It’s the conversation nearly everyone dreads: the job interview. Having snuck away from the office, dressed in your nicest suit, you sit at a conference room table and try to please the people with the power to hire you. In just a short period, your task is to convince these strangers how good you’d be at this job.
But behind this uncomfortable ritual, so essential to securing employment, admission to school, and other plum rewards, is an even more discomfiting fact: As a tool for picking the best candidates, it’s almost entirely useless. Social scientists have argued in vain for years that the standard open-ended, unstructured interview does essentially nothing to improve the quality of the candidates selected. In fact, a new study now suggests, unstructured interviews aren’t just a waste of time—they may actually harm companies’ and schools’ ability to select the right people.
This may seem incredibly counterintuitive. As much as we hate job interviews ourselves, few of us would want to hire employees without meeting them. However good a person looks on paper, it’s tempting to think there’s some x-factor that will tell us if she’s right (or not) for the job and that our instincts, especially if based on long professional experience, won’t steer us wrong.
“This is just not something that people like to hear,” says Jason Dana, the new paper’s lead author and a visiting assistant professor at the Yale School of Management. “For whatever reason, we have a deep-seated need to feel that we can judge character.”
Given other criteria to judge by, however, like sales numbers, references, or test scores, the data say that an unstructured interview—a changeable series of questions that elicits a somewhat random set of information—can cloud our judgment. Whatever your gut tells you about that job candidate you interviewed, the researchers say, you shouldn’t listen.
It’s not easy to accept the notion that our gut is getting in the way—or that it’s not to our advantage, as job seekers, to have the chance to charm a potential boss. Even if experts advise that unstructured interviews are nothing but trouble, whether the rest of us can absorb that wisdom is another matter.
Though employers have been slow to catch on, studies since the early 1980s have shown that, when compared with other types of tests, unstructured interviews are one of the worst choices for accurately judging how well a particular person will do at a particular job.
Even experts in decision-making, though, have acknowledged that interviews have other advantages: They may help identify candidates who have better attitudes towards hard work, for instance. At the very least, the wisdom ran, interviews don’t weaken hiring decisions. But, Dana wondered, what if they do?
For a paper ultimately published in September’s issue of Judgment and Decision Making, he and his colleagues set out to test whether unstructured interviews might actually interfere with people’s ability to predict applicants’ performance. The study’s 76 participants, undergraduates at Carnegie Mellon, were asked to predict two fellow students’ GPA for the semester. Each participant was given data on the two students (age, major, class standing, course schedule, and cumulative GPA), and predicted one student’s GPA based on that information alone. The participant interviewed the other student before making a prediction. In a further twist, sometimes
interviewees were told to answer questions truthfully, and other times to answer according to a system that randomized their responses.
The students worried that their peers would notice this disjuncture. But that never happened. The study’s subjects found both the truthful and the random answers “somewhat informative,” the paper reports. And no matter what, conducting an interview led the subjects to make worse predictions than working only from the information they had on paper.
Dana’s study isn’t necessarily definitive: For one thing, it’s looking at a narrow situation in which the interviewer has a wealth of other data to work with. But it does point to where problems arise. Interviews give us so much information, the researchers say, that they pull our attention away from more relevant data. Human beings are natural storytellers. Given a series of facts, we use them to build a story that makes sense to us and keeps the randomness of the world in check. When the facts come from a meandering interview, those stories have a poor record of predicting how well an applicant will perform.
There are other options. Researchers have found that adding more structure to an interview, by asking all the candidates a set list of questions or coming up with a specific rubric on which to assess them, dramatically increases the validity of the information gathered.
“Let’s say you’re hiring insurance salespeople. You want to predict how many sales they make in a quarter,” says Scott Highhouse, a professor of psychology who’s looked at why we’re so attached to open-ended interviews. “You look at their sales records. Usually, the criterion we use is performance ratings by supervisors. But it doesn’t matter what the criterion is—the standardized approaches always outperform the traditional interview.”
In light of their findings, the new paper’s authors, who specialize in behavioral and decision research, have a simple recommendation: Don’t use unstructured interviews at all. “The assumption is, if I meet them, I’ll know,” Dana says. “People are wildly overconfident in their ability to do this, from a short meeting.” Instead, he suggests, talk to their references, who have hours of experience working with the person. Or look at their GPA, which, Dana points out, can be seen as “an aggregated opinion of a bunch of people.”
Of course, in the real world it’s not always so simple. References won’t or can’t talk. Not every job produces easily identifiable data on which employers can judge applicants. Whether it’s empirically valid or not, interviewers often want to establish a personal and informal contact with a potential hire—a 2004 study found that’s one of the main reasons that people stick with the unstructured format. And while big corporations generally care most about predicting job performance and retention, small businesses or households, for example, may place greater value on interpersonal factors. “If you want to like the person, it becomes more interesting,” says Dana.
Still, even for choices like picking a nanny, where no one would make a hire without meeting the candidate, Dana cautions against attaching too much importance to a short in-person encounter. “For a nanny, I’d have them meet my child and see how they interact,” he says. “If they rub me wrong when I meet them for 20 minutes, I’d still probably hire them, based on the recommendation of a person that had them working for years.”
There are also situations where the overall group dynamic and balance trump plain performance. In choosing a college class, for instance, admissions officers might be trying to predict not only GPA but also students’ engagement with campus life—and that might make interviews seem more enticing. In 2008, North Carolina’s Wake Forest University did the exact opposite of what Dana’s study recommends: The admissions office made it optional for students to submit SAT scores and put greater emphasis on unstructured interviews. One of the school’s goals was to increase diversity on campus. “We’re not saying that we can find the student who can finish their freshman year with a 3.6 rather than 3.4,” says Martha Allman, the school’s dean of admission. “We’re looking for students who have different backgrounds and different interests and talents.” So far, it’s worked: The student body has become more racially, more socioeconomically, and, says Allman, even more philosophically diverse. “It’s not a scientific process that’s going to give an absolute prediction,” she says. “But what is?”
Organizational psychologists like Highhouse recommend that, if you’re going to bring candidates in for interviews, the interview should have a predetermined structure, and that the person should be judged in measurable ways on criteria related to the job. Structured interviews can offer useful information where unstructured ones just cloud the case, he says. By creating a fixed set of criteria, they can combat the unconscious bias that most people have towards hiring someone like them—who shares their hobbies, personality type, background, skin color, gender, or age—and keep the focus on actual performance. “If you’re hiring somebody to fold clothes, have them fold clothes. If you’re hiring someone to do consulting, have them do a consulting case,” says Dana. “The more structured the interview, the better the predictive potential is.”
Beyond the very human desire to trust our gut, there’s a practical difficulty with structured interviews: They take more time and effort to prepare and evaluate. Given this logistical hurdle on top of the emotional one, organizational and decision making experts have spent so many years trying to wean people off unstructured interviews that they’re almost resigned to how attached people are to them.
Even Dana knows his new research will probably not be enough to convince you to hire a new co-worker sight unseen. Asked what was the very least people should take from his research, he offered this advice: “Whatever you think you’re getting out of interviews, it’s probably too much.”