In an age when data increasingly drives decision-making, the ethical implications of artificial intelligence (AI) models are under scrutiny. And the insurance industry is no exception.

Insurance companies are using increasingly sophisticated algorithms to estimate the risk represented by their customers, based on a growing mass of information. “For example, in Quebec, in 2021, the Société de l’assurance automobile du Québec recorded 27,888 injuries caused by road accidents. In each of these cases, each person behind the wheel of a vehicle can represent a different risk for insurers,” explains Olivier Côté, a doctoral student in actuarial science at Université Laval.

And different risks mean different policies and different costs. All this, with the risk of bias and discrimination depending on the data used, the weight of each variable, the training of the models, and even the models themselves. Olivier Côté is working to mitigate these discriminatory biases. The initial findings of his thesis project have been published on the Social Science Research Network website, in collaboration with Professor Marie-Pier Côté, from Université Laval’s School of Actuarial Science, and Professor Arthur Charpentier, in the Department of Mathematics at the Université du Québec à Montréal, and presented on May 16 at the Journée sur l’équité et la discrimination en assurance.

“Given the insurance industry’s dependence on personal data, and the major role this industry plays in society, companies today must be particularly concerned about the fairness of their practice among insureds,” explains the student, who is working under the supervision of Professor Côté and Professor Charpentier.

Read more on the ULaval Nouvelles website