Acts, omissions, and votes (written Dec. 10, 2000)

Jonathan Baron

Research on human judgment has revealed a common and widespread bias: harm caused by action is weighed more heavily than harm caused by omission. For example, suppose your child has a ten percent chance of getting a type of flu. A vaccine, the only one available, will prevent this kind of flu. But the child has a five percent chance of suffering from the side effects of the vaccine, and these effects are just as serious as those of the flu itself. Should you get the vaccine for your child?

Many people would not get the vaccine. They feel that they are not responsible for the harm that might result from inaction, even though they had the option to prevent this harm. If we always chose the option with a lower probability of harm, we would all suffer less harm in the long run. Yet, human intuition often favors inaction, even when the risk of harm from inaction is greater.

This intuitive principle applied in three ways to the arguments against further recounts in Florida, as expressed in the dissents to the Florida Supreme Court decision of Dec. 8, and elsewhere.

First, manual recounting creates unfairness. People are treated unequally, thus potentially violating the equal protection clause of the U.S. Constitution. Different standards will be used by different counties, and some counties would not be recounted at all. (The remedy to the last problem, however, was embraced by the court majority, namely, to recount all undervotes in the state.) On the other side, though, recounting reduces unfairness, so inaction leads to unfairness too. It is unfair that votes are left uncounted just because they were in a county that used punched-card ballots.

Second, manual recounting violates rights of voters to have their vote counted, because it leads to error. Yet, rights are also violated by failure to recount. The language of rights does not lend itself to computation, but computation seems to be the only solution. Which group of violations is larger? As the computer-scientist John McCarthy states (in his Stanford University web page): ``He who refuses to do arithmetic is doomed to talk nonsense.''

Third, manual recounting creates error. Repeated examination of the ballots and liberal standards for recording intended votes (such as counting dimpled chads) will cause votes to be counted that were not intended. Surely such errors happen, but failure to recount manually will lead to error of a different sort: the failure to record intended votes.

The important question is which error is greater. Expert testimony and even experimentation (which could be conducted quickly) would be relevant, but, in the absence of such testimony or experimentation it seems likely to me that the errors created by repeated recounting or by counting dimpled chads are small compared to the errors that result from failure to recount. This is, in any case, the question that the courts should address.

The standards of the law are affected by a common human intuition, that harms of action are worse than otherwise equivalent harms of omission. This in itself is not surprising, as the law is the result of human judgments accumulated over centuries. It is disturbing, though, because it leads to consequences that are worse than they could be.

It is also disturbing because the court's remedy is easy: estimate which error has a greater probability, as best this can be done. (If this is difficult, that is no more of an excuse for inaction than for action.) The bias toward harms of omission thus reflects a deep innumeracy, not just a lack of knowledge but also a disrespect for numbers and numerical thinking. Because the problem is deep, the remedy for it must also be deep, a change in culture, beginning with education and journalism.

File translated from TEX by TTH, version 2.73.
On 17 Dec 2000, 16:56.