Last week the New York Times published an interview with the authors of Sway, a new book that documents a series of psychological forces that lead people to disregard logic and act irrationally. One chapter in Sway deals with the phenomenon of the job interview, and describes the "first date" format of job interview that is so ubiquitous in America today. Most job interviews resemble first dates because employers utilize an unstructured "get to know the candidate" approach in which the interviewers try to establish a rapport with the interviewee, discover common interests, and form an impression as to whether the candidate will be a "good fit" at the company. The problem with "first date" interviews is that asking candidates to "describe themselves" or assess their "strengths and weaknesses" too often leads to canned answers that don't reveal much about future performance.
The authors argue that managers consistently overestimate their ability to form objective opinions based on interviews, and argue that structured interviews are much better predictors of future performance because they focus on relevant, objective data. The fact that unstructured interviews aren't a very good way of gathering objective data on candidates isn't news to anyone who is familiar with research in this area. What's surprising is that the authors actually significantly understate their case when they conclude that:
"As counterintuitive as it sounds, you don't need interviews at all. Research shows that an aptitude test predicts performance just as well as a structured interview."
Actually, what most research shows is that employment aptitude tests are far better predictors than are structured interviews. (The authors refer often to a meta-analysis, but don't reference which meta-analysis they are using in the Notes. The most comprehensive meta-analysis of employee selection techniques is the ubiquitously cited Hunter and Schmidt from the 1990s, which shows that aptitude tests are better predictors than are structured interviews.)
Employment tests predict performance more accurately than interviews do precisely because they yield objective, relevant data about a candidate's problem-solving ability, critical thinking skills, and job-relevant personality traits. In short, they provide the kind of data that is not susceptible to "Sway" -- the authors' catch-all phrase used to describe the effect of irrational impulses on human behavior. Although I think "Sway" simplifies the issues somewhat, the authors' basic point about job interviews is very sound. After all, if you wouldn't marry someone based on hitting it off on a first date, should you be hiring based on an interview?