O
On Topic | Catalyst
Kim Nilsson, CEO of Pivigo, which matches
data scientists to businesses, believes using AI for
screening CVs, based on crude metrics or keywords,
can close off a huge potential talent pool.
“Recruiters are missing out on people who could
be fantastic members of their team but don’t fit the
mould,” she says. “When I left academia, I wanted to
move into project management but my experience
didn’t match the keyword searches. To be fair, that’s
not just a machine thing, humans do this too,”
she adds.
Dr Terence Tse, associate professor at ESCP,
shares her concern. “If everyone is looking for the
same keywords, and all the companies are looking for
the same candidates with these keywords, anything
non-standard is not going to show up,” he says. “We
risk profiles starting to standardise. If we want
creativity in the future what is far more important
than academic credentials is cognitive diversity.”
In other words, just as, in the early days of the
internet, canny porn websites learned to ‘game’
Google’s algorithms by including terms such as
‘cars’ and ‘EastEnders’ in their metadata to rise up
the ranks, we might end up seeing applicants swiftly
learning how to game the CV screeners with perfectly
pitched robot-pleasing phraseology.
Additionally, “there is a danger of AI robbing
humans of the chance to develop the experience they
need to progress in the recruitment field,” according
to Andrew Wayland, chief technology officer at
Alexander Mann Solutions.
“Recruitment requirements often relate to subtle
contexts that are not easily measured in the data
available to the AI, but are obvious to a human,” he
says. “In the past, human recruiters were immersed
in the workplace demands and culture, and saw
the impact of good hiring decisions, building their
own ‘neural net’ for success. Experienced recruiters
helped new staff learn more quickly and informal
feedback loops were established. Complexity such
as competitor activity, new business requirements
and social context could easily be layered on top by
a skilled human operator.”
He argues the need to build in opportunities to
review and learn, both in AI models and people,
as they work together. “Humans can be incredibly
adaptable and how we interact with our robot
sidekicks will be unpredictable,” he says.
“My bet is a new breed of recruiter will emerge
who can exploit and perfect AI tools but also
understand the limitations. We developed our bot,
RHYS, to handle initial interactions with candidates,
review pools and recommend jobs. This freed up the
recruiters to spend more time with the most relevant
candidates. But we still added feedback loops for
alexandermannsolutions.com
36
What does AI in
recruitment encompass?
• Screening: natural language analysis
looks out for keywords in CVs, LinkedIn
profiles and online applications. This can
be augmented through analysis of social
media usage or other public data sources.
• Video analysis: algorithms pick up
keywords but also facial expressions
and tone of voice from a video interview.
These attributes can be matched against
the attributes of successful employees.
• C
andidate communication: AI-driven
chatbots can answer frequently asked
questions, send out alerts at different
stages of the recruitment process,
and even support onboarding.
• A
ssessments: AI can manage stages of
the recruitment process where candidates
partake in games or other cognitive
assessments, and merge this data
with other sources to suggest whether
someone is a good fit.
• Measuring new hire performance: AI
can analyse data after someone has been
hired and use data from good performers
to support hiring mechanisms.
• A
nonymising applications: stripping
out data points that could identify
characteristics such as race, gender or
sexual orientation means human social
bias is reduced.