The Problem with DEI
Lack of data literacy can turn a good idea bad.
Background
Most DEI (Diversity, Equity & Inclusion) programmes are failures. Recent research also shows that DEI training simply does not work. One of the highest attrition rates is for the role of DEI head. The bad news just keeps coming. Even the Harvard Business Review (HBR) published an article on this phenomenon as recently as April 2024.
DEI has been in play since the 1960’s as a direct response to the civic and social justice movement. While the intent has been good, its entry into the professional corporate domain has been burdened with a lack of clarity, a lack of budget, and a lack of talent. The domain today is puffed up with buzz words like ‘authenticity’ and ‘collaboration’, and ’safe space’, etc. No one really understands what is being said, and no one really trusts what is being communicated. It is also not uncommon for the workforce to view DEI as a zero-sum game where those who were previously seen as privileged are having their ‘advantages’ taken away to be given to those who were under-privileged; a holdover from its social justice roots. It’s being increasingly perceived as ‘reverse discrimination’ or ‘reverse racism’.
What is almost never talked about is the data side of DEI. How do we know if something is not diverse? You can’t expect to change what you don’t understand. And so I’m dedicating my 38th weekly article to calling out some of the bad thinking I see in the DEI movement.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
Kill the Buzz (Words)
I quote from the aforesaid HBR April 2024 article: “Diversity efforts build workforces that reflect the communities they serve by giving everyone a fair chance to enter and rise through each level of our organisations. Equity efforts design organisational systems and processes that prevent discrimination and equip everyone with the resources they need to succeed. Inclusion efforts create working environments where everyone is treated respectfully and is valued for their unique contributions and backgrounds.” Sounds great, but how do you measure “fair chance”? How do you know what is “unique contribution” and whether they are “valued” for it?
And this is precisely the issue with DEI — it lacks practical definitions. Terms like ’diversity’, ‘equity’, and ‘inclusion’ are what we call latent variables in data analytics. Latest variables are either hypothetical constructs or they are unobservable, and we need to identify or create proxies to be able to measure them. A classic latent variable is ‘satisfaction’ and we proxy-measure it in part through the number of compliments received. So, DEI needs to be better defined in data-specific terms to enable us to properly measure its progress objectively.
Too often, we default to convenient shorthands on diversity by narrowly focusing on ethnicity, gender and sexual orientation. But for countries that are more culturally homogeneous, like Japan or Korea, we can do better in the way we define and support diversity in the workplace. This narrow focus unfortunately plays into the narrative of social injustice, which distracts from the broader objectives.
Data Savvy
The other main issue is that DEI officers are rarely data savvy. In my data analytics consulting work, I have almost never come across an interaction with a DEI officer looking to use data to deep-dive into their work. With AI the headline of the day, I am seeing DEI officers scrambling to understand if this new technology breakthrough will negatively affect their DEI agenda. They are worried that AI will scale and reinforce the existing bias in workforce decisioning. DEI officers are compelled to be data sensitive or fade into irrelevance.
The objective of DEI is about correcting for negative imbalances by either neutralising it or creating positive imbalances. We cannot simply rely on our observational senses to ascertain those negative imbalances. We can’t be chasing ‘red herrings’. DEI is therefore naturally aligned with data analytics. In fact, it requires a high level of data fluency (not just literacy) to be successful in the job. Consider if you are the DEI officer in the civil service. Is the objective of DEI to reflect the same ethnic ratio as the population mix within the workforce (i.e. equality) or is it to over-weight in favour of the disadvantaged minorities (i.e. equity)? And if we over-weight to correct for inequities and non-inclusions, how much do we over-weight to be ‘fair’? Without structured data-oriented thinking, DEI simply devolves into experimental policies and programmes without any clear understanding of the causes and effects.
I fired off a prompt into Google’s Gemini (Gen AI solution) and got back some useful answers in terms of how to measure progress on DEI. One such recommendation was to see if the ‘diversity ratio’ of employees remain constant throughout the seniority pipeline — e.g. if we are hiring 50% women into the workforce as fresh graduates, do we continue to have 50% women in middle management and senior management? But if you are not data savvy, simply relying on such stock approaches can lead to all kinds of misinterpretations. Men and women are drawn unequally to different academic disciplines. And once hired, those unequal disciplines have different career growth opportunities in any given organisation due to both corporate-macro and organisation designs. So men and women will rise at unequal rates and achieve unequal seniorities. Unless you are a statistician or work with one, you won’t know how to factor in these considerations in your assessment of whether ‘diversity ratio’ is maintained. Just because you observe fewer women in senior positions doesn’t mean there is gender discrimination going on. Our brains are not tuned for conditional probabilities.
Conclusion
The whole DEI movement has been poorly conceived. It claims to be evidence-based, but it is anything but. The lack of data-anchored definitions leaves DEI officers grabbing at straws. The lack of data savvy prevents DEI officers from shifting the needle in the right direction. Instead, the DEI efforts results in distrust and dissent.
During my time as the country CAO in a large internal bank, I ran one of the most diverse teams in the organisation. We were labelled as a mini ‘United Nations’. I had at least 7 different nationalities in my relatively small team. I wanted a multi-national team to prevent the occurrence of small pocket-groups, so that everyone had to learn to trust each other and work together. And to ensure that we had a broad perspective to problem-solving, I hired them from a variety of backgrounds. It was engineered diversity. Was I supporting DEI? I was definitely thinking about how diversity would make for a more cohesive and capable data analytics function, and I had certain notions of how to achieve that. But I certainly wasn’t thinking about how to objectively measure the outcomes. But then, I wasn’t the DEI officer.
