 In February of 1999, a 23-year-old immigrant from Africa, a man by the name of Amadou Diallo, was shot to death by four plainclothes police officers. Amadou was returning home from a late meal at around 1 a.m., when the four officers in street clothes exited their car. Their account of events says that they yelled that they were NYPD officers, and they ordered Amadou to stop and show his hands. Instead he ran for his porch, where, thanks to a burned-out bulb and the porch light, he was only backlit by the light from the interior. Realizing he couldn't escape these four men, he turned and reached into his jacket and retrieved his square wallet. One of the officers shouted, Gunn! And all four officers opened fire on Amadou. In the initial volley, one of the officers tripped over a step, leading the others to believe he had been shot. In all they fired 41 rounds. Amadou was hit 19 times and pronounced dead on the scene. The NYPD Investigations Team ruled it a justified shooting, in spite of no weapons being found on the scene, because the officers acted consistently for the situation they were presented with. The officers were indicted for second-degree murder by a Bronx grand jury, but after a change of venue to Albany, which has less than half the black population of the Bronx, they were all acquitted in jury trial. For one officer, Kenneth Boss, this was his second fatal shooting of an unarmed black man in a two-year period. He still worked for the NYPD and in fact was promoted to the rank of Sergeant in 2015. How often does this happen? The correct answer is short. We really don't know. The statistical tools used for crime reporting and officer homicide aren't consistent or national. Our best attempts have centered around tallying events from newspaper reporting. The 1999 incident is upsetting, but it presents a puzzle. Why do police officers keep shooting unarmed black men? And why do they frequently avoid any legal consequence? There are many points of view on this question. You could argue that it can all be explained by centuries of internalized racism, with cops acting out racial violence in a new legal form. You could argue that the whole thing can be explained by the prevalence of crime in the black community and the siege mentality of gangs and the police who combat them. It's easy to choose the explanation that best conforms to our political views, our worldview. I want to see if we can set all that aside, not as unimportant, but as unproductive lines of inquiry at the moment. I want to propose a theory that's rooted in mental information processing. There's some groundwork I have to go over before we get there. I want to introduce three concepts, the diagnostic matrix, automatic versus controlled responses, and the role of anxiety in decision making. First, the diagnostic matrix. For every type of physical test for a condition, there are four possibilities. A true negative, true positive, false negative, and false positive. A classic example might be a home pregnancy test. If you actually are pregnant, the test will either read as a true positive or a false negative. The percent of time that the test detects correctly when you are pregnant is the sensitivity of the test. If you aren't pregnant, the test can return a true negative result, otherwise it's a false positive. The percent of time it returns a negative result for negative conditions is the specificity of the test. Actual pregnancy tests will have both sensitivities and specificities of 99% or greater. We then have to factor in the prevalence of the condition to determine how useful the test is. For very, very rare conditions, returning a false positive only 0.001% of the time could still mean that most positive results are actually false. For example, a mammogram screening in women under 40, who only develop breast cancer very rarely, most positive findings turn out not to indicate actual cancer. The stress and anxiety and risky biopsy procedures this false positive creates is why most physicians don't recommend mammography screening for young women at very low risk. In the case of a police shooting, let's imagine the officer as a diagnostic test for people attempting lethal harm. If they shoot and the person posed an imminent risk, that's our true positive. A false positive result is an officer shooting an unarmed person, as in the case of Amidu. At this point it doesn't matter why the error was made, it's merely a way of describing the condition and the response. The false negative is significant, because it represents the risk of bodily harm to the police officer, and a salient is armed and intends to kill you, your hesitation allows them to get a shot off. So there's sensitivity to threat and specificity of response, and the two almost always require tradeoffs. If the officer hesitates to increase specificity of response, the sensitivity necessarily suffers. If the officer is maximizing sensitivity to threats, specificity will necessarily suffer. This is the basic game theory of police response as viewed by perfect actors, a robot or a computer, but we're going to have to modify it for real humans. The next factor is an awareness of how our brains process information. We aren't simple computers, each new stimulus is interpreted depending on the other stimuli leading up to it. For example, if we see dark clouds gathering, we're more likely to correctly detect and recognize the distant rumble of thunder. We're primed by one stimulus to sense or interpret the next stimulus. Advertising is very good at priming us subliminally, combining images and close succession like farms and fast food, social gatherings and beer, so that our response is primed to interpret the fast food is healthy. Beer drinking is primarily a happy social event. We call this type of association the automatic process. We're often aware of it, but it operates primarily at the level of automatic effortless response to stimuli. Intention with the automatic response is the