DIRE Warnings, Part 4: Engineering a Representative Solution
How Diversity, Inclusion, Representation, and Equity are supposed to work and where they go wrong.
This article is Part 4 in a series. Part 1 summarized what the series will demonstrate. Part 2 addressed the definitions of equity, diversity, and inclusion including misconceptions. Part 3 addressed relevant (Canadian) legislation and psychology.
Equity
Part 2 of DIRE Warnings highlighted two nearly opposite definitions of diversity. The Canadian Centre for Diversity and Inclusion (CCDI) defined diversity as being about the individual and the unique dimensions, qualities, and characteristics that we each possess. It includes personality dimensions which are core to our identity (Layer 1), immutable traits that we have little to no control over such as race, ethnicity, and gender (Layer 2), more controllable characteristics such as parental status and education (Layer 3), and organizational dimensions such as seniority and job title (Layer 4).
By contrast, Misconceived EDI (equity, diversity, and inclusion) focuses on multiple people having “differences in race, colour, place of origin, religion, immigrant and newcomer status, ethnic origin, ability, sex, sexual orientation, gender identity, gender expression and age”, and often targets head-counts by immutable trait within the layers of an org chart.
Parts 1, 2, and 3 of DIRE Warnings note that the CCDI, research literature, legislation, and psychological foundations of EDI all agree on the correct use and risks of misuse. Under the correct use (Part 2, Figure 3), equity aims to accommodate people’s needs as individuals, rather than blanket “one size fits all” approach, so that “they might receive different, flexible treatment at times in order to have the same opportunities for success as everyone else”.
It is important to note that this is not a new concept. In law and other applications, the term equality has long been applied to situations with differing needs and accommodations. For example, the Canadian Human Rights Act (CHRA) was originally written in 1977. It has as its purpose, as described in Section 2,
“… the principle that all individuals should have an opportunity equal with other individuals to make for themselves the lives that they are able and wish to have and to have their needs accommodated, consistent with their duties and obligations as members of society, without being hindered in or prevented from doing so by discriminatory practices …”.
Assessing individual needs and accommodating them to produce equal opportunities is the very basis of the CHRA, exactly matching that of the CCDI’s description of equity.
Similarly, the Canadian Charter of Rights and Freedoms, Section 15, sets out equality rights as,
“Every individual is equal before and under the law and has the right to the equal protection and equal benefit of the law without discrimination and, in particular, without discrimination based on race, national or ethnic origin, colour, religion, sex, age or mental or physical disability.”
These four components of equality (before, under, protection, and benefit) are more formally defined in law as Substantive Equality (EDI “equity”) vs Formal Equality (EDI “equality”) and take into account people’s differences1. The conceptual difference was built into the Canadian Charter from the start between 1980 drafts and 1983 completion2, but was perhaps first applied in 1985 under R. v. Big M Drug Mart,
“The equality necessary to support religious freedom does not require identical treatment of all religions. In fact, the interests of true equality may well require differentiation in treatment.”
A more general application came in 1989 under Andrews, stating,
“It must be recognized at once, however, that every difference in treatment between individuals under the law will not necessarily result in inequality and, as well, that identical treatment may frequently produce serious inequality. This proposition has found frequent expression in the literature on the subject but, as I have noted on a previous occasion, nowhere more aptly than in the well-known words of Frankfurter J. in Dennis v. United States, 339 U.S. 162 (1950), at p. 184: It was a wise man who said that there is no greater inequality than the equal treatment of unequals.”
That now takes us to U.S. law going back to 1950. The point is that the concept is neither new nor a change in approach; this distinction between equality and equity in EDI merely serves as a formalization of the concept in common parlance instead of esoteric legal language.
It is also important to note that the accommodation of individual needs in the CCDI and legal definitions applies to individuals who fit the need. It can apply to many individuals who define a group based on fitting the metric that is the identified need. Misconceived EDI often applies equity to groups of individuals based on the correlation between the metric of interest and other traits.
Figure 1 shows this discrepancy in definitions by analogy. A, B, and C will look familiar to many readers and similar imagery can be found widely on the internet. Three different height individuals are trying to look over a fence. One is tall enough to see over and two are not. In image B, all three receive identical support (EDI “equality”) even though the tall person doesn’t need it and the shortest person still can’t see over the fence. In image C, enough help is given to each individual so that they can see over the fence including no help to the tall person and a lot of help to the shortest (EDI “equity”).
However, it is very common in Misconceived EDI to then take image C as an an analogy based on social groupings such as race or gender and then apply help to only that social group that is statistically most in need, typically based on some statistical average. But doing so simply recreates the problem in image B, not the solution in image C.
This error is seen in image D. Social groupings are represented here by blue, red, and green. Blue people are in least need of help statistically so receive no help. Red people need more help. Green people need the most help. In providing help in this manner, indeed the average heights of blue, red, and green people are made equal, as are the tallest, shortest, and middle individuals. They are statistically identical.
