Building a diverse data science organization for a diverse world

By Kavita Sangwan, Director of Technical Programs, AI and Machine Learning, Intuit

When I graduated in 1998, I was one of only four women in an engineering class of roughly 30 students.

Then and there, I decided that I wanted to do something to change this — to move the needle and broaden our field to include more women, and more people of all kinds who didn’t fit the narrow stereotypes of the time. I’ve made this a priority for the organizations I choose to work with, and I’ve been pleased to see important gains in tech in recent years.

In this blog, I’d like to look at the current state of diversity in data science, why these historic inequities matter, and what we can all do to do better.

In one sense, my current role as Intuit’s director of technical programs, AI, and machine learning represents important progress as a woman in a senior, strategically important position. It also gives our company the opportunity to help advance the cause by participating in programs and events like the Global Women in Data Science (WiDS) Conference, which draws more than 100,000 participants each year through 150 regional events in countries around the world.

But it’s important to remain clear-eyed about the challenges women in data science continue to face.

In a world in which 2.5 quintillion bytes of data are created every day — driving tremendous demand for data science and analytics talent — multiple studies show that women in data science remain woefully underrepresented. For example, a 2018 Burtch Works Study reported the following statistics:

• 15 percent of data scientists are women
• 22 percent of early-career data scientists are women
• 10 percent of data science executive leaders are women

This isn’t just a matter of fairness. Study after study has shown the positive impact of workplace diversity on everything from employee satisfaction and engagement to productivity and profitability. A 2018 Boston Consulting Group study found that companies with more diverse management teams earned a far greater share of their revenue through newly-launched products and services. In other words, diversity fosters innovation. In the fast-changing tech industry, that’s a benefit no business can afford to overlook.

I’m proud to work for a leading fintech company that strives to create a diverse workforce that reflects the demographics of those we serve as a key strategy for creating products that solve important problems for Intuit’s consumer, small business and self-employed customers. As a consequence, women represent 39 percent of our global workforce, including 27 percent in technology positions and 31 percent in executive positions.

I’m especially gratified that this commitment has resulted in far greater representation of women in data science here at Intuit, when compared to data science studies cited earlier in this blog.

Why diversity & inclusion matter in data science

Intuit has invested heavily to embed AI and ML across our product financial software and services portfolio, including QuickBooks, TurboTax and Mint. By making our products smart, we can offer our customers more personalized tools to get more done with less work, save money, and achieve their financial goals.

Delivering relevant, seamless, and high-value experiences requires a deep understanding of customer needs and perspectives. That’s why empathy is the foundation of the Design for Delight (D4D) principles that guide Intuit’s product development (a simplified approach to what the Stanford University d.school calls Design Thinking). It’s also why we place such importance on diversity and inclusion.

The more our organization represents the full breadth of our customer base, the better we’re able to deliver on Intuit’s mission to power prosperity for people of all kinds.

Fairness in itself should be reason enough to embed diversity in our data science teams and processes, but there are important practical considerations as well:

• Initiatives in artificial intelligence and machine learning are fueled by customer data that can only be used with their consent. To gain customer consent from our customers, you have to earn their trust as a responsible steward and advocate for their interests. Policy statements are all well and good, but organizations also need to embody their ethics and values in their workforce, and develop products that reflect a nuanced understanding of the full range of human experience.
• While it matters for every kind of company, diversity is especially important for data science-powered businesses. A model is only as good as its training, and it’s all too easy to build blind spots and biases into the system. Joy Buolamwini has done important work on the social implications of AI and the unintended consequences that can result from a lack of awareness on the part of data scientists.

Buolamwini’s research has found that facial recognition error rates in commercial AI services are 0.8 percent for light-skinned males, but 34.7 percent for dark-skinned females. It’s troubling that such a disparity would have reached the marketplace unnoticed by the data science team — or worse, that the team might not have considered it an issue. Having greater diversity in the room increases the chance that such inequities will be flagged and addressed so that the resulting products can serve people of all kinds equally well.

• We often think of data science skills in traditional terms — computational mathematics, engineering, data analysis, and so on. These are certainly important, but so are the ability to collaborate, communicate, feel empathy, and be a team player. A data scientist’s work can involve complex cross-organizational interactions with product managers, machine learning engineers, data analysts, business personnel, and legal staff, among others. Having a diverse data science organization fosters an environment that places a high value on working with people of all backgrounds and perspectives, with respect and sensitivity.

Discovering a new world for ourselves

Reflecting on my experiences at the Global Women in Data Science Conference earlier this year, I found myself thinking about the little girl I once was — and the countless little girls around the world today whose futures lie ahead of them.

My advice to these children, and to anyone else seeking a direction for their life and career: experiment and explore. Only by trying a variety of different things can you find out where your passion lies. Don’t be afraid to throw yourself into new situations — it helps you become bold, gives you confidence, and can help change the world around you. It’s our responsibility as adults and industry leaders to ensure equal access to opportunities.

What you do with those opportunities is up to you. And I can’t wait to see it.

This has been an Intuit contributed blog for The Hive Think Tank, Medium.