In discussions with senior leaders and boards, it becomes very clear that AI is the new hammer seeking a nail. “Why can’t we do this with AI?” is a refrain from people keen to get on the bandwagon but not clear what they are really asking for.
Bias in hiring is one such problem many think are amenable to fixing with AI.
Except for this Amazon experience where the algorithm, trained on past data about mostly male employees, actively deprecated women candidates’ CVs such as by penalising the use of words such as “women’s”, as in “women’s chess club captain”, and downgrading graduates of all-women’s colleges, which shows AI not quite solve that one.
It should go without saying AI will not fix what is embedded in an organisation’s thinking and processes.
Only fixing those processes will.
And no, just anonymising the CVs does not help.
A stellar example of this first-principles thinking can be seen in the British civil service hiring.
Yes, you read that right.
It is not the depths of the tech industry that is thinking innovatively about bias but the rank and file of the civil service with which many are familiar probably only thanks to Yes Minister.
Instructions are clear for an individual from outside the civil service applying for a role. When submitting your CV, you — the candidate — are required to remove any references to your name, gender, university name and graduation date.
Sounds simple? And baffling?
It is.
Solutions can be simple and genius at the same time.
Name bias against names that sound Hispanic, African American, non-white is well-known and backed by research. Removing the name is a simple slam dunk.
Anonymising has been found to benefit women but not immigrants. And yet gender bias is real. Removing mentions to gender can generalise achievement without linking it, say, to women’s chess club captaincy.
Removing the University name is a fascinating one. That the civil service is all homogeneous is widely believed and that it is all Oxbridge is an idea that the civil service is serious about challenging. While someone like me has to remove her Cambridge reference, someone else who attended a former polytechnic is no longer at a disadvantage compared to someone like me. This levels the playing field a bit.
Finally, there is the removal of the Graduation date. Age bias cuts both ways, against those who think they are too young and those that are treated as being too old (though the language of the potential employers gets especially creative when dealing with the more experienced person).
In the last two decades, I have heard some fun comments about my age and gender and possibly my tongue twister of a name. However when I applied to a senior role, the civil service called me in for an interview. I won’t say more about why I applied and what the role was, but the fact remains I got through the process, was invited to the interview, and was made an offer after having a rich discussion with the interview panel.
One of my friends, Dowshan Humzah, makes the important point that blind CV processing like this still disadvantages people when they turn up for an interview.
It’s a fair point.
But getting to the table is how the current “normal” will get used to “diversity” whose CVs and experience have passed the blind checks.
We have so much to fix in this uneven hiring environment that every step that brings people of all races, ages, genders, sexuality, disability status and neuro-diversity sitting across the table from decision-makers is valuable.
As a leader, if you are worried about hiring bias, don’t seek solutions in technology.
Ditch the thought of AI as the panacea to the hiring problem.
Use first principles thinking to fix your processes first.