controlled response. This is the slower process, requires effort and is under conscious control. It's not always rational or measured, but it is the process that we feel more in control of. In our example, while our automatic response to the farm and fast food might be a tendency to view the product as more healthy, our controlled response might be more aware of the fact that fast food is mostly premade in giant factories and stored in ship frozen. There are some experimental ways to assess automatic processes by exploiting the fact that they're faster than controlled processes. By comparing results between tests where the subject has only fractions of a second to respond to time unlimited responses, the scientists can differentiate between them. In a key study on this subject, a researcher at Washington University took a small group of 32 non-black undergraduate students. And asked them to identify whether an object they saw on a computer screen was a gun or not a gun. The computer flashed up the face of a black or a white man for only 200 milliseconds, a fifth of a second, followed immediately by a second image of either a gun or a hand tool for the same time, 200 milliseconds. The screen then went blank until the student responded gun or no gun with a button push. The delay in response was carefully measured and accuracy scored. The result was consistent with expectation. When a black face was used as the prime, subjects identified the gun more quickly. There was no correlation to survey responses on attitudes about race. The time scale over which it occurred means that the effect was primarily due to the automatic response. The experiment was repeated, but this time with a time limit to respond of only 500 milliseconds or half a second. The error rates went up for all categories, but the error that increased the most was misidentifying a hand tool as a gun after priming with the face of a black man. So now we have the beginning of a cause and an answer to how race plays into the situation of police officers shooting unarmed black men. In our diagnostic matrix, the automatic response, which dominates our rapid responses, is error prone and more susceptible to priming. It taps into our subconscious and isn't a part of our response that requires effort on our part. I hate to say we're not in conscious control of it, but it's more determined by conditioning. It's entirely possible for a police officer to have no conscious racist intent to be acting in good faith in their decision making. And yet the finger that pulls the trigger isn't responding to that conscious sensibility. Into this mix, we also bring the realities of being in law enforcement. Even simple traffic stops can be risky to an officer's life and their response time and good judgment might be the only thing that saves them from serious injury or death. Police, like soldiers, deal with a level of anxiety on a daily basis that is hard for most people to understand. There's a robust body of evidence on how anxiety impacts judgment. What we know is that people who are anxious, as measured by heart rate and blood pressure, tend to over-detect threats. They produce more false positives for weapons. A good example is a 2012 study where 36 police officers went through a shoot-don't-shoot computer video simulation where a subject popped up and either fired their digital weapon or had no gun and threw their hands up to surrender. The experimenters added anxiety by including a shoot-back cannon that could fire plastic bullets at the participants for wrong answers, just enough to sting and be unpleasant. When the high anxiety scores were compared to the low anxiety, no shoot-back cannon scores, the officer's shot accuracy was lower and they fired at unarmed suspects almost twice as often. Their response time was faster with the anxiety, but more error-prone, specifically in the direction of false positives. There were no racial cues or priming given this experiment, so we're just looking at how anxiety to harm makes police and indeed probably anyone more likely to detect a gun where none exists. In light of the data on this subject, I wanna propose an answer to the question in the title. Why do cops shoot unarmed black men? I think there's room for everyone to be a little right and a little bit wrong. Cops can be explicitly racist and that can't be ignored, but I suspect that the more common case is a level of implicit racism that is ingrained beyond conscious control. Older officers lived in an era of segregation even if they were enlightened or have had personal experience of threat that became part of their ingrained response. The very nature of the job that law enforcement officers do, relying on snap judgments, constant anxiety, and error-prone threat assessment will serve to magnify any innate racism or stereotyping. A good cop, one with no conscious sense of racism, can still be threat-oriented when confronted with a simple traffic stop as in the tragic case of Philando Castile shot during a traffic stop when he calmly reached for his ID. I wouldn't prevent anyone from protesting about what they view as injustice. Protests are a good way to raise awareness and move toward change. What I would propose is that more funding is needed to study how to reduce the likelihood of these events. Better training can actually help. Officers who train in high anxiety scenarios like being shot with soap pellets for incorrect decisions, learn to better manage their automatic responses. Hiring practices for new cops can be informed by research on how personality traits and anxiety management improve appropriate responses. The point is that we can do something about this and it will be better for police relations, it will be better for individual officers and it will go a long way towards rebuilding the lost trust between the police and black communities. It will take time and funding and research and most importantly, a serious will to get it done. Thanks for watching.