The killing of innocent people at Newtown, Conn., and other sites has brought about a renewed effort to find measures to keep guns out of the hands of those who are likely to commit such killings. Preventing the sale of certain types of guns and ammunitions used in these acts has received the most scrutiny. At the same time, while background checklists have accorded almost unanimous support, there has been surprisingly little conversation about what types of people should be on those lists, and even less about any possible harmful effects to people who get on the list but would never kill anyone.
House Bill 1229 will expand the Colorado Bureau of Investigations ability to keep guns out of the hands of criminals to include those who are mentally unstable and perpetrators of domestic-violence. These extensions have been justified from a common-sense viewpoint. In this column, I will address those extensions from a behavioral science and statistical viewpoint hopefully to add a cautionary note to their acceptance.
There is an important social policy distinction between gun control and background checklists. Gun control restricts the sale of certain types of guns and ammunition to all people. Background checklists restrict the sale of all guns to certain types of people. Gun control has constitutional concerns. A background checklist in addition has discrimination concerns. Many people will be on that list who would not use guns inappropriately, and being on the list could have a harmful affect on their lives. Proponents will argue that all it would do is prevent a few innocent people from purchasing a gun. But it also tells society that people on the list should be feared. Furthermore, lists of this kind almost always end up with broader uses than just preventing gun purchases.
A thorough discussion of these types of issues can be found in Kenneth Hammonds book Human Judgment and Social Policy. Hammond is a professor emeritus at the University of Colorado, Boulder.
Hammond explains the issue in statistical terms. Because we do not have a perfect correlation between types of people and their propensity to act in certain ways (in this case, to kill), any proposal to identify such killers will have four consequences: Some people who will commit such killings will correctly be on the background checklist (called true-positives); some people who would never commit such killings will correctly escape being listed (true-negatives); some who would never commit such killing will be wrongly on the list (false-positives); and some who will commit such killings will not be on the list (false-negatives). The concern is that false-positives are never acknowledged in background checklist discussions, let alone the possibility that those peoples lives may be detrimentally affected by being on the list.
A legitimate question to ask at this point is why havent statisticians been able to find characteristics that identify killers (true-positives) without implicating innocent people (false-positives)? A common perception is that these types of killers are so obviously deranged, and their actions are so blatantly bizarre, that any close observation should be able to detect them before they commit such acts. Unfortunately, as Ivars Peterson describes in his book The Jungles of Randomness, such after-the-fact perceptions are misleading. Statistically, these are rare events, and as such, detecting them in the noisy world of people with mental problems and unusual behavior quirks is extremely difficult.
Because statistically, the killing of innocent people is rare, the net cast to identify potential killers will be unreliable. Some killers will be identified, but a large numbers of people who would not commit killings also will be included. It is also statistically assured that any attempt to add types of people to the list so as not to miss likely killers, as HB 1229 is intended to do, will increase substantially the number of innocent people listed. The relevant question is how many false-positives is society willing to tolerate in order to identify one true-positive? That judgment is not easy to make when there is the possibility that harm can result to innocent people by simply having their name on the list.
This discussion is not meant to suggest that background checklists be eliminated, but rather to caution those rushing to do something that there is a price to pay in adding new types of people to the lists. Until more is known about what makes certain types of people turn into killers, it would be prudent to reject legislation such as House Bill 1229. As professor Hammond reminds us, if the accuracy of the judgment that forms a social policy is uncertain, errors will be inevitable, and injustice unavoidable.
Garth Buchanan holds a doctorate in applied science and has 35 years of experience in operations research. Reach him at firstname.lastname@example.org.