Connecting the Dots: Investigating Systemic Discrimination

Workplace Investigations Under the Microscope Part 3 of 3

Procedural fairness in workplace investigations is facing increasing scrutiny
and is a defining factor in whether an investigation will be upheld. This three part series looks at the lessons emerging for employers, HR leaders, investigators, and legal counsel from a recent arbitration decision.

Part 1 examined how a series of communications with the employer created a reasonable apprehension of bias. Part 2 turned to witness selection, credibility, and the weighing of evidence, and what the panel described as a closed mind. The final article turns to what an investigation into systemic discrimination requires, and where this one fell short. [1]

The grievance arose from an investigation into serious allegations of harassment, discrimination, and systemic discrimination. The union alleged that the employer failed to meet its duty to investigate and that the third-party investigator applied a flawed process. The panel agreed.

The panel found that the employer failed to reasonably investigate the complainant’s allegations of systemic discrimination in two ways. First was an unduly narrow approach to fact-finding that did not gather the context surrounding the specific allegations. Second was that the employer, having reserved for itself the job of deciding whether the facts amounted to discrimination, never actually took that step.

Why systemic discrimination investigations are different

It can often be difficult in systemic discrimination claims to identify a single rule, policy, or procedure that produces adverse effect discrimination. Drawing on the Court of Appeal’s decision in Ontario Midwives, [2] the panel explained that there is often “no single identifiable ‘villain’, no single action identifiable as ‘discriminatory’.” It can emerge instead from an “invisible structure” of practices or a “complex web of seemingly neutral, systemic barriers.”

That has direct consequences for how an investigation into systemic discrimination must be conducted. The Human Rights Tribunal of Ontario summarized the approach this way: alleged incidents, along with particulars of historical practices, policies and attitudes, must be viewed comprehensively and in aggregate. It is that “interwoven amalgam of conduct, actions, inaction, policies, practices, systems and attitudes” that is said to produce the differential treatment. The connections between incidents may not be obvious, linear, or continuous, but together the interconnected web is what constitutes the series of incidents that produce systemic discrimination. [3]

The panel explained how this shapes the analysis. Investigating systemic discrimination does not mean making findings about each incident in isolation and then standing back to see whether a pattern emerges. It requires considering each alleged interaction in the broader context of conduct, actions, inaction, policies, practices, systems, and attitudes, and then considering all the evidence in the aggregate. In other words, the context is not something layered on at the end. It is part of the evidence gathering from the beginning.

The problem with incident-by-incident fact-finding

The investigator in this case took a deliberately narrow view of the fact-finding role given their mandate was only to make factual findings, not legal findings. Their report explained that they had applied a considered understanding of what constitutes a fact to avoid straying beyond factual determinations into broader conclusions. At the arbitration hearing, the investigator maintained the same distinction: the experiences of others who allegedly had similar experiences might be relevant at the later stage of assessing and making a finding about systemic discrimination, but they were not relevant to their fact-finding role in the investigation.

The panel saw it differently. It accepted that an investigator must be cautious about drawing broad inferences from a single interaction. But it added an important qualification. In an investigation into systemic discrimination, an investigator must be equally cautious in first confirming that interactions and exchanges are in fact discrete, and do not belong to a broader pattern. Deciding at the outset that an incident stands alone is itself a conclusion, and one that shapes everything that follows.

The risk is that a narrow and incident-by-incident view of what counts as a fact can screen out the very information needed to understand whether a pattern exists. That is what the panel found happened in this case.

Investigating the dots

The panel pointed to the contextual evidence that went unexamined by the investigator. Other racialized faculty members in the same department had left and reported experiences similar to the complainant’s, evidence the panel noted was not even appended to the report. The same malicious rumours had circulated about two junior racialized faculty members at different times. There were also “red flag” statements, including comments describing someone as a “diversity hire” and those invoking the “angry Black woman” stereotype.

The investigator set aside this evidence of patterns and attitudes and treated it as irrelevant because the witnesses, or the events in which the comments were made, did not speak directly to a specific allegation the complainant had raised. That is, the evidence was excluded because it was contextual rather than incident-specific, yet that is the category of evidence a systemic discrimination analysis depends on.

The panel’s concern was not that the investigator had been retained only to make factual findings. Even on a pure fact-finding mandate, and especially where systemic discrimination is alleged, it is essential that the investigator reasonably investigate the dots and set them out in a way that it is possible for whoever is making the legal findings to determine whether they connect.

Minding the gap between fact-finding and legal analysis

The panel confirmed that an employer is not required to contract out the ultimate determination of legal liability to a third-party investigator. It is entitled to consider the facts and reach its own conclusions about whether they amount to discrimination.

