Virginia Tech Shootings and Hindsight Bias
Megan McArdle, guesting at Andrew Sullivan’s place, argues that the initial reaction to yesterday’s Virginia Tech massacre may be clouded by hindsight bias.
[E]ven if all mass-murderers did write scary prose, or make sweeping apocalyptic statements, or otherwise give some signal of their impending meltdown, the signal wouldn’t do us any good, because mass murderers are really, really rare. You’ll have a thousand false positives for one false negative. In hindsight, we can always pick out some clue to what was about to happen. That doesn’t mean that we can, or should, see those things beforehand.
Related is the criticism of administrators for sending students to class after the first murder, or of police for not locking the campus down immediately. This is a classic problem with recriminations: we tend to assume that the fact we had a bad outcome means we made a bad decision. But in an uncertain world, this is ludicrous. Good decision making concentrates on the most likely events, not the wild outliers.
The errors of a false negative (what social scientists call “Type II errors”) can be quite catastrophic. Had Virginia Tech officials sounded the warning bells the instant they knew of the first shootings, they may well have saved thirty innocent lives. Similarly, had everyone who had the slightest suspicion that Cho Seung Hui was emotionally unstable brought that to the attention of the proper authorities, thirty two innocent lives may have been saved.
As Megan suggests, however, the consequences of overreaction to the innumerable false positives (“Type I errors”) would be far more debilitating. Witness, for example, the millions of man hours wasted each year providing the illusion of security at the airports.
Do we want every odd teenager turned in to The Powers That Be for certification that he’s not a potential serial killer? Given what we know of bureaucracy, we can rest assured that a whole lot of people would be unnecessarily locked away in order to ensure that false negatives don’t occur. (Which is why it’s unlikely we’ll ever see a terror alert level below Yellow unless the system is scrapped.)
Several years ago, during the first year Steven Taylor and I were teaching at Troy State, a young student was driving recklessly in a severe rainstorm and was killed. Following that tragedy, university officials decided that they would shut down the university during heavy rain storms and, on occasion, even the forecast of same. Worse, if it was raining at all, the policy was to allow students an excused absence at their discretion if they did not feel safe driving. This applied, incidentally, even if the student lived on campus and was walking to class. There’s no way of knowing if this policy saved any lives. It did, however, deprive students of thousands of hours in the classroom for which they had paid.
Human lives are precious and people in leadership positions have a responsibility to ensure the safety of people under their care. But let’s not pretend that safety doesn’t come at a price. Protecting ourselves against infinitesimal risks may make us feel better but it almost certainly senseless.
UPDATE: I should add that I’m not arguing that the officials at Va Tech acted properly, just that we shouldn’t come to the knee-jerk reaction that they screwed up just because something bad happened. My inclination is that they should have notified the campus that shootings had occurred and perhaps even canceled classes while someone was on the loose. Still, they may well have been justified in thinking it was a murder-suicide situation. We simply don’t know at this point.