Computer models of officer speech reveal racial disparities in respect

Evidence

Body-worn cameras can be used as a window into the substance of the actual interactions that transpire between police officers and community members. As discussed earlier, our research team developed a method to analyze the footage from these cameras to quantify police-community relations (Voigt et al., 2017). We transcribed the interactions shown in the footage and built computer models that reliably predict respect based on linguistic features—such as apologies, expressions of gratitude, and the use of formal versus. informal titles—that mark politeness, power, and social distance. In so doing, we were able to systematically examine the respectfulness of the actual words officers speak to the drivers they pull over. We found racial disparities in the level of respect: officers spoke to Black drivers less respectfully than they spoke to White drivers. — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 432