The first major US study of body-cam footage concluded that police, at least in Oakland, California, showed more respect to white people than to black people.
The study from Stanford University researchers analyzed the transcribed text from 981 traffic stops caught on body cams by 245 Oakland Police Department officers in 2014. White people pulled over were more likely to be called "ma'am" or "sir," and they were more likely to hear the words "please" and "thank you" from police officers. Black people, however, didn't get as much respect, and they were more likely to be called by their first names and even "my man."
"Indeed, we find that white community members are 57 percent more likely to hear an officer say one of the most respectful utterances in our dataset, whereas black community members are 61 percent more likely to hear an officer say one of the least respectful utterances in our dataset," according to the study. (PDF) The outcomes held fixed regardless of the race of the officer, the research stated.
The researchers level out that their survey underscores that knowledge collected from physique cams can have extra makes use of than for simply police work. They are saying their analysis mannequin might be duplicated with different police departments.
"This work demonstrates the facility of body-camera footage as an essential supply of knowledge, not simply as proof," the researchers discovered.
The researchers cautioned that the research was carried out with solely the transcripts of 183 hours of body-cam footage, not the footage itself. And privateness considerations however, the researchers stated that extra could possibly be discovered from body-cam research in the event that they listened to the audio and watched the footage. They wrote:
Nevertheless, learning body-camera footage presents quite a few hurdles, together with privateness considerations and the uncooked scale of the info. The computational linguistic fashions introduced right here supply a path towards addressing each these considerations, permitting for the evaluation of transcribed datasets of any measurement, and producing dependable scores of respect mechanically.
A assessment of the particular footage might additionally act as a educating second for police departments. "As well as, footage evaluation might assist us higher perceive what linguistic acts lead interactions to go properly, which may inform police coaching and quantify its impacts over time," they stated.
The research introduced right here open a path towards these future alternatives and characterize an necessary space of analysis for the research of policing: computational, large-scale analyses of language give us a strategy to look at and enhance policeâ€“group interplay that we've got by no means had earlier than.
They stated they're releasing all the code of their "computational linguistic fashions" for others to make use of.
The analysis was accredited by the Stanford College Institutional Evaluation Board with the cooperation of the Oakland Police Division.