This has not solved the problem. There is still one blue, one red, and one green person who cannot see, and the tallest red person was helped even though they didn’t need it.
Image E corrects the problem. It applies the principle of equity from image C properly. People are not helped by social group, but instead by individual need. Now everybody can see, and anybody that didn’t need support didn’t receive any.
The fundamental problem with image D is that it doesn’t group people by the relevant metric. The issue in this analogy is ability to see over the fence. In image D there is certainly a correlation between social grouping and height, but that doesn’t make them interchangeable. The metric of interest is height, so the proper groupings is based on height or need for help based on height.
Image F shows this proper grouping. It is the same solution as image E, but the individuals have been regrouped by height rather than social category. Image E shows three groupings for “tall” (who don’t need help), “medium” (who need a little help), and “short” (who need a lot of help). In principle, it could be simplified to two groupings of those who do or do not need help. Or, instead of grouping people they could simply be ordered from tallest to shortest and giving individual proportional help to their individual need without grouping anybody.
The Misconceived EDI definition of equity in image D not only replaces the metric of interest with one that correlates (social category), it also creates a first-past-the-post winner-takes-all competition. The group that has a slightly worse average than others become the sole focus of getting support. This misconception of group average needs (image D) as analogous to individual needs (image C) is common when tribal psychology is evoked, a topic discussed in great detail in Part 3 of DIRE Warnings. The article describes the psychology as functioning somewhat like a spreadsheet roll-up function to replace individual variations with a group average, and then “people tend toward seeing only a single representative value for the whole group and anthropomorphize the group summary (“roll-up”) as a single person.”
When evoked, tribally minded intuitively think of image C and D as being essentially identical, where the individuals in image C are “representative” roll-ups of the whole group. Hence a solution in C must also apply to D. But it doesn’t, as the figure shows. Image C works because it is about individual need, and so it is analogous to E or F in which groupings are ignored in favour of treating people as individuals.
Diversity and Inclusion
From Part 2, diversity is about the different dimensions individuals bring and is a given in any organization. Inclusion is the goal that respects diversity in individuals and treats them equitably (Figure 1) by being flexible to their needs and working to ensure their input is included and valued.
None of equity, diversity, or inclusion are about head-counts. None are specifically about immutable traits or social categories. This clear and significant discrepancy between EDI used properly and Misconceived EDI raises the question on where this confusion comes from.
The CCDI training in Part 2 provides a hint in the confusion between diversity and representation. The CCDI training was designed from beginning to end to dissuade people from that misconception. From Figure 3 in that article, representation is defined as “the numerical presence of individuals who have specific visible characteristics or diversity dimensions, such as women or visible minorities.” Hence representation does refer specifically to Layer 2 social categories defined by immutable traits, and it does refer to the relative statistical variability of these categories. This does appear to map closely to the Misconceived EDI definition of diversity.
Representation also has components of head-counts in org charts which are not directly included in the Misconceived EDI definition of diversity so it isn’t a 1-to-1 mapping. But, the incorrect definition of diversity is often used in conjunction with statistical head-counts at different levels of an organization. It appears that the common misuse of the term may come from extracting the Layer 2 social category component from representation and thinking that is what is meant by diversity.
However, as Part 2 of DIRE Warnings showed, the Misconceived EDI definition is almost the exact reverse of what diversity is supposed to mean: the treatment of individuals as multi-dimensional rather than one-dimensional or simply as a head-count. This means there is still a missing piece since representation is not the opposite of diversity.
This missing piece is found in the denominator of representation; it is not merely a head-count, but in comparison to a specific baseline. That specific baseline is defined very tightly and for very important reasons that define the very basis for why representation matters. The CCDI definition for representation (Part 2, Figure 3) contains this additional context:
“A company is representative if its workforce reflects the differing groups that are available for work and have the skills the company needs.”
This framework and definition is also directly consistent with, and built into, Canadian legislation. Representation is defined in the Employment Equity Act (EEA)3 under Section 5, along with the means to achieve equity through inclusive practices, meaning,
“…instituting such positive policies and practices and making such reasonable accommodations as will ensure that persons in designated groups achieve a degree of representation in each occupational group in the employer’s workforce that reflects their representation in (i) the Canadian workforce, or (ii) those segments of the Canadian workforce that are identifiable by qualification, eligibility or geography and from which the employer may reasonably be expected to draw employees.”
That is, representation is not a comparison with the general population or some a priori selected target number. Similarly, representation is not about increasing the numbers of any social category in the availability pool. This doesn’t mean there isn’t value in trying to do so, just that such efforts are not the purpose of representation.
