What Officers Actually Do: Large-Scale Behavioral Coding of Officer–Supervisee Interactions in U.S. Community Supervision

Community supervision oversees about 3.7 million adults in the United States, yet the officer–supervisee interactions at its core remain largely unmeasured. This study represents the largest objective behavioral coding effort of naturalistic U.S. community supervision to date, applying three independent observational systems (officer behavior coding, global motivational interviewing (MI) fidelity rating, and observer-rated relationship quality) to 2,192 sessions from 759 supervisees and 46 officers, with adequate to good inter-rater reliability. Multilevel variance decomposition revealed that session-to-session fluctuation within established dyads dominated the variance structure across all three systems, limiting what single-occasion self-report measurement can capture. Most sessions clustered at behavioral neutrality on MI-consistent dimensions, establishing a naturalistic baseline for routine supervision contacts. Cross-system correlations from non-overlapping observer teams linked specific officer behaviors to independently rated relationship quality, revealing an observable behavioral architecture underlying the officer–supervisee relationship. Implications for measurement, officer training, and behavioral fidelity infrastructure in community supervision are discussed.