After developing an innovative algorithm predicting the personality of web users as part of his doctorate, Golnoosh Farnadi quickly experienced success: many companies wanted to use his algorithm in their hiring process. Alarmed, she decided on the spot that her career would take a different turn.
“My doctorate aimed to develop a model to predict the personality of Internet users based on the traces they leave behind: the texts they write, the photos they publish”, explains Golnoosh Farnadi, assistant professor at HEC Montreal.
But along the way, various companies have contacted her, signifying their interest in using these models to help them with employee recruitment.
“I was interested in the theoretical aspect, but when I realized what we could do with these models in terms of hiring, I was scared,” laughs the 37-year-old researcher.
This is where his career plans changed.
Focus on risk
Before completing his doctorate, Golnoosh Farnadi wrote an additional thesis chapter. In it, she explained why the kind of model she was developing should not be used as part of a hiring process – the opposite of what a doctoral student would reflexively do.
“It’s worrying because this kind of tool is rarely 100% reliable,” she says. These models make mistakes, and those mistakes could result in one person being hired over another. That’s why I decided to change my research directions. »
During her postdoc, she therefore decided to focus more on the risks associated with artificial intelligence (AI). A field of research that she continues to explore today, as she focuses on issues of discrimination and bias in AI models.
Better health equity
Principal member at Mila, the Quebec Institute of Artificial Intelligence, and holder of a Canada-CIFAR AI chair, Golnoosh Farnadi is currently working on a project related to kidney transplantation.
She explains that there are software programs using AI all over the world that are used to match patients with kidney donors. However, these software are not yet perfect, because not all patients – despite the fact that they are also compatible in terms of medical characteristics – have the same probability of being selected by the algorithm.
“We are trying to develop another model, another approach, which ensures an equal chance for all potential patients to receive a kidney,” explains the researcher, who admits that the pace of progress in knowledge in her field is a constant challenge. .
“The models are progressing so fast, think of ChatGPT, that we are always behind: the algorithms are launched, and it is only afterwards that we, the researchers, try to make them fair and equitable. It would be better to do it before, but there are few regulations on this. »
Strong support
Established in Montreal for five years, Golnoosh Farnadi grew up in Iran. It was there, at a very young age, that she developed her interest in everything related to computers and technology.
I have always been fascinated by computers. As a kid, I loved playing with the Commodore 128 my dad brought home.
Golnoosh Farnadi, assistant professor at HEC Montreal
After completing a research journey that took her from Tehran to Belgium via the Netherlands, however, she now thinks that she has found, here in Montreal, a springboard to pursue her research.
“I love Canada, and Canada has treated me very well,” she says. My research chair supports me enormously, the Mila is fantastic, and I am surrounded by people who firmly believe in the research that we do. I am proud to collaborate with people with whom I share convictions. »