Why Social Science Cannot Predict the Extreme

Posted by Matt Grawitch

Do you remember the shootings at Columbine? Did you hear about the recent shooting at the University of Alabama? Both of these events – and many others just like them – are horrific tragedies. Whenever these types of events occur, it seems that psychologists rush to the scene to “explain” why they happened.

For example, after the Columbine shootings, social psychologists rushed out to explain the behavior as a function of low self-esteem. As a consequence, over the past decade we have been inundated with attempts to ensure that we help maintain people’s self-esteem so that they won’t engage in a similar behavior. Yet, a retrospective analysis of events leading up to the incident suggests that high self-esteem may have been more of a culprit than low self-esteem.

Regardless of which side of the self-esteem fence you find yourself on, one fact is very much ignored by those that seek to explain the incident: Previous research does little or no good in helping us to predict these extreme events. Whether it is a school shooting, workplace violence, a terrorist act, or a high school athlete who dies of a rare genetic defect, all of these events fall outside the scope of what contemporary researchers examine.

Before you jump all over me, remember that I, myself, conduct applied and laboratory research all the time. However, I also know the limits of that research.

In research, we use analytic methods that usually require us to exclude participants who are at the extremes. Such cases are called “outliers.” and when they are drastically different from the other people we are studying, they have an inordinate influence on the effects we observe in our research. Hence, in many cases, the practice of removing outliers helps to avoid cases where a very small number of people can bias our results in one direction or another.

However, the practice of eliminating outliers also has unintended consequences for generalizing our results. Most (though not all) research emphasizes people who do not fall outside the scope of what we would deem to be “normal” or “average.” So, to attempt to explain behavior from the “outliers” responsible for the Columbine shootings based on the enormous amount of self-esteem research conducted on “average” individuals is actually to draw inappropriate conclusions.

You see the same problems when trying to explain terrorism and workplace violence. You cannot study moderate members of a group and then attempt to explain the behavior of extreme members of that group. It just doesn’t work (though many psychologists would lose their camera time if people acknowledged this).

Does this mean that research is useless in predicting behavior? Well, the answer to that is not so clear cut. Research is useful only in predicting the behavior of people who have been studied. Because extreme behaviors do not become a popular focus of study until after the extreme behavior has occurred, it is difficult to study these behaviors before they occur. Furthermore, because actual incidents of workplace violence are few and far between, it would be difficult to study these incidents systematically. So, all we can do is try to see what led up to some extreme incident and do our best to re-construct it.

But, looking at things after the fact (post hoc) brings about its own problems. This archival approach typically leads to a constructivist perspective, which can be fraught with bias (conscious or unconscious) and errors.

This is no more evident than in the plethora of books touting some organization or company as great or wonderful. When the company that is touted as great or wonderful explodes, such as in the case of Enron, the constructivist flaw becomes obvious (though some “researchers” simply use this as the springboard to write another book about why they really weren’t wrong to begin with).

So, my caution for psychologists and other mental health professionals is twofold.

  1. If you are seeking to explain an extreme incident, make sure that any research you use to support your “reasons for why it happened” actually specifically studied the type of incident to which you are referring. Very few studies on workplace bullying will be useful in predicting extreme responses to workplace bullying. Don’t over-generalize if it hasn’t been studied.
  2. When working one on one with a client (as in counseling or clinical psychology), don’t assume that what comes out of that interaction necessarily applies to other clients or to people in general. One of the things we know about research is that sometimes tailored interventions for an individual work because of the idiosyncrasies of that person, but they don’t necessarily generalize beyond that person. So, don’t assume that what works for one will work for all, or that what works with extreme patients will do anything for the average person. That just isn’t how it works.

As for you (the average reader), when someone tries to explain away extreme behavior using previous research, consider whether that extreme behavior is something that could have been studied in the past. If not, ignore them.

Photo credit: http://www.flickr.com/photos/tmunki/507351879/

Comments are closed.