Systèmes d'IA de recrutement et d'embauche produisant des résultats discriminatoires

Canadian employers are increasingly using AI-powered tools for hiring and recruitment — automated resume screening, video interview analysis, candidate matching algorithms, and predictive workforce analytics — with limited transparency about how these systems evaluate candidates and growing evidence that they can produce discriminatory outcomes along protected grounds.

The adoption is substantial and accelerating. Statistics Canada reported that 12.2% of Canadian businesses used AI as of Q2 2025, more than double the rate from the previous year, with human resources and recruitment among the most common applications. LinkedIn's AI-powered hiring tools are used by thousands of Canadian employers. Major Canadian organizations use platforms like Workday, iCIMS, Greenhouse, and HireVue that incorporate AI for candidate screening and ranking. The Canadian government itself uses AI-assisted tools in some hiring processes.

The evidence of bias in AI hiring tools is well-documented internationally and directly relevant to Canadian deployments. Amazon's internal AI recruitment tool, developed to screen resumes, was found to systematically discriminate against women — penalizing resumes that included the word "women's" (as in "women's chess club captain") and downgrading graduates of all-women's colleges. Amazon abandoned the tool by early 2017 after failing to eliminate the bias. Reuters publicly reported the project in October 2018. The root cause — training on historical hiring data that reflected past discriminatory patterns — is present in virtually all AI hiring systems trained on employer data.

Video interview analysis tools raise particular concerns. HireVue and similar platforms assess candidates based on facial expressions, tone of voice, and word choice, generating scores that influence hiring decisions. Research has demonstrated that these tools can discriminate against candidates with disabilities (different facial expressions, speech patterns), candidates of different racial or ethnic backgrounds, and candidates whose first language is not English. HireVue discontinued its facial analysis feature in 2021 following criticism, but other companies continue to offer similar capabilities.

Canadian human rights law prohibits employment discrimination on grounds including race, national or ethnic origin, colour, religion, age, sex, sexual orientation, gender identity, marital status, family status, disability, and genetic characteristics. The Canadian Human Rights Act and provincial human rights legislation apply to hiring processes regardless of whether a human or an algorithm makes the decision. However, the mechanisms for detecting and proving algorithmic discrimination are underdeveloped. An applicant rejected by an AI screening tool typically receives no explanation and has no visibility into the criteria that were applied.

The Canadian Human Rights Commission has recognized the risk. The CHRC has stated that "algorithms that are trained on historical data can perpetuate and amplify existing patterns of discrimination." However, no specific enforcement action has been taken against discriminatory AI hiring practices in Canada.

The structural concern is that AI hiring tools create a high-throughput discrimination machine. When a biased algorithm screens thousands of applications, the number of people affected is far larger than traditional human bias — and the discrimination is invisible because it occurs inside a black box that neither the employer nor the applicant can inspect. Canadian employers may be unknowingly violating human rights law by deploying tools they cannot audit, evaluate, or explain.