THANK YOU FOR SUBSCRIBING
Hrtech Outlook Weekly Brief
Be first to read the latest tech news, Industry Leader's Insights, and CIO interviews of medium and large enterprises exclusively from Hrtech Outlook
No-one is going to question that companies that support diversity, equity and inclusion perform better. A 2019 report carried out by McKinsey identified that companies in the upper 25% for gender, racial and ethnic diversity are “more likely to have financial returns above their national industry medians”. In my opinion there must be a great opportunity for recruitment technology to remove bias and drive diversity.
Almost all recruitment tech is focused on helping recruiters to streamline processes, enable shorter cycle times; be slicker, faster and more efficient. The technology aims to winnow down/cull the applicant pool, so that line managers are presented with a small, targeted shortlist of relevant candidates to meet face-to-face.
The traditional paradigm of sending in CV’s or filling in application forms seems to be the primary first step. And the first intervention of technology to start to reduce the numbers, seems to be key word searches. So recruiters list the key words they want to see on CV’s and application forms and the AI tech searches for the key words in the documents. Does it drive diversity? It seems to me that the people who can write a great cover letter or have had access to coaching will have an advantage. And just because you have a lot of key words on your application form/CV doesn’t mean that you are right for the role?
There have long been debates about removing the CV from the application process. Not sure that any company has quite done this yet, and as an executive recruiter (who primarily still does search in the old-fashioned way), knowing where someone has worked is important. But for those large recruitment campaigns where there is a big applicant pool and relevant capability and expertise can be drawn from a number of different sectors/experiences, should there still be so much reliance on a CV?
We are all aware that people with “foreign-sounding” names are less likely to be invited to interview. To respond to this, there is technology that scrubs names, age, and references to gender and clues about ethnicity from CV’s and application forms. This technology levels the playing field. It would be good to get first-hand experience of how this has been used at scale and whether it has resulted in more diverse hires. I would think it adds lots of value.
Quite a lot of the recruitment tech out there, rather than making the recruitment process more streamlined, seems to add in steps. I guess if these steps can have an overall beneficial effect on the removing bias and levelling the field, and can be delivered in a slick, fast and user-friendly way, then the additional step won’t be too much of a problem. But if there are too many steps and stages in a recruitment process, there is a much higher likelihood of candidate drop-out. Which candidates will be more likely to drop out? Does the addition of extra steps drive diversity?
What do I mean by additional steps? I am talking about things like ability tests, assessments etc. Technology that will test candidates’ numerical, verbal, abstract thinking or personality profiling tools. These tools have been around in one form or another for a long time. How leading edge are they? Do they drive diversity? There have been loads of papers written that suggest that ability tests are biased.
Slightly more technologically advanced, is the introduction of assessment “games” into a recruitment process. These are probably focused on abstract thought and potentially inherent intelligence (says who?). They are games that look and feel a bit like a computer game with flashing lights and noises and count-down clocks etc. I have a lot of nervousness around bias against those with diversity of thought. This was raised in a BBC3 programme (“computer says no”) where a young applicant with ADHA had major issues with the format. I also have concerns around gender bias (more males play computer games) and age (I haven’t progressed beyond Space Invaders!)
However, in my research, I see outputs of tests/assessments incorporated into recruitment technology tools, that provide a score on “trustworthiness” to “dependability” to “likeability” (says who?) Where are these scores on personality traits coming from?
Maybe some of these scores come from the use of facial recognition AI?
Before I discuss facial recognition AI, I must state I am not in favour of reading too much into faces or body language in any form of interview. That is because MOST people are generally a bit nervous or self-conscious in an interview. If someone is looking down and to the left, that is probably NOT because they are lying, it is more likely that they are just a bit nervous, or they might have some dust in their eye, or the lights are too bright, or they might have some form of issue that makes them blink and look down!
Facial recognition AI was developed by law enforcement, not for recruitment, and the AI behind it is so complex that it is not auditable. There are several case studies that show that facial recognition AI has “problems” reading non-white faces. And there was an article in the Telegraph recently (“why you shouldn’t wear glasses to an interview with a robot”) where software engineers from Cambridge University suggested that some facial recognition technology represents little more “automated pseudoscientific software” making “spurious correlations” between facial expressions/visuals and personality.
I do appreciate that technology is moving forward at warp speed. If developers can build an algorithm that can beat a chess grand-master, they must be able to build some kind of reliable assessment. But how do we know whether it is biased?
Legislation coming that will safeguard against potential bias driven by the use of AI in recruitment. The days of “dark algorithms” in recruitment are numbered. Let’s not forget when Amazon developed an algorithm to recruit engineers, they discovered that it was biased in favour of male candidates. Amazon’s (now withdrawn) recruitment tool was built using data given to it mainly by males and the algorithm then “learned” and skewed results towards men.
There are a number of vendors out there marketing technology that enables companies to “attract, retain and develop diverse talent”. There are phrases like “deep learning to match talent with roles”, and “delivering total talent management across an organization that is non biased”. There is AI that can (apparently) predict what candidates are capable of as opposed to looking at what they have done. There are tech companies out there with AI that can (apparently) help you understand if people will “thrive in your environment”. What is the technology exactly? How do the candidates interact with it? How does all this AI work?
AI has to be part of the solution in the quest to streamline recruitment processes and deliver diversity. But it must be explainable. If you can’t understand why the AI has delivered the results it has, how can you be sure that the algorithm is not biased? It is crucial that AI can be audited and everyone (even me) understands what is going on.
In my opinion, before integrating AI technology into their processes, recruiters must understand:
How the technology works?
What it claims to do?
Can it deliver on those claims?
Does it drive diversity?
