AI-enabled instructor care and physical fitness performance in police cadets: the mediating roles of emotions, engagement, and recovery

BackgroundPolice training is physically demanding and psychologically stressful, which may undermine training adaptation and physical performance. Artificial intelligence (AI)-supported learning analytics systems have been proposed as scalable tools to support training management, yet empirical evidence remains limited regarding how such systems may operate when embedded in human instructor care practices.MethodsWe conducted a quasi-experimental pre-post study among police cadets (N = 224) from four natural classes in a second-level police college. Two classes (n = 112) received an AI-enabled instructor care dashboard intervention, while two classes (n = 112) followed routine training. The dashboard generated weekly care signals based on non-sensitive training and self-report indicators and prompted instructors to provide standardized supportive actions, including brief conversations and individualized training suggestions. Physical fitness performance (0, 100) was assessed at the beginning (T0) and end (T1) of the semester. Difference-in-differences (DID) models with covariate adjustment and class fixed effects were used to estimate intervention effects. Exploratory mediation analyses examined whether post-intervention emotional states, training engagement, sleep quality, and burnout were associated with the observed intervention effect.ResultsDID estimates indicated a significant net improvement in physical fitness performance in the AI-enabled instructor care condition (β = 3.48, p < 0.001). Exploratory mediation analyses suggested that this improvement was statistically associated with more positive emotional states, higher training engagement, better sleep quality, and lower burnout.ConclusionAI-supported instructor care may represent a low-risk and scalable approach to supporting physical training performance in high-stress educational settings. The findings provide quasi-experimental evidence consistent with an intervention effect, while the mediation results should be interpreted as exploratory evidence of associated psychological and recovery-related pathways.