Imagine you’re the CEO of a major ride-share company. Someone accuses your organization of sexism, pointing out that nearly all your drivers are men. Your first instinct is to launch into the familiar tech industry response about engineering demographics—fewer women in computer science, skewed applicant pools, and so on. But then they clarify: they’re not talking about engineers, but the actual drivers. Before you can explain that your platform doesn’t actually “hire” drivers—anyone can sign up with minimal requirements— such that there couldn’t possibly be any sexism restricting women drivers, you ask about their source. They cite “Chickipedia,” apparently an alternative to the “sexist” Wikipedia because unlike Wikipedia, more than 10% of its editors are women.

This scenario highlights a broader issue: conversations about diversity, equity, and inclusion (DEI) often become quite dumb. One camp insists demographics should never influence decisions—may the best candidate win! The other sees any demographic skew not mirroring the broader population as evidence of discrimination requiring intervention.

10 years into DEI going mainstream, it seems shocking how little progress the discourse has made. Both extreme perspectives seem obviously wrong to me, yet it seems like there have been no bona fide good faith attempts to actually come up with a thoughtful framework for analyzing DEI decisions.

To this end, this post aims to establish a comprehensive taxonomy of DEI considerations. The goal is to highlight all relevant factors that may be taken into account when evaluating the potential value of a DEI intervention. Given that there are over 1,000 possible permutations of these variables, a thorough evaluation of each is beyond the scope of this post. Instead, by outlining the possible relevant factors to consider, I hope to contribute to the conversation on when various DEI interventions are justified and to understand how to implement in the least costly and most effective way possible.

Discrimination Types

There are many different “kinds” of discrimination

1. Actual Racism/Bigotry: Direct, intentional discrimination based on race, gender, or other personal traits.
2. Government-Approved Discrimination Unrelated to Personal Traits: Certain roles may require candidates from specific backgrounds. For instance, Canadian Supreme Court judges are appointed to ensure representation from different regions of the country.
3. Role-Related Discrimination: Some roles may justify demographic preferences, such as hiring female gynecologists for patient comfort or male actors for specific parts.
4. Disparate Impact Due to Necessary Requirements: Physical demands, like strength tests for police officers, can lead to demographic imbalances—not due to bias but due to the role’s demands.
5. Arbitrary Requirements Impacting Certain Groups: Seemingly neutral standards may disproportionately affect certain groups, like dress code laws preventing Sikhs from becoming police officers.
6. Questionable Requirements with Disparate Impact: Sometimes, arbitrary standards, like math requirements for teachers, inadvertently reduce the number of certain groups (ie Black teachers), when Courts find there to be inadequate evidence to support the need for such a standard to be ie a geography teacher.
7. Culturally Hostile Environments: Work environments may foster cultures that alienate certain groups, like environments where sexist jokes are common.
8. Background-Based Preferences Affecting Immutable Traits: When employers prefer candidates from particular socioeconomic or cultural backgrounds (ie law firms like hiring students who look good in a suit, have a shared interest in skiing etc), this can indirectly favour certain demographics without direct discrimination.
9. Culture Rationally Favouring Traits Tied to Group Traits: Some fields’ cultures align with traits that correlate with specific demographics. For example, Wikipedia’s preference for detail-oriented obsessive individuals attracts more men, as does investment banks who seek analytical people suited to high-stress confrontational environments.
10. Culture Favouring Traits Unrelated to Role: Some workplaces favour certain demeanours, like confrontational styles, even in roles where this isn’t necessary (ie some restaurant kitchens), which may inadvertently reduce diversity.
11. Unconscious Bias in Evaluation Criteria: Without overt discrimination, unconscious biases in evaluations may skew hiring and promotions.

Context

Environments where there is an expectation that everyone be welcome, or exclusion may impact your livelihood are more critical to immunize against discrimination.

1. Livelihood-Based Contexts: Settings such as employment or universities, which are intended to be inclusive for all and are typically subject to extensive regulation.
2. Recreational Contexts: Environments like amateur sports teams, where the impact of discrimination is lower, inclusivity is less critical, and government regulation is typically minimal.

Types of Interventions

Based on discrimination type and context, interventions may take different forms, each with unique implications and trade-offs.

1. Assigning Additional Points for Targeted Groups: Programs like university affirmative action programs consider underrepresented demographics more favourably.
2. Setting Minimum Hiring Requirements: Some organizations mandate a minimum number of hires from underrepresented groups.
3. Imposing Maximum Limits on Overrepresented Groups: Instead of boosting underrepresented groups, this approach limits the majority group.
4. Requiring Certain Groups for Certain Rationally Roles: Some roles might specify particular demographics, such as hiring Indigenous scholars for research on Indigenous issues.
5. Requiring Certain Groups for Certain Arbitrary Roles: Some roles might specify particular demographics, such as hiring “equity deserving groups” scholars for research on topics not benefited by their identity background.
6. Interview Minimums Without Selection Quotas: The NFL’s “Rooney Rule” requires diverse interview pools for certain positions, but no stipulations on hiring.
7. Cultural Modifications for Inclusivity Workplaces might adapt norms to encourage inclusivity, such as fostering less confrontational communication.
8. Increasing the Pipeline: Investing resources to promote a specific opportunity or field within an underrepresented region, community, or group. For example, a tech company might work to increase female enrolment in computer science programs, which will eventually help when it comes to hiring.

Costs of Intervention

Different interventions come with different costs.

