AI could transform DEI policies for the better — or make them worse

Experts say companies should proceed with caution when using AI to enhance diversity, equity and inclusion

Article content

In the quest for a workplace that’s diverse, equitable and inclusive, some businesses are turning to artificial intelligence for salvation.

AI could be the objective, data-driven recruiter company leaders have long dreamt of — able to see only the skills and qualifications of a potential candidate without judging based on race, sex, accent or appearance, and craft job descriptions free from biased language.

Advertisement 2

Article content

Article content

AI could make a workplace more accessible for people with disabilities, such as through speech recognition systems, and could call out negative patterns impacting some groups, leading to fairer performance evaluations and promotions.

But it could also do the exact opposite, which is why many experts are warning companies to proceed with caution when using AI.

“The benefits exist only when AI is designed effectively and that’s a really big caveat,” said Thomas Sasso, an assistant professor at the Gordon S. Lang School of Business and Economics at the University of Guelph.

Cases of AI perpetuating biases, rather than eliminating them, have emerged because of limitations to how the tech is getting programmed. One in five Canadians has personally experienced discrimination from AI technology, including misrepresentation, stereotyping and reduced access to resources and opportunities, according to Telus Corp.’s 2024 AI report. People fear its potential to do harm, and among those who identify as LGBTQ+, 61 per cent worry AI may be used against individuals or communities.

Article content

Advertisement 3

Article content

Hiring managers should be cautious when using AI to fill roles. While the tech has moved past gaffes like suggesting only male candidates, the programs are still trained on historical data, and may reflect a time when workplace hiring was less diverse. “A user of this technology might think you’re identifying the best candidate, but instead it’s just identifying the typically successful candidates for organizations,” Sasso said.

Even more concerning, said Sasso, is that organizations put themselves at legal risk if managers can’t explain why they’ve hired someone. “We often talk about AI systems as a black box and so we don’t know how decisions are being made and what information is going into it,” he said. “If you can’t justify a decision, if you’re relying on this technology, you’ve now opened yourself up to a significant potential case of human rights violation.”

Still, many companies are using AI as part of their DEI strategies. For example, AI platform Diversio Inc. promises to take a company’s DEI “pulse” and provide the tools needed to create a “more inclusive and engaged workplace.” Such software is attractive to companies, because it has the ability to scan and analyze vast swaths of data quickly and efficiently.

Advertisement 4

Article content

Some leaders may think AI Is a good replacement for DEI-dedicated employees. But Sasso said an AI bot can never replace a human’s innate skills. “It will be able to identify some issues if those issues look similar to past ones, but being able to identify new issues or new ways that these biases emerge is going to be very complicated for these systems,” he said. For example, he said LGTBQ+ issues remain a blind spot for many businesses that AI doesn’t have the capacity to tackle.

A key part of any DEI strategy, he adds, is creating a way for people to learn from their mistakes. “So much of DEI is the human connection and building trust. It’s really hard to feel that trust when you’re talking to an algorithm,” Sasso said. “We know that talking to a manager, talking to a human, sometimes can be problematic or difficult, but it can also be a space of healing through that conversation.”

Those tasked with coding AI programs have been trying to correct biases for years by making coding teams slightly more diverse, updating the data sets algorithms are trained on, or purposefully encoding safeguards into the technology. But in their effort to do so, some have overcorrected. For instance, Gemini, the generative AI chatbot from Alphabet Inc.’s Google, recently came under fire for creating historically inaccurate images, such as Black Vikings and a female pope.

Advertisement 5

Article content

The lesson for business leaders who want to use AI in any capacity — be it to help create a presentation, recruit, or foster DEI — is that they need to understand what they are buying. “If companies buy something off the shelf they have no idea (what they’re getting themselves into). It’s whatever the developer or the vendor tells them,” says Eddy Ng, Smith professor of Equity and Inclusion in Business at Queen’s University. “We can’t rely on Google.”

That’s not to say there’s no space for AI in the workplace at all. Both Sasso and Ng describe themselves as cautious embracers of the technology. Indeed, Ng said using AI is crucial so future data sets can become more reflective of current workplace diversity. “If you don’t get put in that data set, then down the road when they develop, when they find solutions, your needs are excluded,” he said. “We need to embrace AI so that our data gets included, but at the same time we have to be mindful.”

Sasso said companies can use a three-step process to safeguard against inherent AI bias. First, business leaders should consider why and in what capacity they want to use AI and then recognize its limitations. To do that, they will need highly skilled people who can understand what information the AI is collecting and how it will be used.

Advertisement 6

Article content

Then, instead of rushing into a company-wide rollout, companies should pilot the technology in low-risk situations for a few months. This allows a business to assess whether outputs align with its values and practices and determine if it’s helping company productivity or improving customer and employee experiences, he said.

Recommended from Editorial

Finally, companies should consistently assess if AI is still working to their benefit.

“Through that process, you can have a more ethical, controlled and responsible way of using AI in organizational practices,” Sasso said.

Article content