Diversity in technology isn’t just about fairness, it’s about bringing together teams that can build the best, most innovative products. While the statistics clearly show that having more a diverse team leads to greater innovation, the industry overall still hasn’t moved away from the stereotypical “bro” culture. Artificial intelligence research and development is, unfortunately, no exception.
At the AI Forum at C2 Montreal, BDC and Element AI brought together a panel of women AI leaders to discuss the obstacles and issues that women face in the AI industry, and propose ways to address a host of issues, from the stereotypes that hurt recruitment efforts to the measures that need to be taken to encourage more girls and women to study and work in AI.
In the panel were Erin Kelly, CEO of Advanced Symbolics, a company that uses AI to assess public opinion, predict outcomes, and prevent suicide; Joëlle Pineau, associate professor of computer science at McGill University, where she co-directs the Reasoning and Learning Lab; and Terah Lyons, a former advisor for the White House Office of Science and Technology Policy, now a Mozilla Foundation Technology Policy Fellow.
“There’s huge pressure in the field, and it’s up to us to allow for imperfection in female leadership.”
The discussion was led by Peter Misek, BDC IT Venture Fund Partner, who pointed out that one of his daughters is keen to work in computer science, but doesn’t want to be entirely surrounded by boys. When Misek asked how to fix this imbalance, Pinelle noted her experience at McGill, where they are already using creative approaches to draw more women into their CS programs.
“The majority of females who graduate in computer science did not come into university thinking they wanted to do CS,” Pinelle said. “Boys grow up thinking ‘I want to program computers’; Some girls do, many of them don’t, so we have to have that flexibility in the system.”
The way that McGill has encouraged this flexibility is by changing the way students are recruited into, and can access computer science courses. There are now two streams of introductory courses: one for those who already know how to code, another for beginners. They have also changed the way that students get into the Honours program, by removing a requirement that necessitated making that choice in the first year of study.
“Our challenge is, how do we make sure that at the end of this first year course, they want to take a second one?” Pinelle said.
By promoting joint programs, encouraging student advisors to recommend computer science to women students, and creating numerous access points into the program, McGill now has 50 percent women in their introductory computer science classes, many of whom continue on to do minors, majors, or even switch into Honours. The number of women in graduate programs still isn’t ideal, but Pinelle is hopeful that this will change.
Kelly believes that the reason so few women are interested in working in artificial intelligence is more a marketing issue than a policy one. She sees McGill’s approach as brilliant, because it exposes people who had never thought about computer science or AI to an exciting career.
— Johanne Cadorette (@JCadorette) May 24, 2017
She has worked to build a company that’s a balanced environment where her employees are social and diverse.
“A diverse set of people has a diverse set of preferences, and [they can] communicate those preferences to AI systems.”
“It’s about getting out there and talking about the things we do, not how we do it. I don’t talk about our algorithms; I talk about the people we’re helping, the companies we’re working with, the countries we’re working in,” said Kelly. “[The work] is very intellectual, but women don’t have a problem with that. They have a problem with the image of it; it looks boring and the people you’re working with are weird. We have to get out of that stereotype.”
Lyons noted that there is a significant body of social scientific research that proves that “women are more drawn for cultural, and also for emotional reasons to pro-social fields and to concerted applications that actually have real-world impact for computer science.” By marketing AI for good as their research focus, programs can have greater success in recruiting a more diverse group of researchers.
Lyons also believes that institutions are at fault for maintaining the stereotypes of technology as a field for men.
“One of the things that I encountered working in defense and intelligence in the US government was walking into rooms of uniformed men to talk to them about AI or robotics, or civil or commercial aircraft systems, and experiencing that very male-dominated, very homogenous culture,” said Lyons. “And it’s perpetuated by the institutions — the way the media talks about AI and technology in general — the hoodie stereotype is perpetuated by everything, from Hollywood to sitcoms to the way that companies market for new hires.”
— Kaliop Canada (@KaliopCanada) May 26, 2017
The panel went on to debunk arguments for why women can’t “keep up” in this field if they want to take time off to have a family. While the technology certainly changes quickly, Kelly remarked that women who take time off can re-enter in management roles rather than as coders.
Pinelle also explained that the field is flexible enough that women don’t actually have to leave. “It’s a field where you can work part-time very easily…you can work from home. You can dictate the hours and how you’re going to do it.”
She and Kelly described friends and colleagues who kept their babies by (or under) their desks as they worked, and spoke of their own positive experiences bringing their kids to work and to academic conferences.
The panel also discussed the need to shift away from the focus on perfection from women working in AI and technology in general.
— Event Marketer (@EventMarketer) May 30, 2017
“There’s a huge amount of pressure in the field, and it’s really up to us to allow for imperfection in female leadership and to allow for some ability to take the kind of risks that men in their position [can].”
Teaching women to take risks and fail starts in childhood when, Kelly argues, it’s important to emphasize effort over results, and improvement over perfection.
“The type of women who go into technology is a very specific type of person,” said Pinelle. “It’s women who don’t care much about failing and are willing to take risks, and are maybe more confident.”
“The second, that we diminish that expectation [of perfection in women] and reset our cultural stigmas around what it looks like to be a successful business person or technologist, I think we’ll be in a much better place,” suggested Lyons.
For the AI industry as a whole, and society in general, having women and other underrepresented populations involved in this work, will be a vital component in diminishing bias within AI systems.
“Why do we need a diverse set of people to be building AI systems?” asked Pinelle. “Because a diverse set of people has a diverse set of preferences, and [they can] communicate that diverse set of preferences to the AI systems, and that will change how the AI behaves.”
Disclosure: Lauren Jane Heller is an employee of McGill University