Profiling in a security conscious age

    Profiles in society describe the characteristics of people who are
    likely to behave in particular ways, and, to the extent that they are
    accurate, they increase the likelihood that individuals engaging in
    certain behaviors will be identified.

    The use of profiles is ubiquitous _ for example, to catch criminals
    (from terrorists and drug couriers to car thieves and tax cheats), to
    parole prisoners, even to determine the rates we pay for auto
    insurance. It’s next to impossible to scrutinize every person crossing
    a border or boarding a plane, so a profile can tell authorities who is
    more likely to be doing something criminal.

    The Internal Revenue
    Service cannot carefully review tens of millions of tax returns, but a
    profile can tell the IRS whose tax return deserves special scrutiny.
    Our prisons are overcrowded, but some prisoners are too dangerous to be
    paroled: Which ones would make better risks? Some drivers are more
    careless than others or drive on more dangerous roads. Doesn’t it make
    sense for insurance companies to use a profile to determine who should
    pay more for insurance so that you and I, the good drivers, can pay
    less?

    Consider also that all of us use profiles of one sort or
    another in ordinary ways. These profiles may not be based on
    statistical evidence, but they are profiles nonetheless. Who hasn’t
    used experience to make educated guesses at a party about who might be
    interesting to talk to, based on how they look or comments they have
    made? Parents engage in profiling of sorts when they encourage their
    children to look for romantic partners who are ambitious, rather than
    lazy, because experience tells them that ambition increases one’s
    chances for a happy life.

    So how could it be argued that
    profiling _ looking in a certain direction for certain kinds of people
    who are likely to behave in certain ways _ can have very negative
    consequences?

    First, we need to remember that profiles direct
    our attention toward some people and away from others. We should
    recognize, however, that “shifting our gaze” actually increases the
    likelihood of finding evidence that increases our confidence in our
    profiles _ even when individuals who do not fit the profile are
    engaging in the targeted behavior just as frequently as those who do.

    If there’s just as much mischief occurring on the left and right sides
    of a classroom, but the teacher looks twice as much to the left as to
    the right _ because of suspicion resulting from a profile about who
    will misbehave (e.g., the boys with the baggy pants) _ guess what
    happens. The teacher will find twice as much misbehavior on the left
    side. And the teacher’s observation will give credence to the profile,
    even though the profile is wrong.

    Second, we should keep in mind
    that when using a profile to find the people we’re looking for, there
    are two mistakes we can make: a false-positive error and a
    false-negative error. For example, the former may identify an innocent
    person as a potential terrorist, and the latter may identify an actual
    terrorist as an innocent person. Given a choice, most people would
    prefer to err on the side of caution _ avoiding a false-negative error
    _ considering the potential consequences of failing to identify a
    terrorist.

    Of course, our reaction might be very different if we
    were the ones being misidentified, and experiencing the consequences of
    being labeled a terrorist.

    Third, we need to recognize that
    decreasing false-positive errors usually increases false-negative
    errors, and vice versa. When a profile casts too small a net, it
    catches only a few of the people we’re looking for, and misses many
    others. However, a profile that casts too big a net catches not only
    some of the people we’re looking for, but also people who aren’t
    engaging in the behavior we’re trying to prevent.

    Unfortunately, profiling is not a perfect science, so the nets that are cast are always too big or too small.

    The above concerns demand that we ask and try to answer two intimately related questions:

    1) How can we evaluate the magnitude of the consequences of
    false-positive and false-negative errors produced by less than perfect
    profiles?

    2) Who does this evaluation?

    Answers to these
    questions are vitally important as we search for the proper balance
    among people’s demands for security, civil liberties, and dignity. This
    last element, dignity, has been ignored in most discussions of the
    benefits and drawbacks of profiling _ except by those who experience
    the effects of being wrongly identified.

    Those of us not singled
    out by authorities or fellow citizens at an airport, in a car, or on a
    nearly deserted street in someone else’s neighborhood probably have a
    very hard time understanding the impact that repeated false-positive
    identification errors can have on the individual, and the resulting
    alienation toward society.

    Alienation may seem to be of interest
    mostly to sociologists and psychotherapists, but if we’re prone to
    wondering why patriotism isn’t what it used to be, or why people aren’t
    buying into the political system we think is the best in the world,
    perhaps a consideration of people’s alienation becomes less esoteric
    and more pragmatic.

    Some have suggested that certain
    characteristics shouldn’t be used in a profile, even if they improve
    the profile’s quality, because they produce too much harm for those who
    are innocent but fit the profile (a false-positive error).

    Some
    might respond that such a false-positive error is regrettable, but that
    those misidentified must “get over it,” or “that’s the price we have to
    pay for security” _ although they usually mean “that’s the price
    someone else has to pay.”

    If these thoughts have crossed your
    mind, consider the sense of unfairness you experience if you pay a lot
    for auto-liability insurance because of the town you live in, your age,
    your sex, or your car _ even though you’ve never been in an accident or
    gotten a speeding ticket.

    Of course you can move to another
    town, and you’ll eventually get older _ and, these days, you can even
    change your sex if you’re so inclined. But the color of your skin, your
    ethnicity, and your country of origin will always be what they are now:
    immutable.

    Imagine if Catholics, Jews, or Muslims cheated on
    their taxes a bit more than people of other religions. Would we
    encourage the IRS to use religion as part of a profile to spot tax
    cheats? Or would we ignore the small statistical advantage that might
    be produced by including this characteristic in the profile of a tax
    cheat because it would produce an unacceptable number of mistakes with
    major consequences for others?

    Is it important to think about
    how we might assess the magnitude of those consequences and how that
    assessment might interact with the number of people who experience
    them? You bet it is. Errors produced by less than foolproof profiles
    (i.e., all profiles) cause injustices that hurt others _ our fellow
    citizens, our neighbors, even members of our own family _ in ways that
    can scar them, ourselves, and our society.

    As a society, we not
    only have to decide whether we want to give up our civil liberties to
    increase our security. We also have to decide whether, or when, it’s
    wise to give up our dignity in the service of security.

    Finally,
    we must determine whether the increase in security with profiles is
    real or illusory, so that we can determine whether the price is
    ultimately too high.

    (Arthur Frankel is a professor of psychology at Salve Regina University in Newport, R.I.)