To understand the purpose of representation as defined in both the CCDI training and EEA legislation, why that definition is a necessity, and where increasing the numbers can fit in, it is perhaps best to start from understanding the employment system as a dynamic flow from an engineering perspective.
Consider an availability pool of people that are potential hires for a job for a specific need that includes some form of skills, education, and experience. Flowing into this pool are the newcomers to the those qualifications which can include sources from universities and colleges, skilled employees leaving other jobs, immigration of skilled workers, re-trained workers, or any other potential source. Flowing out of this pool are people getting hired into positions into various organizations. The inflow, pool, and outflow can be separately analyzed by demographic social categories such as by Layer 2 traits.
Consider as a hypothetical fictional example: Suppose it is a smoothly running system with 20% of the software engineers being Klingons including 20% of the input flows, 20% of the availability pool, and 20% of the outflows from the pool to the organizations. (For statistical purposes, suppose the variation is +/- 2%.)
Now suppose one organization looks at its percentage of Klingon software engineers and discovers it is only 10%. That is well below the range of all organizations as a whole. Clearly there is something unique at this one organization. Following from Section 9 of the EEA, the organization is supposed to investigate, understand, and fix the reasons (e.g., remove barriers) for the shortcoming (under-representation), or identify there is a bona fide explanation.
That takes hard work. The organization can look at demographics of applicants and see if the problem is attracting them in the first place. If applications are high but interview percentages are low, the issue is in the selection process. If interviews are high but offers are low, it’s in the interviewing and selection 16 process. If offers are high but acceptance is low, it’s something that turns off the applicants. If hiring is high but retention is low, it is likely something about the environment of the organization.
The explanations can take time to discover and fix. On attracting talent, it might be that there are confounding factors. If it is a governmental organization that has bilingualism requirements then the proper availability pool needs to change to exclude all people who are not sufficiently bilingual. The comparison then needs to be with the percentage of the availability pool that are both bilingual and Klingon. If that is 10%, the source has been found. If the bilingual requirement is a necessity for the position, then the 10% is a bona fide correct number, free from unfair bias.
Alternatively, if there are no confounding variables, perhaps the issue is that job postings are written in a way that fails to attract Klingons. That is, the hiring process did not consider the different dimensions of applicants, meaning their diversity. They may all have the same education and occupation as software engineers (Layer 3), but they may have different employment interests based on different cultural backgrounds (Layer 2). The organization can better attract talent by being more inclusive in the postings by considering that applicants have multiple dimensions.
They might also be failing to attract introverts, older applicants, or people with children – though these are not analyzed through the EEA lens of representation. The point of EDI is that these characteristics should be looked at anyway. The specific Layer 2 traits in the EEA are based on specific discriminatory exclusions from large scale historical patterns, but EDI is more generally about removing as many unnecessary limiting factors as possible across all traits of all dimensions. In this manner, EDI and the EEA can work in tandem to address issues of representation. EDI does not have representation as its target goal but good EDI policy can help to remove existing barriers from the bottom up, and application of the EEA Section 9 investigation can localize where the problems exist and what might be causing them to result in under-representation of the EEA categories.
From an engineering perspective, such biases act like clogs or filters in the system for specific components of the flow, representation and the EEA acts as a sensor feedback network comparing the flow in the system to the availability pool and localizing the clog, and EDI helps to unclog and free the flow unfiltered.
Neither EDI nor representation serve to increase the inflow or availability pool concentrations. Representation is not a quota system that sets a desired target; it is a detection system for potential barriers inside an individual organization. It is not a comparison to general population but to the comparable availability pool for an occupation in the region. It is an alerting system; it evokes an investigation into the causes of the statistical difference and build report on what problems were found and how to fix them.
Increasing the percentages of the inflow is a separate outreach effort that needs to be done at the sources. Perhaps 30% of the people who want to software engineers are Klingon, but they are limited by the inflow to the sources themselves (e.g., university). In that case, EDI policies – following the proper definitions described in this DIRE Warnings series – applied at those organizations can help to unclog the flow.
Conversely, if the system is not clogged, attempting to artificially increase the flow at any stage will very likely lead to a failure mode. The concentration (percentages) of specific components (social categories) in the outflow (hired into organizations) cannot sustainably exceed the concentration in the availability pool or inflow as a matter of simple mathematics.
Part 5 of this series will look at failure modes for the combined diversity, inclusion, representation, and equity (DIRE) system.
Department of Justice, Chaterpedia, Section 15 – Equality rights, retrieved from https://www.justice.gc.ca/eng/csj-sjc/rfc-dlc/ccrf-ccdl/check/art15.html
Anne F. Bayefsky, Mary A. Eberts (1985). Equality rights and the Canadian Charter of Rights and Freedoms, Carswell, Toronto.
Employment Equity Act, S.C. 1995, c. 44, retrieved from https://laws-lois.justice.gc.ca/eng/acts/E-5.401/FullText.html