Is Automating the Hiring Process a Mistake?
Posted on January 12, 2022 by P.K. Maric
Automation is increasing in many industries, and even the hiring process itself. However, automated hiring software often fails spectacularly.
From pre-employment tests to applicant tracking systems, or application forms, hiring depends on various software, with varying degrees of automation. When it works well, it saves time and makes things far more convenient, for both candidates and employers. But it doesn’t always work well. In some cases, it fails more often than it succeeds, but this failure often goes completely undetected.
When Automated Hiring Goes Wrong
Big companies have many applicants for every job. Google, for example, receives millions of job applications every year. Each corporate job opening receives 250 applications, on average. There is no way to screen all these applicants in a reasonable amount of time without relying on some sort of automated process.
Traditionally the first step in the hiring process is resume screening, as the first thing candidates generally do when applying for a job is to send in their resume. So naturally, there have been attempts to automate this part of the hiring process.
Unfortunately, this fails more often than it succeeds, because automated hiring software mistakenly rejects millions of qualified job candidates. With the rise of various automated hiring tools, the situation has gotten so bad that, according to a report from Harvard Business Review, automated hiring tools have “exacerbated the very talent shortage they were intended to address.”
Part of the problem is faulty software, as it would filter out candidates if they were missing only one of the required skills, even if they perfectly matched all other requirements, or even if they used a slightly different terminology than that used by the employer. But the bigger issue is simply that resumes are utterly useless and should not be used as the main criteria for hiring.
Research has shown that most information on a resume has almost no validity in predicting job performance. Things candidates typically put on resumes, such as their age or interests, don’t help you make a hiring decision. But surprisingly even things like previous job experience and education level don’t have any correlation with job performance.
Resumes simply aren’t good criteria for making a hiring decision. Regardless of whether that decision is being made by a person or an AI. Automation can’t solve this problem because automating a bad process that shouldn’t exist to begin with doesn’t make it better.
Another problem is that automated hiring software can also be biased. Amazon developed an AI recruiting tool and then scrapped it when they realized that it discriminated against women. We tend to think of AI as unbiased and objective, so how could this happen?
Well, AI is built by humans, and will therefore have any error that humans make while building it. The problem here was the data it was trained on. People are biased, so the AI was fed biased data and learned to discriminate against women because that’s what the people that built it told it to do, albeit unintentionally.
Based on the available data, Amazon’s system learned that men were preferable to women because most of the resumes that were used as an example of a good candidate were from men. The system penalized resumes that contained the word “women’s”, for example in the phrase “women’s chess club”, as well as penalizing all-women’s colleges.
Luckily, this AI recruiting tool was never actually put into use (or so Amazon claims), but it shows that automating a process doesn’t immediately make it better.
The Problem
The problem with automated hiring software is that they rely on overly simplistic criteria to separate “good” from “bad” candidates. They do this by filtering applications based on certain keywords, skills, years of experience, education levels, and similar criteria.
The core of the issue is that the traditional hiring process itself is deeply flawed. Automating a flawed process doesn’t fix it, it just makes problems happen faster.
There has been a lot of research into the effectiveness of different hiring methods. The problem with the classic hiring process is that the most common methods used to select candidates, such as resume screening and unstructured interviews, have a very low correlation with job performance. The methods that are accurate and effective in selecting candidates, such as skill tests, aptitude tests, or structured interviews, typically come too late in the process for it to be effective, if they are even used in the first place.
Automated hiring can’t help if it’s only sorting resumes. But to fix the problem we first need to fix the hiring process by using more effective methods and then find a way to automate them. Fortunately, we already know which methods to use, and how to automate them.
When Automation Works for Hiring
Many employers use various types of tests in their hiring process. Such as:
- Personality tests
- Aptitude tests
- Skill tests
- Knowledge tests
These tests can be automated without any issues, and speed up the process significantly while avoiding human error, which also makes things far more accurate.
While the validity of personality tests for hiring purposes is debatable at best, work-sample tests (which target job skills and knowledge), and cognitive ability tests (also referred to as general mental ability tests, aptitude tests, or reasoning tests), are some of the most accurate and effective hiring methods available.
Pre-employment testing software focuses on automating this part of the hiring process. The way testing platforms work is that candidates take a skill test, set by you, the employer, and the software automatically scores it. The main automated part is the scoring of answers, but the software also makes it easier for employers to administer tests, and easier for candidates to take them as well. Since the tests are online, candidates can take them from wherever they prefer, at a time that suits them.
There are a number of pre-employment testing platforms available online, including TestDome. You may have also heard of HackerRank, which is probably the most famous one. The most important thing isn’t the platform though, it’s how you test.
What automated scoring does for this process reduces human error significantly, as well as speeds things up so that you can immediately see results, instead of having to evaluate them yourself. This saves time for you, but also for the candidate since you can get back to them quickly.
While many employers do use skill or aptitude tests in their hiring process, some still prefer the old-fashioned and error-prone approach of manual scoring.
I’ve experienced employers using pen-and-paper tests in interviews. Even for programming questions. The whole time I wondered: “Why can’t I just do this on a computer?”. Even in school computer science tests were done on an actual computer. Why are employers, especially those with a bigger budget than most schools, relying on pen-and-paper or whiteboard tests and manually grading results? This is the easiest part of the hiring process to automate, and one that’s least prone to errors.
Conclusion
The traditional hiring process is backward, so attempts to automate it, starting from resumes and applications, have backfired completely. Worst of all, many companies aren’t even aware that their process isn’t working, and are missing out on millions of potentially great candidates. Automated tools that focus on applications and resumes have shown to be just as ineffective as humans, if not more so, considering they make errors far more quickly and on a greater scale.
To fix the problem, we first need to fix the hiring process by getting rid of the parts that don’t work, and then automating the parts that do.
