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.
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
Jibo, Tell Me About This Job!
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.
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.
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.
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.
Alexa, Interview These Candidates!
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.
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 Future of Recruiting
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