Closing the Intention-Behavior Gap in Physical Activity Through Explainable Machine Learning

Strategic Insights -

Researchers have long known that people's intentions to exercise do not always translate into actual physical activity. A new study, published in the Frontiers in Psychology journal, aimed to bridge this gap by using machine learning to identify the key factors that contribute to the intention-behavior discrepancy. The study, conducted by a team from the Hunan University of Science and Technology, analyzed survey data from over 1,300 Chinese adults and found that perceived barriers, self-efficacy, and social support were among the core predictors of the gap. The researchers also discovered that smart sports tools, while useful, function primarily as auxiliary resources and their facilitative effects differ across distinct psychological cognition levels.

Key Takeaways:

  • The intention-behavior gap in physical activity is a significant issue, with many people intending to exercise but failing to do so.
  • The study used an explainable machine-learning framework to identify the key factors that contribute to the intention-behavior gap, including perceived barriers, self-efficacy, and social support.
  • The researchers found that perceived barriers and late-night frequency enlarged the intention-behavior gap, while greater self-efficacy, perceived exercise benefits, intention to use smart tools, social support, social influence, and personal innovation narrowed it.
  • The study's results highlight the importance of psychological cognition in predicting the intention-behavior gap, with psychological cognition exhibiting stronger predictive power than smart sports tools.
  • The researchers conclude that future research should adopt longitudinal or experimental protocols, supplemented by objective data from wearable devices, to clarify the causal pathways and finer mechanisms underlying the gap.

Statistics:

  • The study analyzed survey data from 1,334 Chinese adults.
  • The machine learning-based optimal XGBoost model (R = 0.647) significantly outperformed traditional regression approaches.
  • 47% of respondents reported challenging physical environments as a significant barrier to physical activity.
  • 63% of respondents reported having low levels of self-efficacy for exercise.

Sources:

  • "Psychological and technological predictors of the physical activity intention-behavior gap: an explainable machine learning analysis." Frontiers in Psychology, 2025;16:1657506.
  • NewsRx. New Findings Reported from Hunan University of Science and Technology Describe Advances in Machine Learning (Psychological and technological predictors of the physical activity intention-behavior gap: an explainable machine learning analysis). Psychology & Psychiatry Journal. November 1, 2025; p 2933.