Gradual Privacy Paradox in AI-Enabled Fitness: An AI Ethics Interpretation of Privacy Satisficing Under Bounded Rationality

AI-enabled fitness services rely on continuous collection of activity, physiological, and location data to support monitoring and personalized feedback, which raises persistent privacy and security concerns and ethical tensions regarding data use and user autonomy. Nevertheless, sustained engagement with these services remains common, indicating a divergence between privacy concern and continued use. Using online survey data from 596 adults aged 18 years and above, this study examines AI fitness use from an AI ethics perspective grounded in bounded rationality. A Deviation index is constructed as the standardized difference between privacy concern and risk acceptance. High willingness to use AI fitness services is analyzed using a parsimonious probability-based approach. Logistic regression models examine how the likelihood of high use varies across the Deviation range, while accounting for perceived transparency and safety, measured as Information Control Level, and stated privacy trade-off attitudes. The results show that continued use varies systematically across the Deviation spectrum. Higher Deviation values are not associated with a collapse in use probability. Instead, predicted probabilities change gradually across the observed range. Privacy concern and continued AI fitness use therefore coexist within this adult user sample. This pattern supports a descriptive AI ethics interpretation of privacy satisficing under bounded rationality rather than a binary privacy paradox.