Profiling and Self Defense

The shadows are on the darker side
Behind those doors, it's a wilder ride
You can make or break, you can win or lose
That's a chance you take when the heat's on you
And the heat is on
--Glenn Frey

Nice discussion of profiling by Professor Williams. Profiling is a method of economizing information costs by using readily visible characteristics as a proxy for more difficult to observe characteristics. It is not unlike reductionist statistical operations that seek to obtain high predictive value from a few variables thought to be predictors.

Humans constantly profile other humans. When conversing with friends we study facial expressions and body positioning to gauge their responses to our points of view. Dating services study characteristics of members to infer good matches. Colleges use standardized tests of applicants to predict future scholastic performance.

If people did not engage in profiling, then they would need to spend more time and other resources collecting information about people before drawing conclusions. Because people seek the most benefit from the least cost, profiling is an attractive way to economize scarce resources.

However, since profiling generates predictions or conclusions under conditions of uncertainty (in this case, imperfect information), it is subject to error. Two categories of error are possible. A type 1 error (a.k.a. false positive) occurs when profiling leads to the conclusion that a relationship exists when it really doesn't. A type 2 error (a.k.a. false negative) occurs when profiling leads to the conclusion that no relationship exists when it really does.

To exemplify, let's employ a context where the use of profiling attracts considerable attention: predicting criminal behavior. A friend and I sit down to eat at a casual restaurant. Customarily, I like to sit with my back to a wall so that I have a good view of things. I proceed to scan the room for anything unusual. An unshaven younger man dressed in a long coat stands over by the bar. He appears to be alone and does not converse with others at the bar. He looks down at his watch and then looks out the front door. He does this a few times.

Perhaps he's just waiting for someone. Or perhaps he is preparing to engage in criminal activity. Robbery, shooting, bombing...

I am engaging in profiling. I am observing a limited number characteristics and behaviors of this man in the context of the surrounding situation and using my observations as the basis for forecasting criminal conduct.

If the man's wife subsequently walks in and they proceed to a table, or if the man finishes his drink and walks out the door never to return, then my forecast would be likely be wrong and I may allocate no further attention to this man. If so, I will have (thankfully) committed a type 2 error--I proposed a possible relationship that did not turn out to be true.

While scanning the room I also notice three mid-aged women sitting at a table across the way. They are dressed casually and chatting among themselves. They appear to be enjoying each other's company over dinner. Because I notice nothing suspicious, I give this group no further thought as to potential criminal behavior.

If, subsequently, one of the women suddenly stands up, pulls out a machine pistol, and screams that everyone here is going to die for jihad, then I will have (unfortunately) committed a type 1 error. I saw no possible relationship between these people and criminal activity when in fact there was one.

Note that in these two cases, the cost of the type 2 error was small while the cost of a type 1 error was large. This is typically the case, and implies that it is often worthwhile to profile in order to identify potentially dangerous situations.

As long as I do not invade the pursuits of others to do so (if I did so I would be an aggressor rather than a defender), profiling can be an effective means for economizing valuable resources, such as time, that assist in effective self-defense against aggression.