PRE-READING QUESTIONS
Think about the following questions.
1. What experiences have you had with interviews? How did you feel about them?
2. How do you think these would have been different if you had been interviewed by a machine?
3. Would you be happy if your next interview were to be done by artificial intelligence? Why or why not?

AI INTERVIEWS

Today, many of the job advertisements and job specifications we see have been written by computer programs known as artificial intelligence (AI) bots. Having found a job we think desirable, we send in our applications and résumés. These are often read by AI too; bots search for key words and details to filter out¹ weaker candidates. Until recently, this was where humans would take over. Increasingly, though, AI is being used at the interview stage too, so applicants² may find themselves interviewed online in their own homes, perhaps even by a virtual head on a screen.

AI systems ask applicants questions, record their answers, and assess them against the job criteria. These bots also analyze³ the way applicants speak and behave by measuring tone, listening to speech patterns, checking for grammatical errors, and assessing eye movements. The systems then produce a report with a series of scores for employers. For now, most employers only use this report to produce a list of candidates to interview in person.

Primarily, employers like using AI because it saves a considerable amount of time. In a competitive job market, companies often find themselves overwhelmed with applications. This can leave several members of staff tied up for weeks in the search for just one new employee. Using AI, companies can "interview" as many as 20,000 applicants in one day. This is especially useful for entry-level jobs where there are many applicants, almost all of whom are inexperienced. Instead of asking detailed questions about that particular industry, AI can quickly evaluate⁴ applicants based on their enthusiasm and general manner. In fact, HireVue-a company that provides AI interview technologies—claims to interview up to one million college students each year.

There are advantages for applicants, too. As well as allowing applicants to remain in the comfort of their own home, most AI interviews allow candidates to complete practice questions or allow thirty seconds to check notes and prepare an answer, and many even allow candidates to record their answers again. None of this would be acceptable in a traditional interview setting. And, while they lack human understanding, AI bots also lack other human traits that can lead to unfair advantages in interview situations. They do not get tired or grumpy⁵ during a long day of interviews, nor do they take a liking to certain candidates based on abstract opinions.

Likewise, there is no chance of a personality clash⁶-AI bots remain completely neutral toward candidates. Thus, 20,000 applicants receive exactly the same interview and in theory, at least-have an equal chance of success.

However, there are also several disadvantages for those seeking work. With AI saving so much time, companies have no reason to limit the number of interviews they grant. So applicants may be given interviews despite having very little chance of getting the job. Even though AI interviews take place at home, applicants still spend time researching the job, preparing their clothes and homes, and completing the interview, which may just be wasted time.

On top of this, the criteria AI interviews use are not standardized and, without transparency, it can be hard to know exactly how one is being assessed. In truth, AI interviews will never be entirely objective as they are programmed by humans to look for certain "warning signs." When these are found, though, AI bots may not be flexible enough to interpret⁷ their real meanings. For example, a long gap in someone's employment history may reduce an applicant's score significantly. But this may have been the result of illness or time taken out to raise children. Similarly, AI bots that assess candidates' speech patterns, expressions, eye movements, and body language may not be able to differentiate between "unsuitable" candidates and those with disabilities.

In such situations, rather than being fair, some AI systems may lead to discrimination. This is not only a problem for applicants. Even though they have given AI bots the job of interviewing candidates, companies remain responsible for everything that happens in interviews. With time, we may even see companies sued⁸ over AI's actions. For many reasons, companies must be very sure that the Al they use reflects⁹ their company's standards, attitudes, and beliefs. Making AI interviews as transparent as possible may be in everyone's interest as an interview is not only about a business choosing an employee but about both finding a good and lasting match.

• disability 
• neutral 
• specification 
• transparency 
• grammatical
• differentiate

AI Interviews: Business and Practice

One company leading the way in AI interviewing is HireVue. Founded in 2004, in Salt Lake City, Utah, the company provides a complete AI assessment service for companies looking to hire new staff. As of 2021, over 700 companies use HireVue, and the company has interviewed over twelve million applicants on their behalf.

A major corporation now using HireVue systems is Unilever, the parent company of familiar food brands like Hellman's, Knorr, and Lipton tea. Unilever has a combined staff of more than 155,000 worldwide, and Leena Nair, its head of human resources, has said that using HireVue saves the company more than 100,000 hours and $1 million every year.

But not everyone is convinced that HireVue's model is fair or effective. A prominent AI researcher in the US has called HireVue's system "pseudoscience," and a campaigning group, the Electronic Privacy Information Center, even sent an official complaint to the Federal Trade Commission. They claimed that the way HireVue's AI systems worked was "unfair, deceptive, biased, and unprovable."

In response, Loren Larsen, head of technology at HireVue, has said that during a typical thirty- minute interview, the company's AI system collects 500,000 pieces of data about an applicant-far more than any human might manage. Larsen was less clear about exactly what data the system records or how it interprets this information.

For applicants, then, there remains the question of how to succeed when taking an AI interview. The internet is full of tips and suggestions, which include sitting in front of a blank background, talking more slowly than normal, smiling often, looking directly at one's webcam, and repeating words that appear in the job specification. However, as AI interviews do not provide feedback for applicants, for now, these ideas are mostly guesswork.