Asymmetrical Tests

Abstract

It is often thought, or presupposed, that investigative tests done for evaluative purposes should be symmetrical—that is, they should be equally capable of giving positive and negative results. This is an error in the logic of evaluation, and it arises from confusion between the asymmetrical ability of a test to yield information and its propensity to yield biased information. Although ‘asymmetrical’ does connote one-sidedness and we often use ‘one-sided’ to mean biased, we can and should use the terms more carefully. Some tests can only yield information about the faults of a program (or product, or person, etc.)—or only about its virtues—while others can yield one of these more reliably than the other, and yet others are symmetrical in their treatment of merits and demerits. All of them gather relevant information for evaluative purposes, and one cannot conclude that any of them are biased simply because they are asymmetrical. The point is important because in many situations, one may only have access to asymmetrical tests and this does not support the claim of a biased approach as long as there is more than one test in the battery used to evaluate, and one test’s asymmetry is balanced out by the other test(s).