Separating fact-finding from the legal determination is permissible and common. But the panel identified it as a fine line, particularly in a complex systemic case, and attached two conditions. First, the fact-finding must be broad enough to capture the contextual evidence the second stage will require. Where the investigation fails to consider alleged incidents in context, unduly narrows its scope given the breadth of the allegations, or does not set out the relevant contextual evidence, the employer will not be able to complete the analysis. In this case, the employer was not provided with the “raw material” required to reasonably assess whether the complainant had experienced systemic discrimination.

Second, the employer must actually perform the second stage. Here, there was no evidence that it did. It appeared to have relied on the finding that the allegations were not substantiated and dismissed the complaint on that basis, yet the investigator had deliberately refrained from making any systemic findings. The systemic discrimination analysis fell into the gap between the two roles.

The reach of a flawed investigation

The decision also illustrates that the consequences of a flawed investigation do not end when the report is delivered. After the investigation concluded, a respondent posted a widely viewed tweet referencing the investigator’s conclusions. The complainant’s evidence was that it “basically killed” their career, including being approached at a social event by a stranger asking about it.

The employer’s position was that the investigation had concluded by the time the tweet was sent, so it was not a breach of confidentiality it was required to address. The panel viewed it differently and found a causal connection between the failure to carry out a reasonable investigation and the harm the tweet caused, for which the employer bore some liability. The tweet drew its force from the outcome of an investigation the panel found unreasonable, and the employer took no steps to address the harm once it knew about it.

Although the panel accepted that confidentiality generally does not extend beyond the duration of the investigation, its decision shows that a flawed investigation can generate organizational and legal risk even after the file is closed through the reliance others place on its findings.

Key takeaways

Systemic allegations demand a wider lens than complaints about isolated incidents. These considerations are helpful to keep in mind:

• Scope the mandate to the allegations. Where systemic discrimination is alleged, the terms of reference should be broad enough to allow the investigator to capture patterns, practices, attitudes, and historical context, not only the individual incidents listed in the complaint.
• Gather context. Contextual evidence should form part of the fact-finding throughout the investigation. Assess each incident in its broader context and then consider the evidence in the aggregate.
• Do not screen out pattern evidence. Evidence about similarly situated employees, repeated conduct, and “red flag” comments can be relevant even when the witness did not observe the specific incidents a complainant alleged.
• Investigate the dots so they can be assessed. If the mandate is limited to fact-finding, the report still needs to lay out the contextual evidence clearly enough that someone else can determine whether a pattern exists.
• Confirm who completes the analysis. If fact-finding is separated in the mandate from the legal analysis, make clear who will perform the second stage and be sure it happens and is documented.
• Do not treat “not substantiated” as the end of the inquiry. A conclusion that individual allegations were not found to have happened is not the same as a considered assessment of whether the evidence, viewed as a whole, discloses systemic discrimination.
• Obligations can extend beyond the report. An employer’s obligations may not be
  discharged once findings are made and the report is issued. Where post-investigation harm flows from conduct that is causally connected to flaws in the investigation, the employer may be required to act to address and remedy it.

Bringing the threads together

Across the three articles, the same theme recurs. This is not a decision about an investigator who lacked the credentials for the work, and it is not about a single glaring error. It is about how a series of subtle procedural choices, each one somewhat defensible on its own, can accumulate until an investigation no longer meets the standard of procedural fairness: communications that create an appearance of seeking approval, evidence gathered and weighed unevenly, asymmetrical credibility assessments, and a fact-finding lens drawn too narrowly to capture the broader context needed for the systemic allegations being investigated.

Procedural fairness cannot be delegated to any single step in the process. It lives in how the mandate is carried out, how the investigator communicates, which evidence is pursued, how that evidence is tested, and what the organization does once the report is in hand. Investigations are conducted so that organizations can understand what happened, address it, and move forward with confidence. A sound process also ensures the parties are treated fairly and provided a meaningful opportunity to be heard. Given the scrutiny investigations are attracting, getting the process right is no longer just good practice. It determines whether the outcome holds.

Natasha Savoline is a partner, HR advisor, and workplace investigator at Bernardi Human Resource Law LLP. She supports organizations in cultivating healthy workplace cultures through HR/employment law advisory services, training, conflict mediation, leadership coaching, culture assessments, and workplace investigations.

Connect if you would find a conversation helpful.

---

1 Governing Council of the University of Toronto v. University of Toronto Faculty Association (June 1, 2026)
(Grievance Review Panel) (unreported).

2 Ontario (Health) v. Association of Ontario Midwives, 2022 ONCA 458 (CanLII) (Ontario Midwives).

3 Ibid, note 2; Association of Ontario Midwives on behalf of Ontario Midwives, 2014 HRTO 1370 (CanLII) at para. 37.