Steering ethical use of AI in Canada’s financial sector
There could be a version of the future in which banks discriminate against certain borrowers by charging them higher interest or giving them fewer loans, on the advice of a “smart” robot.
Whether or not this reality manifests in Canada depends on what financial institutions understand about artificial intelligence—and how they decide to use it.
“If an algorithm is biased against females, for example, then females will be disadvantaged and will not be able to get loans to get an education, start a business, buy a car,” says Alexey Rubtsov, a TMU professor specializing in financial mathematics research. “We want to avoid perpetuating biases in our society.”
Considering the numerous benefits of AI, including increased efficiency and automation of complex tasks, it becomes even more crucial for financial firms to develop algorithms properly and ensure they can detect and mitigate any instances of discriminatory decision-making. With the global AI market expected to grow 37% annually (external link) until 2030, maintaining fairness and trust in AI operations is of utmost importance, according to Rubtsov.
It’s one of many insights Rubtsov contributed to a new initiative by the federal government to help banks, investment firms, and other financial institutions safely and ethically use AI. As part of the Financial Industry Forum on Artificial Intelligence, he was one of 57 AI thought leaders from financial service firms, regulators, financial technology companies, and academic institutions who produced the April 2023 report: “ (PDF file) A Canadian Perspective on Responsible AI (external link) .”
Initiated by the Office of the Superintendent of Financial Institutions in partnership with the Global Risk Institute, the 52-page report outlines key challenges, considerations, and best practices for AI adoption in four interrelated areas that comprise the “EDGE” principles: explainability, data, governance, and ethics. Rubtsov shared his expertise on these principles—having the right values, techniques, and regulatory guidelines in place to ensure AI is used in ways that are moral, fair, and protect privacy.
Rubtsov also weighed in on explainability, which refers to financial professionals being able to understand the theories, data, and methods that AI models use to make decisions. Rubtsov says when executives recognize the outcomes of AI algorithms, they can determine whether they are ethical and align with their corporate values.
“Explainability ensures that the right decisions are made for the right reasons,” Rubtsov says. “Some AI tools are extremely complicated, and even those with technical expertise sometimes cannot explain why algorithms made certain decisions. In finance, where decisions carry high risks, the way this technology works needs to be widely transparent.”
As head of the Financial Mathematics Research Group, Rubtsov and his team of research fellows and students study and develop mathematical models that can support financial firms as they deal with emerging developments in AI, data mining, cryptocurrencies, and fintech. They also devise approaches related to optimizing investment returns; managing market, environmental and systemic risks; and pricing derivative contracts.
Among the specific research endeavours taking place is an examination of how to promote fairness in AI-generated decisions in service of equity, diversity, and inclusion. Graduate student Alejandro Hernandez is exploring mathematical methods to limit biases in AI models to prevent outcomes that discriminate against individuals based on their race, gender, age, or other factors.
Meanwhile, graduate student Luka Milić is exploring how financial firms can stabilize their operations in the face of global warming. He is specifically studying how to optimize investment portfolio allocations to mitigate the impacts of adverse climate effects and the transition to a lower-carbon economy. The study is supported by Rubtsov’s Discovery Grant from the Natural Sciences and Engineering Research Council (NSERC) investigating “A Financial Mathematics Approach to Climate Change Risk.”
Similar inquiries take place in TMU’s Financial Mathematics program, the first and largest undergraduate program of its kind in Canada. Rubtsov engages students in developing their fundamental knowledge—linear algebra, probability and statistics, calculus, geometry—as well as building key sector competencies relating to analyzing markets and investments, computational methods, evaluating risks, ethics in finance, AI, and business leadership.
“We want students to be versatile in their skills, so we incorporate coding, quantum computing, blockchain technologies, and managerial skills into the curriculum so that when they join financial institutions, they are well-equipped to guide their organizations in this fast-developing field.”