Just How Friendly and Unbiased can a Job Interviewing Bot Really Be?

Over the last two years the use of artificial intelligence (AI) technology has been radically changing how companies source, screen and manage job applicants in the talent pipeline. In my research, I refer to it as Algorithm-Based-Hiring (ABH). Some companies are taking AI further and starting to use bots (robots) to conduct job interviews.  An article about the job interview robot (bot), Tengai, confirmed that if development stays on schedule, job interviews with live human subjects applying for real jobs, will happen in 2020.  

The frequently stated goal of recruiting bots like Tengai, is to conduct interviews free from the unconscious biases that recruiters could bring into the hiring process.  These bots want to remove bias, while still making the experience seem human.  Two other benefits of this technology often cited are: First, bots will stay consistent and logical despite repetition, and second, the massive amounts of data enabling AI decision making, will drive selection of better quality candidates without unintentional discrimination.  

As these AI tools become more pervasive, there needs to be new conversations about how humans will handle these face-to-bot interactions in interviews.

The goal to reduce bias in hiring is a noble undertaking and while developers of bots progress towards that end, we need to keep in mind the other side of the interviewing equation - the job applicants.  You know - those people who have been conditioned to think that first impressions actually do matter in a job interview. Since bots do not engage in pre-interview chatter or do the warm-and-fuzzies, the whole notion of first impression may no longer even be a factor when assessing competing candidates.  So, while the bots’ lack of awareness of demographic variables like gender, age, ethnicity and race may be beneficial, it is unclear just how other applicant attributes that have been predictable indicators in job interviews of the past, will be assessed.  

Characteristics like eye contact, speaking ability, interpersonal skills and even appropriate interview attire that formerly impacted candidate choice in the interview process could have little to no value with bots.  How will job seekers now demonstrate traits like likeability or even the maturity of a handshake when interviewing bots can’t make small talk?  

It is very easy to see the value of a recruiting bot that will have the stamina to ask question after question precisely and without fatigue in interview after interview. However, how tolerant will job applicants be for answering question after question from a bot with no visual cues? Given the widely accepted paradigm that 55% of our communication is nonverbal, how, for example, will the nervous behavior of an applicant be evaluated by bots?  Emotionally intelligent job seekers routinely rely on body language cues to get a sense of how they are performing in an interview.  What happens to an applicant without feedback?

Without face-to-face time, recruiters could be left trying to assess applicant qualifications primarily from bot transcripts.  If the bot only delivers an interview transcript to a recruiter, what, beyond the words, will the recruiter assess to evaluate the applicant? How will those words be evaluated? How does this new emphasis on the interpretation of a written transcript impact the applicant’s candidacy or improve the recruiter’s ability to evaluate applicants?  Recruiters could see replacing face-to-face conversations with hours of transcript reading, as less than progressive.

The bottom line is that just because an interviewing bot isn't processing demographic information does not mean there is no guarantee that unconscious bias, or other selection biases will not be introduced.  Also, touting the reduction of unconscious bias doesn’t mean that job applicants will trust the interview process because they are sitting across from a bot instead of a human. An erosion of trust in the selection process or the company’s brand could arise, for example, if job seekers have concerns about where their interview transcripts might end up through data security breaches.   

The hope as we keep expanding the use of AI in recruiting is that we periodically take the pulse, yes-the pulse, of the humans across from the bots.  Although humans may be comfortable asking Alexa’s opinion on food, lodging and entertainment; they may not be too keen on interview bot screening them in or out of job application processes.  

It is important that those humans developing the AI tools understand more about human communication, reactions and tolerances for interactions with these machines in job interviews.  Hopefully, the analysts and developers will involve subject matter experts on the job applicant side of the equation and will ask the right questions. The more we understand human interactions, the better tools we will build.  So, maybe this push for AI will help us better understand what it means to be human in an interview, even if we can’t claim to totally eradicate bias.