Machine Learning Approach for Predicting Older Adults' Responsiveness to Cognitive Training Interventions: Data from
Petra Vargek1,2, Sašo Karakatič3, Karin Bakračevič1
1Department of Psychology, Faculty of Arts, University of Maribor, 2000 Maribor, Slovenia.
Journal of Intelligence
|April 27, 2026
Summary
Machine learning models predict cognitive training success in older adults. Models showed potential for personalized interventions, suggesting initially advantaged individuals benefit most from cognitive training.
Area of Science:
- Cognitive neuroscience
- Artificial intelligence in healthcare
- Gerontology
Background:
- Personalized cognitive training aims to maximize benefits for individuals.
- Machine learning (ML) offers predictive capabilities for tailoring interventions.
- Elderly individuals are a key demographic for cognitive enhancement research.
Purpose of the Study:
- To develop supervised ML models predicting near and far transfer of cognitive training.
- To identify baseline characteristics associated with training response.
- To predict individual responsiveness to memory, reasoning, and speed-of-processing training.
Main Methods:
- Utilized publicly available data from the ACTIVE study.
- Applied multiple supervised ML classification algorithms.
- Used sociodemographic data, cognitive function, everyday functioning, and depressive symptoms as features.
Main Results:
- Models predicting near and far transfer showed better-than-chance discrimination (AUC 0.56-0.74).
- Far transfer prediction models showed limited performance.
- Predicted responsiveness varied by participant characteristics, indicating potential magnification effects.
Conclusions:
- Developed ML models show practical potential for personalized cognitive training selection.
- Tailoring interventions to individual characteristics could improve program implementation.
- Further external validation of the predictive models is necessary.
