-R) scores under each strategy. Predicted between-strategy differences and Youden-index cutoffs generated recommendations: RPT, WTR, Either, or Neither. Because treatment assignment was randomized, outcomes could be compared between participants whose assigned treatment matched versus mismatched their model-derived recommendation.
Results: In the development trial, matched participants had lower posttreatment API-R scores than mismatched participants (p=5.6×10−24; d=1.90) and higher likely-response rates (p=1.8×10−15; RR=4.6). In the validation trial, each treatment-specific model explained 9% of outcome variance. Matched participants again had lower API-R scores (p=.009; d=.53) and higher response rates (p=.04; RR=1.3). Differences remained at six months (p=.033; d=.40).
Conclusion: Together, these findings indicate that personalization of procrastination treatment can be achieved by modeling differential response to specific behavioral components rather than entire treatment packages, supporting precision intervention in procrastination and beyond. Further, explorative examination of predictors suggests that individuals may benefit from a mechanism-informed, but not strictly mechanism-deterministic, precision framework.
Related
- The Role of Lifestyle Behaviors in the Association Between Behavioral Procrastination and Metabolic Syndrome: A Cross-Sectional Study of 92,184 Workers
- Optimizing Cochlear Implant Programming in Single‐Sided Deafness and Asymmetric Hearing Loss: A Randomized Comparison of ESRT and Behavioral Methods
- Rehabilitative efficacy of cognitive behavioral therapy-based intervention for chronic fatigue syndrome : A systematic review and meta-analysis
- Intensive cognitive behavioral therapy for obsessive-compulsive disorder: a meta-analysis of randomized controlled trials
- Client Satisfaction with Trauma-Focused Cognitive Behavioral Therapy: Findings from a Series of Statewide Learning Collaboratives
BibTeX
@misc{loparo2026optimizing,
title = {{Optimizing Behavioral Strategy Selection during Cognitive Behavior Therapy for Procrastination using Machine Learning}},
author = {Devon LoParo and Magdalena Pietruch and Weronika Maria Browarczyk and Joachim Kowalski and Jarosław M. Michałowski and Marek Wypych},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/7t6j9_v2},
url = {https://osf.io/preprints/psyarxiv/7t6j9_v2/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Optimizing Behavioral Strategy Selection during Cognitive Behavior Therapy for Procrastination using Machine Learning
AU - Devon LoParo
AU - Magdalena Pietruch
AU - Weronika Maria Browarczyk
AU - Joachim Kowalski
AU - Jarosław M. Michałowski
AU - Marek Wypych
PY - 2026
DA - 2026/09/01
PB - PsyArXiv
DO - 10.31234/osf.io/7t6j9_v2
UR - https://osf.io/preprints/psyarxiv/7t6j9_v2/
AB - Objective: To determine whether machine-learning models using baseline self-report data could personalize selection between two behavioral strategies embedded within cognitive-behavioral therapy (CBT) for procrastination. Methods: Participants were 256 Polish-speaking university students aged 18–26 years (M=22.4; 73.0% women) who completed active treatment in two preregistered randomized trials (development sample, N=173; external validation sample, N=83). Participants received five weekly group CBT sessions with identical psychoeducation and cognitive restructuring but different behavioral modules: Realistic Planning and Timely Beginning (RPT) or Working Time Restriction (WTR). Separate partial least-squares regression models used 483 baseline demographic and self-report items to predict posttreatment Aitken Procrastination Inventory–Revised (API-R) scores under each strategy. Predicted between-strategy differences and Youden-index cutoffs generated recommendations: RPT, WTR, Either, or Neither. Because treatment assignment was randomized, outcomes could be compared between participants whose assigned treatment matched versus mismatched their model-derived recommendation. Results: In the development trial, matched participants had lower posttreatment API-R scores than mismatched participants (p=5.6×10−24; d=1.90) and higher likely-response rates (p=1.8×10−15; RR=4.6). In the validation trial, each treatment-specific model explained 9% of outcome variance. Matched participants again had lower API-R scores (p=.009; d=.53) and higher response rates (p=.04; RR=1.3). Differences remained at six months (p=.033; d=.40). Conclusion: Together, these findings indicate that personalization of procrastination treatment can be achieved by modeling differential response to specific behavioral components rather than entire treatment packages, supporting precision intervention in procrastination and beyond. Further, explorative examination of predictors suggests that individuals may benefit from a mechanism-informed, but not strictly mechanism-deterministic, precision framework.
ER -