1. Lowered Quality of Candidate Diversity initiatives may sometimes lead to hiring under qualified candidates which may impair an organization going forward. Identifying if the environment/role is one where preferring a more diverse candidate comes at the cost of selecting someone who is expected to perform worse, or there is a large supply of similarly talented candidates.
2. Negative impact on organizational culture: If an organization has a particular culture and is filled with similar, compatible people, making changes to that culture may impair the organization’s functioning.
3. Inequity Toward Qualified Applicants Denied Admission: Affirmative action in law schools, for instance, can limit qualified candidates from overrepresented groups.
4. Stigmatization and Poor Fit for some Candidates: Candidates who enter under adjusted criteria may feel stigmatized or struggle in poorly fitting environments.
5. Opportunity Costs for Higher-Impact Diversity Initiatives: Resources devoted to top-level diversity, such as high-visibility judicial roles, may reduce diversity efforts in lower courts where representation could have greater impact.
6. Raising the salience of demographics: in a neutral society, people’s immutable traits would matter less rather than more, and by bringing so many benefits by associating with a certain identity, it can cause additional division.

Benefits of Intervention

Interventions can yield numerous benefits for for different kinds of groups: the individuals themselves, the organization making the intervention, and society writ large.

1. Organizational Benefits Through Increased Diversity: More diverse teams make decisions that better reflect a broader range of perspectives, improve decision making or can help an organization appeal/better serve certain communities (ie hiring a female play by play commentator can increase a basketball team’s prospective audience)
2. Societal Benefits Through Increased Diversity: For example, encouraging Arabic-speaking individuals in medicine improves patient communication and care, or enabling various groups to have more potential leaders with different kinds of experiences.
3. Strengthened Future Talent Pipeline: Doing outreach in advance of any DEI decision, as a way to increase the diversity for the future. IE supporting female students in computer science or featuring women speakers at conferences helps increase women in tech, as a way to make hiring in tech easier in the future.

 

Justification for Intervention

1. At times, candidates may seem underqualified due to limited exposure or resources rather than lack of capability. In these cases, intervention can address systemic disadvantages that go away almost immediately. For example, a candidate who couldn’t afford professional clothing or had to work extra jobs during school may initially seem less polished but could excel with equal access and preparation.

—

I genuinely don’t have a solid sense of what’s good or bad here, other than to say this is complex, and most people take too shallow a view of it. Reflecting on the taxonomy outlined above, there are a few things that stand out to me.

1. Many environments have a meritocratic dimension—meaning that if the best candidate is not selected, it may lead to poorer outcomes. It’s important to determine whether the situation truly requires a “best” candidate or if there is a large supply of equally qualified candidates. Similarly, if someone loses a spot due to DEI considerations, especially in environments where people work hard to be selected for competitive positions, candidates may feel personally hurt if they believe diversity initiatives prevented their success, which creates a lot of toxic costs.
2. One notable aspect is how courts or administrative bodies view evidence. In short, adjudicative bodies typically require tangible evidence as justification for something, which creates a problem because for most things in this world, there is no evidence—not because the connection isn’t true, but because evidence is hard and expensive to produce. Said differently, if courts take the position that unequal outcomes can only be justified with evidence supporting both the rational connection and the cause of the biased outcomes, then, with evidence being so uncommon and hard to produce, many things that in fact make sense will not be accepted by the courts. For instance, while math proficiency might not be necessary for geography teachers, many people feel that an inability to do math at a high school level signals lower intelligence, and people of lower intelligence probably won’t make great teachers—which is probably true, just not the type of thing you would expect to have evidence to argue in support of. This means there is an entire class of cases where judicial oversight will lead to worse outcomes, just due to the nature of how adjudicative bodies work.
3. There are a huge number of areas where hiring discrimination is done for self-interested market-related reasons. For example, hiring a female play-by-play commentator for basketball games to increase outreach to non-traditional fans, or a company wanting to increase its internal diversity to broaden internal perspectives to make their product better. It seems that rational self-interest does, and should, guide most of these conversations, so it feels a bit odd that so many of them are viewed as matters of justice or virtue rather than simply good business decisions. The market naturally biases most organizations into increasing diversity, and imposes a significant cost on those that discriminate.
4. One of the biggest challenges in these conversations is that it’s unclear what demographic representation actually entails. Diversity encompasses much more than one’s immutable traits. When broad labels like Hispanic, Black, or Asian are used, they encompass a wide range of identities with significant differences, and an individual might not be representative of the entire group. Similarly, just because someone belongs to a certain group doesn’t mean they will be fully representative of or supportive of that group’s broader concerns. Diversity labels are often so broad that targeted outreach efforts intended for Black Americans or Indigenous people might primarily benefit recent, affluent West African immigrants or those who are only distantly Indigenous and have no meaningful connection to that heritage or experience of discrimination related to it. This is especially challenging with terms like “equity-deserving groups,” which, as currently defined, lack the specificity needed to be a practically useful concept.
5. Given the emphasis on hiring mandates and the contentious nature of this topic, it’s surprising how little focus there is on expanding applicant pools through targeted outreach rather than through direct intervention in admissions criteria. Perhaps this approach requires more time and effort, or perhaps it has been tried and found ineffective, making the more disruptive interventions feel necessary, even if they’re unpopular. One could generalize from this that there’s been insufficient focus on the actual efficacy of all of these measures in the first place. To some, this suggests that the outcomes are not the primary goal—that instead, the focus lies in the satisfaction derived from engaging in the politics and power dynamics of the discourse.