Criminal Investigative Failures: Avoiding the Pitfalls (Part Two)

Article excerpt

Part one of this article focused on cognitive biases and how they can contribute to criminal investigative failures. Part two presents probability errors and organizational traps that can lead investigations astray. It also offers recommendations and additional strategies that investigators may find helpful.

[ILLUSTRATION OMITTED]

PROBABILITY ERRORS

Probability and Psychology

Anyone who has spent a few hours watching people gamble will realize that probability is a difficult concept for the human mind. Individuals often use heuristics--and suffer from biases--when dealing with probability. Police officers find it particularly hard to think probabilistically. Because of their street experiences, they prefer black and white, rather than shades of gray. Probability errors in criminal justice most often occur in the forensic sciences but also can happen in criminal profiling.

Coincidences and the Law of Small Numbers

A common problem with probability results from looking for patterns in, or drawing inferences from, a small number of incidents. For example, an analyst examines the dates for a series of 15 street robberies and observes that none of the crimes occurred on a Thursday. Is this pattern meaningful? Probably not. With only 15 crimes, chances are at least one day of the week will be free of robberies.

Skeptics often say they do not believe in coincidences. However, when looking for patterns within large numbers of items (i.e., events, suspects), coincidences are inevitable. The comparison of Presidents Kennedy and Lincoln provides a well-known example. The list of remarkable similarities is strictly the product of chance (with 43 U.S. presidents, 903 possible comparisons are possible) and cherry picking (noting similarities, while ignoring differences).

What role does coincidence play in major crime investigations? If enough suspects are looked at, by sheer chance, some will circumstantially appear guilty. A few people will just be in the wrong place at the wrong time. Efforts to solve a crime by "working backwards" (i.e., from the suspect to the crime, rather than from the crime to the suspect) are susceptible to errors of coincidence. If you look hard enough, you can usually find some sort of connection. These types of errors often are seen in the proffered "solutions" to such famous cases as Jack the Ripper.

Coincidences can be a trap when offender modus operandi and similar fact evidence are used for crime linkage purposes. Trawl search problems occur when only similarities, and not differences, are examined. (1) Comparisons of common similarities (e.g., vaginal intercourse in rape crimes) lack utility, while misspecifications of similarities can be misleading. Consider two juvenile murder strangulations involving body transportation and concealment.

While the similar crime characteristics suggest a link, more detailed examination reveals important inconsistencies. One victim was a 3-year-old male, manually strangled, his body found in a dumpster 100 yards from his house. The other victim was a 14-year-old female, strangled with a rope, her body found dumped in a river 20 miles from her home.

Double Counting

Extracting two elements of a crime from a common source and then erroneously treating them as separate aspects can mislead a criminal investigation. A rumor heard from more than one person does not necessarily verify the information as both individuals may have received it from the same source. Consider a behavioral profile of a child murderer. Amongst other details, the profile estimates the offender's age and his vehicle type, derived from automobile insurance data. Using the profile, investigators evaluate two suspects--one matches both the age and vehicle criteria, and the other only the age. Who is the better suspect vis-a-vis the profile? Actually, they are equal. Derived from the age estimate, the vehicle type is not an independent profile element drawn from the crime scene (as opposed to a vehicle sighting by a witness). …