Profiling

There has been much coverage in the news media about “profiling”, especially in NYC during this mayoral primary campaign. Much of it has been critical of Mayor Bloomberg and the NYC police department which utilizes a stop-and-frisk program in an attempt to have a positive impact on crime in the city. Each candidate during this primary season is staking out a pro- or con- position on stop-and-frisk. A recent federal court ruling has stated that the NYC program violates people’s rights against unreasonable search. “The judge ruled that police officers have for years been systematically stopping innocent people in the street without any objective reason to suspect them of wrongdoing” (August 12, 2013, New York Times). The stop-and-frisk program is said to be based on profiling, that people who fit certain characteristics or profiles are picked out by the police department and “selected” for participation in the program. The huge number of people who are stopped suggests that the program is in need of an overhaul. It appears to be not so much an exercise in profiling as it is an exercise in stopping a large number of people and seeing what turns up, and apparently a federal judge agrees. But as a consequence of this and other bad publicity over the years profiling has been cast as evil.

If profiling is evil, that means we are all evil. Profiling is a built-in feature of being human. Each of us use profiling every single day to assist with quickly categorizing the vast quantities of information which impinge upon us and to help us make decisions that range from very simple to quite complex. Whether an application of profiling is good or evil is based upon how it is used and the characteristics utilized in creating the profile. Sigmund Freud for instance stated, “When you meet a human being, the first distinction you make is ‘male or female?’ and you are accustomed to making that distinction with unhesitating certainty.” You make that distinction with unhesitating certainty because of a profile you carry around regarding which body shapes, facial characteristics, etc. are classified as female and which as male. Taken in total those features and characteristics represent a profile of maleness and femaleness.

Freud’s view of profiling maleness and femaleness tends towards physical characteristics, but there are other categories, such as demographics and behaviors which are also used to build profiles.  Levitt and Dubner in Super Freakonomics describe a profile that was created in the UK to help pick out potential Islamic fundamentalist terrorists. Now, if you are trying to pick out potential Islamic fundamentalist terrorists, chances are that they won’t have a name like Buffy Willis, or Bruce Fleming, and in fact the research shows that Muslim names predominate among Islamic terrorists (no surprise), just as Irish names predominated among Irish terrorists who were fighting the British previously.

That particular characteristic, name type, is very superficial, and can be very misleading as a profiling tool as it tends to sweep into its net a huge number of false positives, those with Muslim names who are not terrorists (the vast majority). The researcher that Dubner and Levitt cite ended up creating a terrorist profile with predictive ability (being able to pick a terrorist out of the crowd) by adding in addition to demographic characteristics a behavioral characteristic which they cannot disclose. At the risk of getting a little technical here, they state that most of the profile variables being used to predict who is a potential terrorists and who is not tended to be binary (you either have that characteristic or you don’t), and a shortcoming of binary variables is that they don’t have much statistical variance (the degree to which over a large number of people you will get a wide-range of responses). When you don’t have much statistical variance you lose predictive power. The behavioral variable that enabled more accurate profiling of terrorists was on a continuum with higher levels of this particular behavior being much more common among those with terrorist tendencies. This gradation of responses, from lower being less likely to be a terrorist to higher levels being increasing likely has a much greater amount of variance and hence predictive power.

On a more mundane level, one research project we conducted for a high technology firm sought to define what profile would be seen between those sales people who made the President’s Club by exceeding their beginning of the year quota by a given amount, vs. those sales people who fell short. By far, the best differentiator was not gender or tenure or some other demographic variable, but rather a behavioral characteristic, the degree to which the sales person sold in a collaborative fashion. Selling in a collaborative fashion is not binary but rather exists along a continuum and those sales people with higher and higher amounts of this particular characteristic, in this company’s environment, were more successful. The ability to sell collaboratively has all sorts of selection, training, performance management, technology and support systems implications which are then potentially able to be fine-tuned by a client company.

Here is another example of how behavioral variables on a continuum are better than binary one’s commonly used in profiling. If you have ever passed through airport security in the USA vs. Israel you will notice a distinct difference in how passengers are screened. In the USA you are essentially screened for objects, metal objects or liquids that you are not supposed to carry into the airline gate area. You may be asked one or two questions about whether someone has given you anything to carry for them or where have your bags been, but the screening process is about objects (either you have them or you don’t – a binary variable) and not about behaviors. In Israel the focus of the screen is about your behaviors, why are you there, where are you going, what were you doing, tell me about your family, do you speak Hebrew, where did you learn it, etc. The screening process is about your behaviors, past, present and future which are scored and very hard to fabricate out of whole cloth. The risk score determines if the person needs follow-up additional screening. Israel’s security screening track record is exemplary.

With all the commotion about profiling these days it is clearly obvious while some would desire for profiling to just go away – that any kind of profiling is inherently bad. But in reality we profile constantly and it is not going to go away. Profiles you have in your head of store characteristics, which may determine where you will shop, according to the type, quantity, quality and costs of the goods are common. Which of your neighbors will end up being your close friends based upon profiles of shared interests and characteristics happens all the time. A doctor diagnoses a patient by examining and logically thinking through profiles of symptoms commonly seen in various illnesses. The list goes on and on. The issues around profiling are not about it being inherently good or evil, but rather how we choose to implemented it and the validity we are able to ascribe to it in day-to-day practice.

© 2013 by Jeffrey M. Saltzman. All rights reserved.

Visit OV: www.orgvitality.com