35◼ STRATEGIES Bloomberg Businessweek March 1, 2021
at an innovative organization that wants people to be
creative, you don’t necessarily want people who have
done the same job over and over.
● How about screening for GPA?
Grades and prior experience are not necessarily
great predictors of success.
● Aren’t algorithms biased?
Algorithms definitely can be biased, but humans
can be quite biased in lots of ways that are harder
to fix. So when evaluating an algorithm, I evaluate
it relative to the process that firms might be using,
not relative to some perfect benchmark.
● What’s special about your methodology?
Often, algorithms are used as predictors: Based
on who was hired in the past, this is who will be
good in the future. It’s better to think about hiring
as dynamic learning.● What’s dynamic learning?
The algorithm learns by taking selected risks
and making choices it hasn’t made before, like
[earmarking] people with different qualifications
or from different demographics, because that’s
where you have the greatest opportunities to learn
whether those are good choices going forward. It
turns out that when you do this, you improve the
quality of people that you hire and increase gen-
der and racial diversity by a substantial amount.
● How do you define “quality”?
People who are likely to receive and accept an
offer. You want algorithms that are trained on job
outcomes you care about, like performance rat-
ings, or retention, or absences, or promotions.
You want to prioritize those.MAKE
BETTER
Q&AHIRES
Danielle Li, an economist and associate
professor at MIT’s Sloan School of
Management, studies how technology
can help companies hire better. One
key is to broaden interview pools using
algorithms that favor nontraditional job
applicants who might get screened
out by off-the-shelf software and busy
hiring managers. In a recent study Li
conducted with a Fortune 500 com-
pany, she found that her approach
increased the percentage of Black
or Hispanic candidates interviewed
to 23%, up from the 3% to 5% gen-
erated by traditional algorithms and
the 10% the human recruiters consid-
ered. Here are edited excerpts from
her interview with Arianne Cohen:
● Why do companies need to improve these
pools?
Companies often stick to known quantities:
people from elite schools and people who are sim-
ilar to those who succeeded at that firm in the past.
● Why?
Risk aversion. Human screeners lean on variables
like whether someone filled out the form properly. But
those variables aren’t predictive—there are plenty of
great, hard workers who didn’t fill out the form properly.
● Are you saying it’s not essential to look at
experience or attention to detail?
The point is to find someone who will be creative
and seek solutions for this particular position. If you’re
“A lot of our
work has
become
quite
invisible”The need: Growth—learning new skills and
acquiring knowledge.
The problem: If workers don’t grow professionally,
their motivation and engagement wane.
The fix: Ask employees to set quarterly learning
goals and discuss how to meet them. This might
mean pursuing online certificates or degrees, getting
access to fly-on-the-wall sessions with executives,
or having time to pursue a pet project.
The part managers mess up: Don’t make
these opportunities feel like items on a to-do list or
a part of meeting the bottom line. �Arianne Cohen