Stubborn Reliance on Human Nature in Employee Selection: Statistical Decision Aids Are Evolutionarily Novel

The limited use of decision aids has been a conundrum in our field for years, and Highhouse’s (2008) article is one of several attempts to explain and solve the problem. Lawshe and Bolda (1958) argued that man-agers could not understand themetric of val-idity coefficients, so they developed a more user-friendly metric, the expectancy chart. Others suggested that managers were not aware of the economic value of valid predic-tors, and they worked out economic utility equations (e.g., Schmidt, Hunter,McKenzie, & Muldrow, 1979). Dawes (1971) com-plained that decision makers did not under-stand the value of mechanical methods for combining predictor scores, and he encour-aged better communication about their ben-efits. Yet, none of this had much effect on persuading managers to use decision aids (Hastie & Dawes, 2001; Latham & Whyte, 1994). Highhouse offers another interpreta-tion: employers ’ belief systems. People still believe (erroneously) in near-perfect predic-tion estimates and that predictive ability can be improved with experience. He recom-mends that we do a better job of communi-cating to managers what the decision-making literature has found about uncer-tainty and error. These explanations and prescriptions have two elements in common. First, they assume that managers are ‘‘rational’ ’ actors: If presented with convincing data, they can be persuaded to change their habits. Second, there is a communication problem. Industrial–organizational psychologists are not effectively communicating the benefits of selection technologies to managers; if we did a better job communicating, managers would see the light. Of course, people can, to some degree, be persuaded by scientific evidence, and clear communi-cation can be effective. However, the problem goes deeper than this. Some beliefs are readily changed by new information, reflection, and analysis; others are more stubborn. Beliefs about assessing people and predicting their behavior fall into th