Want to Debias Hiring? Change What Hiring Managers Focus On

Take a look at the two orange dots below. Which one is bigger? If you’re like most people, you can’t help but see the orange dot on the right as larger. However, when the blue dots disappear, removing the “context,” it’s clear that the orange dots are the same size. This is known as the Ebbinghaus illusion, and it illustrates a fundamental principle of our psychology: context matters. This is true for judgments of all kinds, ranging from how we see the world to how we make choices in our everyday lives.

Imagine that it’s the late afternoon and you’re craving a snack. You reach into your stash of snacks and pull out two options: a granola bar and a chocolate bar. How do you decide which snack to have? The granola is healthier than the chocolate, but the chocolate is tastier. Now imagine that instead of just these two snacks, you have a third: a gross but extremely healthy protein bar. Rationally, the protein bar shouldn’t affect how you feel about the two original options, and yet it makes the granola bar more attractive, because the granola bar now seems like a compromise on both health and taste. This well-documented shifting-of-preference phenomenon is known as the decoy effect.

What if instead of deciding what to eat, the decision was whom to hire? Our research reveals that how we think about and evaluate people is not immutable; it depends on the context in which we’re evaluating them, specifically the context of other people. Across seven experiments (in press at the Journal of Personality and Social Psychology), we tested whether we could systematically shift people’s preferences between two candidates by including a third “decoy” candidate. We found that we could, in certain circumstances. To differentiate the candidates, we used two fundamental dimensions of evaluation in business settings: aptitude (or competence) and attitude (or warmth).

Our research reveals that how we think about and evaluate people is not immutable; it depends on the context in which we’re evaluating them, specifically the context of other people.

In one scenario, research participants were presented with information about two job candidates: one candidate with high aptitude scores but lower attitude scores (our competent, cold fish) and another candidate with lower aptitude scores but high attitude scores (our beloved bumbler). In this case, unsurprisingly, people said they would be more likely to hire a competent, cold fish than a beloved bumbler. In other scenarios, participants were presented with the same two candidates along with a third, decoy candidate. Just as we predicted, adding a hyper-warm but useless candidate suddenly made people evaluate our beloved bumbler more positively than when our competent, cold fish and beloved bumbler were the only ones in the candidate pool.

This is because the hyper-warm, useless candidate made our beloved bumbler seem like a reasonable compromise on both dimensions. But the converse wasn’t true; adding a hyper-competent S.O.B. didn’t seem to change people’s feelings about our competent, cold fish.

One possible explanation for these findings is that because this was a hiring scenario, all of our participants were (very reasonably) relying more on the aptitude information to make their choices. This meant that adding a decoy that drew their attention to the aptitude dimension wouldn’t really have an impact because they were already focused on aptitude. If this was right, then asking people to focus their evaluations on attitude should have flipped our results. This is exactly what we found. In another experiment, we told participants that the hiring firm wanted evaluators to weigh attitude more heavily in their decisions; in this case, adding the hyper-competent S.O.B resulted in people evaluating our competent, cold fish more positively than when our competent, cold fish and beloved bumbler were the only ones in the candidate pool.

These experiments revealed that decoys seemed to be having their effect by tuning people’s attention more toward either the attitude or the aptitude of a candidate. Of course, one way firms can do this in the real world is by emphasizing one versus the other dimension in a job description (as we did in the experiment we just described). But another strategy that should exert the same effect is by exposing people more to one dimension versus the other. For example, when hiring managers look over résumés, in most situations they are free to focus on whatever information they want. In fact, one of the most salient features of a résumé is the applicant’s name, which usually communicates gender (and in some cases race and socioeconomic status). What if instead of letting evaluators look over résumés freely, we showed them pieces of information one at a time and, more importantly, made sure evaluators saw important qualifications repeatedly?

We tested this idea by asking experiment participants to compare two candidates—sometimes on the attitude dimension, and other times on the aptitude dimension. We showed participants a piece of information that communicated the candidate’s attitude (Candidate A was nominated for a mentoring award; Candidate B won a mentoring award) or their aptitude (Candidate A had a 93 percent managerial satisfaction rating; Candidate B had only a 75 percent rating). Underneath those pieces of information, we asked participants to rate which candidate was either warmer (attitude) or more competent (aptitude).

Take bias out of the decision-making process rather than the decision-maker.

Crucially, a third of our participants made equal numbers of aptitude and attitude comparisons (in other words, they were asked, “Who is warmer?” and, “Who is more competent?” an equal number of times), one third of participants made more aptitude than attitude comparisons (i.e., were more often asked to judge, “Who is more competent?”), and the final third made more attitude than aptitude comparisons (i.e., rated candidates on warmth more often). As we predicted, participants who were exposed more often to attitude-based comparisons said they were more likely to hire our beloved bumbler whereas participants who had made more aptitude-based comparisons said they were more likely to hire our competent, cold fish—even though everyone had received the same amount of information about both candidates in all of the conditions. That is, participants had all the same facts about the candidates; we had simply repeated the aptitude-related facts in the aptitude-emphasis condition, and similarly repeated attitude items in the attitude-emphasis condition.

This work has implications for the quest across industries to make decision-making less biased. Beyond the business environment, there are other consequential domains in which people jointly evaluate sets of candidates, including housing, education, and health. While we only tested the dimensions of attitude and aptitude in business-hiring contexts, we plan to investigate other dimensions (e.g., how black or white the candidates faces are) in other consequential contexts. Currently, our findings suggest that if you want people to focus on qualifications rather than demographics, for instance, you should create structured candidate evaluations that tune evaluators’ attention to candidate qualifications and not demographic information.

With choices as important as these, we’ve learned that we can’t rely on changing the stereotypes and implicit prejudices of evaluators, because these strategies are resistant to long-term change and may backfire. Our research suggests a more promising avenue: take bias out of the decision-making process rather than the decision-maker.