Measuring Racial Profiling with Self-Driving Vehicles
by Bryan Engelhardt

Using self-driving vehicles, we propose being the first to scientifically test whether racial profiling exists in traffic stops, and if it exists, to measure its prevalence.

Self-driving vehicles provide a controlled environment to scientifically test for racial profiling. Specifically, self-driving vehicles allow a researcher to control for all the variables related to a traffic stop (time of day, location, speed of car, physical condition of car, etc.) while varying the experimental variable - the race or other visual cue of the person sitting in the driver's seat.

Irrespective of the findings, the study will provide valuable information. It is valuable to persons widely suspected of being racially profiled, police departments and their constituencies, scientists working in the field, and manufactures of self-driving vehicles. In the search for truth, the project's findings will provide scientific proof for or against the suspected injustice.

For the developers of self-driving vehicles, the results would provide another strong argument for the use of self-driving vehicles. The results would likely provide significant press to this effect. Furthermore, the study may increase demand from the suspected victims of racial profiling as the vehicles provide scientific evidence of the practice and potentially proof for recompense.

Police Departments and their municipality's tax payers and representatives spend significant time concerned about, and resources measuring and evaluating, racial bias. Our findings would provide scientific support of whether the concern is well founded and the effectiveness of their measures to limit it.

Finally, for scientists, the study provides undeniable scientific proof. It is a clean break from the current approaches that are based on imperfect data and flawed statistical methods. The current tests of racial profiling are routinely called into question. Refer to Alpert, Smith, and Dunham(2004), "Toward a better benchmark: Assessing the utility of not-at-fault traffic crash data in racial profiling research." for one of many discussions.

We are requesting four self-driving vehicles with the auto-pilot feature travel a total of up to 120,000 miles. Under reasonable assumptions, our request will allow us to test for racial profiling beyond a 1 in 10,000 chance assuming a 0.5% (0.005) difference in the pull-over rate per mile of those suspected of being profiled and those who are not.

Click here for a detailed report.

Stage of idea: Requisition

Unmet requisite(s): Self-driving vehicles

Created on Sep 27, 2017 and last updated on Oct 13, 2017.

Related Ideas: Worker Productivity and Driveless Commutes

Suggested citation: Engelhardt, Bryan, 2017. "Measuring Racial Profiling with Self-Driving Vehicles.",

Post a comment | Post a private comment to author(s)