Why Profiling Serial Killers Can’t Work
You can’t find a needle in a haystack with a pitchfork.
Posted November 30, 2017
Profiling of serial killers probably doesn’t help catch them. Much of this work is based on interviews of serial killers and reviews of their murders to assist in pattern recognition with future cases. The lore includes such observables as the idea that killers return to the scene of the crime, enjoy participating in the investigation, and know or don’t know the victims prior to the killing. These data points could conceivably narrow down the suspects in a given situation. The lore also includes many unobservables, such as whether the killer was organized or disorganized, merely angry enough to kill or even angrier, has high or low self-esteem, and took the situation personally or didn’t. (This is different from knowing the victim. A killer could take an incident of road rage or sexual rejection personally and become extraordinarily angry even if the killer did not know the victim beforehand.) Any ideas about serial killers that are not observable cannot help the police narrow the suspect field.
Instead of helping catch serial killers, profilers turn the police’s attention to suspects that fit the profile. Fortunately, police typically don’t put much stock in profiles, but when they do, their investigations tend to confirm their suspicions. There is not good data available on the helpfulness of profilers, but we know of some cases, such as that of Tim Masters, where a suspect was identified by profiling and then wrongly convicted by a combination of prosecutorial certainty, expert hubris, and the reluctance of juries to let someone go free who has been painted as a menace to society.
But I want to write about why profiling can’t work, not why it doesn’t. The first thing you need to know to understand this is the number of serial killers operating at any one time. The problem here is that reporting agencies lump together those we tend to think of as serial killers with regular killers who simply murdered more than once. When the FBI says there are 25-50 serial killers operating in America at any one time, only some of those would be of the sort that profilers profile. But say for the sake of argument that there are 50 in America at any one time or roughly one for every 6 million people. That would average to about one in a state like Colorado (population 5.5 million).
Now imagine that the Colorado police believe that a death is the result of a serial murder. They call a profiler who looks at the scene, recognizes some patterns and opines that the killer is a White male, age 15 to 65, who did not know the victim, took a trophy, has high self-esteem, and is following news reports on the murder very closely. The question is not how many serial killers fit this profile, and it is not just whether this killer fits this profile; the question is how many other people fit this profile. You can’t tell by looking at someone whether they have low self-esteem or whether they are following news reports. It looks to me like about 2,000,000 people who were in Colorado on the day of the murder fit this profile.
Now imagine that a truly gifted profiler examines the murder scene and intuits that the murderer “has some medical knowledge,” meaning that the killer either works in a medical field or is smart enough to learn about medicine on the internet. That could narrow the suspect pool by half, a remarkable feat of profiling, but one which still leaves us with a million suspects. Now the profiler tells us that the killer hates women, but there’s no way to use this information. But then we get another truly astonishing insight: the killer drives a truck or SUV. This is the kind of guess that, when they arrest the killer and learn he drives an SUV, makes profilers pat themselves on the back and claim validity for their methods. However, before the killer is caught, even an amazing intuition like this one only cuts our suspect pool in half (I’m guessing, but I see a lot of SUVs in Colorado). Now we have one of the most insightful profilers we can imagine, but we still have half a million suspects.
So let’s take it a step further and give our profiler a touch of genius. The profiler says the killer is married, works full-time, owns a home with a garage, shovels his walk in the winter rather than hiring a company, dresses casually except at formal occasions, and loves crossword puzzles. We still have 10,000 suspects, too many to interview much less make an arrest.
That last item about crossword puzzles comes from this old chestnut: All you know about a man is that he loves poetry and crossword puzzles. Is he more likely to be an English professor at Harvard or a truck driver? The answer is a truck driver. There are maybe 15 male English professors at Harvard, and even if ALL of them love poetry and crosswords, you’d only have to find 16 truck drivers who loved both to make the truck driver more likely, and there are over a million truck drivers in the US alone. Heck, if you think having a Ph.D. in English nails down the poetry-and-crossword profile, I’d guess there are more truck drivers with Ph.D.s in English than there are Harvard professors with Ph.D.s in English.
To find a needle in a haystack, you either have to comb through the haystack or use a huge magnet. If the haystack has a million stalks of hay and only one needle, the magnet works because it includes the needle and none of the hay stalks. To find a serial killer among millions of other people, you would need a profile that includes the killer and none or nearly none of the other people. DNA evidence is that kind of magnet, or getting caught in possession of a trophy from the victim. Profiling patterns are not.