AI is the new new thing in recruiting. Products like Mya claim to use AI and promise to revolutionize recruiting. As usual, the hype far exceeds the reality, since true AI means having a flexible, general-purpose intelligence of the type which allows an individual to learn to complete a vast range of tasks. Today we have products that have mastered individual tasks, like IBM’s Deep Blue that plays chess better than even the best humans, and Watson, that beat the champions on Jeopardy. But such programs are good at just one particular task, and none has the broad intelligence that characterizes humans. No combination of hardware and software available today can learn new things the way even a small child can.
But we don’t need AI to improve recruiting. Software can certainly be made to convince people into believing it is a recruiter. In June 2014 a milestone was reached in computing when a program — “Eugene” — passed the Turing Test for the first time. The test requires a program to convince a sufficient number of people that it is a human by engaging in a textual conversation. Eugene did just that. What was most amazing was not that the test was passed, but that Eugene is just a relatively small amount of code. It wasn’t a gigantic software product, running on a supercomputer, like IBM’s Watson. Eugene succeeded by using a series of ploys that fooled the judges. It does that mainly by changing the subject or responding in generalities when asked a question it can’t answer. It should run for President.
People want to believe they are talking to a human, even if they know they are talking to a machine. Amazon’s engineers have learned that a lot of people talk to Alexa — the software included with Amazon’s smart-home speaker — as if it were a person. Commercially available robots in Japan like Softbank’s Pepper and Toyota’s Kirobo can entertain children, chat with adults and identify human emotions. Jibo, described as a “social robot,” communicates and talks using natural social and emotive cues.
The technology in these products can add a lot of value to recruiting. The greatest impact can be in engaging with candidates. As most recruiters know, engagement is a time-consuming, and labor-intensive activity. There’s no shortage of products that claim to help with candidate engagement, but the yields can be low since they all rely on text-based content and are minimally interactive. They lack a social interface. People interact best with each other in verbal conversations. Text is a poor substitute. The difference is that verbal conversations include a lot of irrelevant banter and quirks reflecting the personality of the people involved. This is what’s missing in Siri. While it sounds human, no one mistakes it for one — the language is too polished, too precise, and too even in tone — utterly lacking in any personality.
But social interfaces are changing rapidly. A report in the Wall Street Journal mentions that Amazon is giving Alexa a personality, by making its voice sound more natural, and writing clever or funny answers to common questions. Jibo punctuates its conversation with stupid jokes on occasion.
The software can be adapted for use on career sites, allowing an employer to better engage candidates. A short conversation between the candidate and a career site is far better than trying to serve up content customized to the candidate. A common complaint of candidates is that they generally have no way to ask questions about a job or the employer, and while the information may be available they have to find it. A social interface can eliminate all that and raise a candidate’s interest by truly engaging her in a conversation at any time on a wide range of subjects.
Job interviews are another aspect of recruiting that can benefit from social computing. Video and audio interviewing products have been around for years, and while they are a convenience, the limitations of the technology restricts the value they provide. The main limitations are that candidates cannot ask clarifying questions, and follow-up questions cannot be asked by a recruiter. Watching the recordings, necessary for scoring candidates, can be a violation of the constitution’s prohibition on cruel and unusual punishment.
But an interview administered through a social interface can be interactive — a true conversation. Questions can be adapted based on responses, and candidates can ask their own questions. The software can even parse out key phrases in a response and do some rudimentary scoring. Additional value from using such systems is that they are not influenced by unconscious bias that may come in when a candidate has a strong accent or is not a native speaker of the language of the interview.
The technologies mentioned here are not about to put recruiters out of work. We are a very long way from having true AI, and we may never get there. That is any empty threat made many times before. Job boards were supposed to become a clearinghouse between candidates and jobs, and social networks were supposed to make sourcing irrelevant. We’re still here, and there’s more of us than ever before, but the technologies will change recruiting, hopefully for the better. The benefits never come without costs and unintended consequences. Overestimating the assessment capabilities of the software will result in the wrong candidates being screened out or in. To support engagement will require creating content, which can also be time consuming. The work done by recruiters will be different, but a revolution is not coming. Watson cannot become a recruiter.