Can a machine make better hiring decisions than a person?
A recent large-scale analysis comparing expert judgments to computerized algorithms found that hiring algorithms beat the experts by a wide margin.
Overall, algorithms improved accuracy in predicting job performance by more than 50%.
Perhaps even more important is the issue of fairness.
Hiring algorithms are not only more accurate, but also tend to increase the diversity of people selected. Hiring based on intuition seems to overlook high-potential candidates from underrepresented groups.
This seems to make sense, given evidence that organizations with greater racial and gender diversity are also more productive and have greater market share, more customers, more sales, and higher profits.
Here’s how hiring algorithms beat human judgment.
1. Computers don’t need to rationalize
When a person makes a decision, it is usually “framed” in a certain way. For example, we think quite differently about rejecting a job applicant than we do about simply hiring someone else.
Even though it is exactly the same outcome.
It can be unpleasant to think about denying someone an opportunity, and it can potentially reveal our own biases. So we often don’t think about it at all. Instead, we tend to focus attention on the people who are hired. We don’t really want to know what we might be missing.
Computers don’t feel uncomfortable about such things. They simply compare everyone based on the available information. No need to ignore any information.
2. Determining what actually matters
People have their own theories about the type of person they want to hire. These theories vary in accuracy, and are difficult to verify because we often look for evidence to confirm those theories, rather than evidence that might refute them.
Computers don’t care if they are right or wrong, and readily update their predictions whenever necessary.
3. Ignoring things that don’t matter
Our theories about the type of person we want to hire might not be that accurate, and therefore tend to include things that are irrelevant or even counterproductive.
Even if we aren’t deliberately applying a theory, there are a variety of biases that can influence our decisions without our intention or awareness. For example, we might tend to hire people that we like (perhaps because they are physically attractive or similar to us in some way). It can be difficult for a person to ignore irrelevant information about an applicant, even when trying to do so.
Computers only consider the information that they are given for a specific task.
4. Applying criteria consistently
Even if we do have a great theory about the type of person we want to hire, and are able to ignore irrelevant information, there is the problem of consistency.
It would be incredibly difficult (perhaps impossible) to treat every person in the exact same way, and hold them to the exact same standards. For each individual, it is likely there will be certain things that stand out and grab your attention. Different things. Comparing candidates based on different criteria is like comparing apples and oranges.
Computers have no problem doing the same thing consistently.
5. Finding multiple pathways to success
The best hiring algorithms do something that human intuition cannot.
They can consider all the potential pathways to success. At the same time.
When you look at a resume or conduct an interview, there are probably a few specific things that you might be looking for. These might be the best things to look for some of the time, but are they always?
You probably have one or two theories about the type of person you want to hire, but a computer can test among thousands of possible combinations that contribute to important outcomes. Why limit your options?
There are often a variety of ways that a person can be successful. Limiting yourself to a specific “type” of person unnecessarily reduces your chances of making the best hiring decisions.
Image credit: Chris Isherwood