Introduction

The tech industry is far from being as diverse as it should be. Women make up around 23-25% of the US “tech workforce” in major tech companies¹. These numbers are well below the total participation rate of women in these companies (29-45%). The situation is even worse for the field of data science, where “of all the various tech fields, data science currently ranks the lowest in diversity”². It’s not difficult to imagine that things are not much better for data science teams in large enterprises such as banks, telcos and retailers. With all the caveats associated with online surveys, the data suggests that only about 20% of data scientists across industries are women³.

If organizations reduce diversity to a matter of meeting quotas, they fail to understand that diverse teams are essential to solving complex challenges like AI.

These numbers paint a dire picture of the AI industry without even considering other kinds of diversity. In addition to the important question of equity, diverse teams are also more innovative and adapt more easily to a rapidly changing environment⁴. A quota-focused approach that disregards the inherent value of teams with varied backgrounds fails individuals and organizations alike. In the following, I will look at diversity through three distinct lenses, aiming to unveil a more nuanced perspective that can help organizations find a better approach to diversity.

Note, I’m not arguing that one kind of diversity is more important than another. Organizations should strive to address all of the following and more.

1. Gender and Cultural Diversity

Gender and culture dominate the diversity discussion. Plans to increase diversity are often measured by how they address the gender ratio and the cultural (including ethnic, racial, and religious) markup of the employee base. Overall, the tech industry has made huge efforts to attract more women, LGBTQ+ members, and people of colour by creating better training opportunities, adjusting their talent search, and making the workplace more inclusive⁵.

Various theories, such as the “leaky pipeline theory” and the “hostile workplace theory” have been brought forward to explain the persistent gender gap⁶. And structural barriers keep certain minorities from participating according to their labor participation rate. Meta, Google, Apple and others are backing affirmative action⁷ to increase diversity in higher education and thus their talent pools.

Gender and cultural diversity are the most visible markers of diversity as they can be communicated as part of one’s identity. A lack of diversity along these dimensions is thus more apparent, making gender and ethnicity targets for diversity quotas. However, organizations that focus on a few visible markers of diversity might get stuck at a very shallow understanding of the “why”.

This is ostensibly the case when increasing diversity is seen as a cost to the business. Too many organizations rely on employees to “volunteer” and organize events with the occasional external diversity representative being invited as a guest speaker. Unfortunately, this can mislead members of these organizations into thinking that diversity is meaningfully addressed, because these events happen.

It becomes quickly apparent how interested an organization is in promoting diversity, when actual change and investment are required. Is money spent to find and recruit women, LGBTQ+ members, and people of colour? Or does the organization put the burden on candidates to present themselves that way on LinkedIn? Are external communities and training organizations supported to help increase the talent pool for the whole industry? How about accessibility, accommodations, and promotions?

All of the above requires meaningful financial and organizational commitment. Deepening the understanding of what diversity really means can help avoid seeing it as a prescribed target that has to be achieved in the most cost-effective manner.

2. Cognitive & Behavioural Diversity

A diversity approach that is monomaniacally focused on achieving quotas misses additional dimensions of diversity. While gender and cultural diversity have been shown to be beneficial for organizations, this has been attributed to the fact that teams with people who think differently and do things differently are more creative at problem solving. A team or organization with low cognitive diversity can suffer from “groupthink” which leads to poor decision making and the failure to address risks. Put differently:

Homogeneous teams produce homogeneous outcomes⁸.

Taking cognitive diversity into account presents another opportunity for organizations to push themselves to create an overall better and more accommodating workplace. For example, individuals who tend to be more introverted might get overlooked in organizations which have a culture of dominance. These organizations deprive themselves of the value that introverted individuals can contribute, while at the same time failing to provide an equitable workplace.

Cognitive and behavioural diversities refer to the different ways that individuals perceive the world, learn, make decisions, communicate, and act. Some of these characteristics have been popularized by the various (pseudoscientific) personality tests that large organizations and consultancies are so enamoured with. Independently of what the benefits of particular cognitive and behavioural traits are, it should be apparent that there is often no “best way” of doing things when it comes to solving complex, novel problems such as AI. Thus, teams that can explore multiple approaches have a better chance of discovering solutions.

Cognitive diversity is important at all levels of hierarchy. It helps improve decision making, increases adaptability in a fast changing environment, enhances strategic anticipation, and promotes innovation⁴; all of which are vital for AI adoption.

3. Diversity of Experience

The different kinds of diversity described above are obviously not mutually exclusive. Gender and cultural diversity are probably correlated with cognitive diversity due to differences in upbringing, schooling, and social expectations. It’s thus not wrong in itself for an organization to address more visible markers of diversity.

However, focusing on a few select markers of diversity can distract from the more expansive definition of what I will call “diversity of experience” for short. Ultimately, an organization (as an entity with a purpose) should have a self-interest in being accommodating to individuals with a wide range of experiences to avoid getting trapped in the thinking and decision making of a particular group.

The fact that we are still far away from achieving diversity goals – even in the AI industry that would benefit the most – suggests that Goodhart’s Law⁹ is still alive and strong. Quotas and metrics can help an organization steer in the right direction, but they come with the risk of simplifying things and overlooking the depth and complexity of the challenge. Metrics will be “gamed” and perceived as another box to tick with the least amount of effort possible.

A wholesome appreciation of the value that “diversity of experience” brings to a team could help all members of an organization unite around a common goal of increasing diversity.

Conclusion

Unfortunately, the tech sector – and the AI industry in particular – are still less diverse than the general labour force. This is surprising, because diverse teams are essential to solving complex challenges like AI.

Equity is often presented in a language of social discourse that enterprises find challenging to integrate into their hiring and workplace processes. They live in a world of key performance metrics, quotas, budgets, and cost optimization.

I tried to outline a complementary framework that could serve as a counter-narrative to the prevailing surface-level understanding of diversity. What if diversity is desirable in and of itself for an organization, because it has value? It should thus be in every organization’s self-interest to provide a workplace that is equitable and attracts diversity.