Living well? The unintended consequences of highly popular commercial fitness apps through social listening using Machine-Assisted Topic Analysis: Evidence from X

Abstract

Objectives: Use artificial intelligence–Human collaboration to investigate the unintended consequences of the most popular commercial fitness apps through social listening via X (formerly Twitter) posts. Design: Machine-assisted topic analysis (MATA). Methods: X posts (n = 58,881) referring to the five most profitable fitness apps were collected via application programming interface and filtered for negative sentiment, resulting in 13,799. MATA was used to generate a structural topic model. This organized the data into topics and provided 20 representative posts per topic for further qualitative analysis, informed by a thematic analysis approach. Results: Six topics were generated by machine analysis and subsequently retained as independent themes during human analysis. These reflected key challenges and unintended consequences of using commercial fitness apps, including negative psychological and behavioural impacts. These centred around the challenges of quantifying real-world activities, implications for accuracy, difficulties associated with achieving algorithm-set goals, and subsequent negative impacts on emotions, motivations, and engagement with apps and health behaviours more generally. Conclusions: This study highlights the negative behavioural and psychological consequences of commercial fitness apps as reported by users on social media. Our findings suggest that these may undermine the apps' potential to promote health behaviour change and well-being. This highlights the need for a more user-centred app design based on psychological theory, prioritizing well-being and intrinsic motivation over rigid, quantitative goals.