Behavioral profiles are bullsh*t

"The killer was a real grouch"

IP address: 127.0.0.1

City/State: [redacted]

Geographical profile follows:

You do not consider yourself an angry person, though you can be quick to anger. You are willing to forgive people who apologize. You focus on the “big picture,” and at times you can be a perfectionist. You are imaginative and thoughtful. You think of yourself as a moral person, and you have your own sense of ethics. You are independent, but you are caring with the people in your life. Learning about the things you love is important to you. You are determined, and you succeed whenever you try your hardest.

A weakness for you is that you don’t like to listen to thoughtless people, and you can be dismissive of their ideas.

---

A professor enters a crowded classroom with a stack of papers. “These,” he explains to the students, “are personality profiles for each of you. My lab here conducts cutting-edge personality research.” He walks down the aisles, placing each paper face-down on the desks.

“Each of you has been subject to profiling since the start of the semester. I’m not at liberty to disclose the data points we have collected, but we only use publicly-available data. Please turn over the profile and read it to yourself.”

The students do as they’re told. Some are excited; some are nervous; some barely care. But as they read, they’re amazed: it describes them each exactly. Murmurs ripple through the classroom.

“Now then,” the professor continues, “once you’ve finished, pass the paper to your lab partner.”

The students are confused. My private profile? But they do as they’re told.

The murmurs grow louder. The profiles are all identical. Even yours, above.

---

What is the purpose of science? For the sake of brevity, let’s agree on: description, prediction, and control. Description is accurately reporting a phenomenon. Prediction is utilizing knowledge of said phenomenon to accurately and reliably specify the conditions under which it occurs. Control is utilizing knowledge of said phenomenon to accurately and reliably cause the phenomenon to occur, or not occur. To a certain extent, these proceed in order from least to most – utility, complexity, etc.

Description is not bad, though. In fact, it’s good. Description is necessary and important. But if you could describe your bladder without predicting or controlling its behavior, you might feel underwhelmed.

---

Behavioral profiling of criminal suspects sounds reasonable. After all, people who commit crimes have features which can be described – and in many cases, that information is helpful. For example, we know for a fact that African American men are disproportionately arrested, tried, and convicted in America; without descriptive crime data, this factual basis for systemic racism would be unknown.

Perpetrators of “active shooter” gun violence in America are, according to FBI data, approximately 96% male, and mostly 20s-40s, white, who legally own a handgun. Generally, they have not been convicted of a crime prior to shooting someone. While there have been concerns about domestic violence and stalkers – the type of person who uses pressure and violence before the shooting event – between 2000-2013, 10% of cases were categorized as clearly related to domestic violence, meaning that 90% were not. (Stalking and domestic violence are very serious crimes, but may not be a prelude to murder – though a pattern of this behavior is a risk factor). They find that 62% of shooters acted in an “aggressive” or “bullying” way prior to the shooting, but in such a way that legal consequences were avoided.

A concept that comes up frequently in this literature is “leakage.” What they mean by leakage is that covert verbal behavior becomes overt. To wit: people talk about their plans. Leakage occurs (in part) because the shooters often plan for more than 1 week. The term “leakage” is weirdly Freudian, but let’s agree that “most shooters discuss their plans” is a key red flag, however we phrase it.

Other details are downright inane, such as the fact that most shooters have experienced stressors and often had an adverse interpersonal interaction. Active shooters are, on the whole, not happy.

But this is a description. Could it be a prediction?

---

The FBI cautions that you can’t really predict a shooting from demographic information alone. That’s probably because the demographics are not very helpful: in America, white male legal gun owners from 18-50 might constitute 24 million people. (252 million white Americans, approx. half are men, approx. half are in the 18-50 age bracket, and approx. 41% own guns)

The profile for an American active shooter, demographically, is nearly the size of the entire population of Australia.

Fortunately we have a few other signs: a man who has access to guns, plans the shooting, and reveals their plans to commit a shooting. That last one might stand out, anyway.

---

Suppose you catch the warning signs – what is the effective intervention for someone who has the potential to be an active shooter? Well, you might want to run, hide, and fight if they start shooting. But before the shooting, you should make emergency plans. Have the emergency exits memorized.

Unfortunately, that is the campaign to which the FBI has dedicated its marketing budget; you may have even heard “run, hide, fight” before. Meanwhile, they have a truly good, comprehensive report on prevention.

The report is worth reading for various excellent points, but what about prevention? The authors describe prevention as akin to good parenting: flexible, supportive, attentive. This is not necessarily possible to do, at least in all cases, as a last resort – relationships and trust must be built. This is an ongoing process, involving help when it is needed, before the boiling point. But this is America, so: run, hide, fight it is.

Are we asking for too much here? After all, how could anyone suggest a unit that intervenes in crimes before they occur? If a crime occurs, but the perpetrator is not captured, perhaps that is the most logical point of intervention. For a mass shooting event, this would be standard detective work, while for a serial killer, this would be “profiling.”

That’s not a terrible answer; stopping a killer before they kill again is inarguably good. But a successful profile, of some type, could lead to an intervention before a murder has occurred. Clearly demographics are not sufficient, but a series of behavioral checklists could capture a risk profile that leads to supports.

Intervention into problematic drug use is one instructive example. Yes, we have programs to help people who are addicted to drugs. But most wealthy countries experiment with methods of preventing drug use. It isn’t possible to prevent all drug use, but harm can be minimized.

---

Well anyway, here in America we’ve only had about 60 years of active shooters, and perhaps the jury is out, so to speak. What about serial killers? The use case is perfect: someone committing murders, who will continue to commit murders, and who must be caught based on patterns of behavior. Surely a behavioral profile can stop a serial killer?

Drat!

While the Serial Murder Symposium sounds pretty neat, it’s unfortunate that there is no generic profile for a serial killer. But that makes sense; there isn’t a generic profile for, just for example, pickleball players, despite their inherent criminality.

Wanting doesn’t really capture the nature of reinforcement, but…

A somewhat interesting aspect of serial killer profiling is attribution of the action to a causal factor that we might call a maintaining consequence. For example, the killer may kill for money, or to escape consequences of prior actions. Nevertheless, the behavior may be maintained by multiple motivations, and the experts say…you can’t really use these motivations as an investigative tool. They believe that too much investment in discovering a motive could take resources away from capturing the criminal. Maybe – that seems to be, frankly, a guess.

The Serial Murder Symposium has some great ideas for reacting to serial killers, for example a clear plan to share information between police departments. Still, these are generic best practice reactions to a crime that has already occurred.

---

The fundamental problem is treating behavioral profiling like it’s a personality quiz on Facebook. Was the killer narcissistic, or shy, or a freakin’ dingus? We might talk about people this way informally, but those traits are not really actionable.

Likewise, smart reporters understand the problem with, for example, looking at a killer’s reading list. The books might provide some insight into maintaining variables, but only ex post facto – only as description.

The public at large is, bizarrely enough, interested in maintaining variables. The motive for a crime is inherently interesting. But ironically the interest wanes with an understanding of maintaining variables – the mystery fades. For example, some of the most complex murder plots revolve around…money. A tangled interior life, a twisted personality, a mysterious series of childhood traumas; these are the intrigues you must abandon in favor of a profit motive.

The functional analysis literature is incredible and interesting, shedding light on the basic functionality of behavior. Yet it removes much of the mystery – behavior is more comprehensible, less fascinating. Some protests against behaviorism are inherently invested in the mystery. People are too complex, behavior is too complex, to understand.

Perhaps understanding behavior is the easy part.