November 11, 2011

Predicting behavior: The case of the missing militant

There is an oft-repeated axiom in our field that the most reliable predictor of future behavior is what a person has done in the past.

But is this axiom valid?

Let's take the example of Ronald Stanley Bridgeforth, featured on America's Most Wanted.

Forty-three years ago, while being detained on suspicion of trying to buy merchandise with a stolen credit card, the 24-year-old Black militant pulled out a revolver and shot at police. He jumped bail and, three years later, became a suspect in the killing of a police sergeant during an armed invasion of a police station in San Francisco.

Would this information lead you to predict that he was likely to engage in more violence in the future?

If so, you would be wrong.

Bridgeforth vanished from the radar screen, and eventually police figured he might have died. But the former community activist was far from dead. After fleeing to Africa, he moved to Ann Arbor, Michigan, married, and raised two sons. Under the assumed name of Cole Jordan, he worked as a janitor, earned a bachelor’s degree from Wayne State University and a master's degree in counseling from Eastern Michigan University, became a licensed professional counselor, and eventually worked his way up to the rank of a full-time faculty member at Washtenaw Community College.

Bridgeforth with attorney Paul Harris (L) and wife Diane (R)
Last week, Bridgeforth finally turned himself in. Authorities were not closing in, but he had a troubled conscience. He plans to plead guilty in the assault case, in which he faces a maximum of five years in prison. Prosecutors announced they will not prosecute him in the infamous murder of Sgt. John Young at San Francisco's Ingleside Station on Aug. 29, 1971. That case unraveled two years ago in part due to allegations (aired in a documentary, Legacy of Torture) that police used torture with electric shock, cattle prods, beatings, sensory deprivation and asphyxiation to obtain confessions from three of the nine suspects.

One might argue that Bridgeforth is an exception to the rule. Only, he's not. Time and again, we hear about a long-time fugitive who lived a quiet life, surrounded by friends and co-workers who had no clue about his or her violent past.

David Gonzales, William Walter Asher III, Katherine Ann Power, Claude Daniel Marks and Donna Jean Willmott, to name just a few.

These cases are testament to the weak validity of the axiom that past behavior is a good predictor of the future. There are several flaws with the theory, among them:
  • The base rate: Most serious crimes have a low base rate of recidivism. That makes us most likely to be correct if we predict that the behavior will NOT reoccur. For example, because of the base rate of rearrest for murder, we would be wrong in the broad majority of cases if we predicted that someone who has killed once will most likely kill again. The same is true for sexual offending. In one recent study, 95 out of 100 people arrested on sex charges had no prior sex crimes; an inordinate focus on the five percent lends an illusion of a higher base rate of reoffending than the evidence warrants.
  • Desistance: A second major problem with static predictions is that people change. In fact, even hard-core criminals almost universally desist from crime as they age. This holds true across all eras and cultures. As scholars Shadd Maruna and Laub and Sampson have shown, crime is mainly a young man's game. As they age, offenders settle down and become less impulsive. Or, they simply burn out.
  • Environmental context: The axiom of past as prelude also ignores the circumstances that contribute to offending. Criminologists have long known about the critical importance of context. For example, peer influence is critical to crime by adolescents and young adults, who have the highest rates of offending. Lifestyle circumstances that can -- and often do -- change over time influence other types of crimes as well, such as robberies and drug offenses. For Bridgeforth and others of the 1960s-1970s era, the context was a militant revolutionary movement that swept up many idealistic young people.
  • Unproven allegations: The Bridgeforth case also highlights the problem of relying on allegations of past misconduct that may not be reliable. Bridgeforth has denied the charge that he was the getaway car driver in the San Francisco police killing, and now prosecutors have chosen not to prosecute him.   

Ultimately, the past-as-prelude axiom may hold better for some behaviors than others. Perhaps it is more reliable when predicting scripted or compulsive acts that a person engages in with high frequency over a lengthy period of time. However, it is less reliable when applied to context-influenced behaviors with low base rates of reoccurrence.

And never should we ignore the influence of aging. Bridgeforth is not the same man at 67 as he was at 24. Think back to your own adolescence or early adulthood; are you the same person now as you were then?

The viewpoint that past is prelude is fundamentally pessimistic, leaving little room to acknowledge that human beings are highly adaptive, and often capable of learning from mistakes and changing our lives.


jedson said...

Karen –

With regard to you well reasoned and thoughtful article “Predicting Behavior: the Case of the Missing Militant,” at first glance it would appear that you pretty well demolished the hypothesis that “the most reliable predictor of future behavior is what a person has done in the past.” As this was the position I took in a recent blog post at your site (see comment under Call for papers on violence risk assessment ), my first thought was, “well perhaps I was wrong.” It wouldn't be the first time. The four points that you raise in order to disclose the flaws in my position are impressive. I agree totally with all of them. Let me add a fifth. Very often some consequence of a behavior – legal or otherwise – may deter its repetition. What these points strongly suggest is that human behavior is simply not predictable – at least not without a serious degree of uncertainty. That, I would concede is really the important point, and it is an insight that should guide our public policy. That said, let me rephrase my position just slightly. Try this: In so far as human behavior is predictable at all, the best predictor is the pattern of past behavior. This position is not refuted by your four points.

Let me illustrate with your illustration. But lets step back. Suppose we are talking about making a prediction back just before Bridgeforth jumped bail. He had been engaging in an illegal behavior and he resisted arrest by means of a violent act. I would on that basis predict that he was more likely to engage in violent solutions to problems in the future than the average person. Next he surfaces as a revolutionary. Set aside the difficult issue of whether violence is justified in revolutionary action. He is again behaving in a violent manner. Then lets look at another aspect of your illustration. The police apparently got their confession with the use of torture. I would predict that the police officers who did this were more likely than the average person to use torture in the future, and also more likely to engage in violence to solve other problems than the average person.

In short, I agree with you that human behavior is not ultimately predictable, and this is one very strong argument against incarcerating people for crimes they might commit in the future. Nevertheless, insofar as predicting behavior is possible (we are dealing with probabilities) the best predictor of future behaviors is a past pattern of behavior. On most points, I think, our differences would not lead to different policies. I think, though, that my position does raise questions about the validity of expert testimony to predict what people will do. Its no better than a description of a person's past habits.


Karen Franklin, Ph.D. said...


Thanks for your comment. I'm not sure that expert witnesses have nothing at all to offer, but I think we need to approach prediction very cautiously, not go beyond the scientific method we are using, and acknowledge our limitations. I have seen reports in which experts state things like, "Based on the results of X, Y and Z, Mr. Jones has a 57% likelihood of engaging in Behavior A within the next five years." Clearly, this level of precision is dubious when faced with something as complex as human behavior.

jedson said...

I probably overstated my objection to expert witnesses. If they can present us with clear ideas that may not have occurred to us, or data that we can check out, I am sure they can be useful. What I often see, however, is just pontification about presumed “facts” and truths that we are to accept simply on their authority. I enjoy your posts.