A new study demonstrates that the so-called "driving while black" phenomenon is not supported by the data in Massachusetts. The study uses a "veil of darkness" approach to estimating racial bias. If the rate of pulling over blacks rises during daylight hours, when police can see the race of the driver, and falls during night-time hours, when they cannot, then that would suggest there is a racial bias in their decision of whether to pull over a car. By looking specifically at the “Inter Twilight Period”—the period that is dark for part of the year and light for the other—it was even possible to compare results at the same time of day with substantially similar driving conditions but in different lighting.
I sometimes worry that the spin of (selective) correlation=causation promoted by progressives causes African Americans to expect and to perceive racism excessively, essentially creating a self-fulfilling prophecy. Hard to know with certainty.
As you so often point, out applying the most logical base rate is vital to attempts to use statistics to evaluate claims of racism. To that end, has the 'veil of darkness' effect been validated in some other way? Merely because it makes sense doesn't make it true.
I don't know that there is an obvious way to validate the findings, short of getting an even better estimate of bias by sending thousands of test drivers of different races, but perfectly identical driving behaviors, to drive the same routes at the same time of day, and collecting statistics on police stops. Interestingly, while that is impossible to do with human drivers, because no two human drivers are identical, it would be at least theoretically possible to put human dummy drivers in unmarked autonomously driven cars, and see if one or the other racial group is stopped more while the car is fully in control. This could give you a nearly perfect estimate of whether there is bias.
Indeed, that data may already exist in the data sets of autonomous car companies, who are already paying backup drivers to sit in cars even though those drivers are not intervening in the autonomous driving 95% of the time. If there is enough racial diversity among their backup drivers, it may be possible to extract an estimate of the bias based on how often police stop autonomous cars. (Of course, there are many factors that might make that difficult. For example, if police only very rarely stop autonomous cars there might not be enough data to make a meaningful estimate.)
Teslas would probably be the best bet here given that there is no external indication that they are autonomously driving, and the sheer number of autonomous miles that Tesla has collected data on. Anyone know Elon Musk and want to float the idea?
Thank you for a thoughtful reply. I didn't make myself clear. Has it been validated that the race of a driver is identified more readily during the daytime? If a cop is close enough to observe a violation the race might be obvious even at night, particularly in urban and suburban areas. I am very appreciative of your entire approach.
Thanks, Zac. I wish other journalists would address the question to the paper's managers/owners.... why ignore the facts? I think they owe their subscribers an answer.
I sometimes worry that the spin of (selective) correlation=causation promoted by progressives causes African Americans to expect and to perceive racism excessively, essentially creating a self-fulfilling prophecy. Hard to know with certainty.
As you so often point, out applying the most logical base rate is vital to attempts to use statistics to evaluate claims of racism. To that end, has the 'veil of darkness' effect been validated in some other way? Merely because it makes sense doesn't make it true.
I don't know that there is an obvious way to validate the findings, short of getting an even better estimate of bias by sending thousands of test drivers of different races, but perfectly identical driving behaviors, to drive the same routes at the same time of day, and collecting statistics on police stops. Interestingly, while that is impossible to do with human drivers, because no two human drivers are identical, it would be at least theoretically possible to put human dummy drivers in unmarked autonomously driven cars, and see if one or the other racial group is stopped more while the car is fully in control. This could give you a nearly perfect estimate of whether there is bias.
Indeed, that data may already exist in the data sets of autonomous car companies, who are already paying backup drivers to sit in cars even though those drivers are not intervening in the autonomous driving 95% of the time. If there is enough racial diversity among their backup drivers, it may be possible to extract an estimate of the bias based on how often police stop autonomous cars. (Of course, there are many factors that might make that difficult. For example, if police only very rarely stop autonomous cars there might not be enough data to make a meaningful estimate.)
Teslas would probably be the best bet here given that there is no external indication that they are autonomously driving, and the sheer number of autonomous miles that Tesla has collected data on. Anyone know Elon Musk and want to float the idea?
Thank you for a thoughtful reply. I didn't make myself clear. Has it been validated that the race of a driver is identified more readily during the daytime? If a cop is close enough to observe a violation the race might be obvious even at night, particularly in urban and suburban areas. I am very appreciative of your entire approach.
Thanks, Zac. I wish other journalists would address the question to the paper's managers/owners.... why ignore the facts? I think they owe their subscribers an answer.