When Fernanda Viégas was in college, it took three years with three different majors before she decided she wanted to study graphic design and art history. And even then, she couldn’t have imagined the job she has today: building artificial intelligence and machine learning with fairness and transparency in mind to help people in their daily lives.
Today Fernanda, who grew up in Rio de Janeiro, Brazil, is a senior researcher at Google. She’s based in London, where she co-leads the global People + AI Research (PAIR) Initiative, which she co-founded with fellow senior research scientist Martin M. Wattenberg and Senior UX Researcher Jess Holbrook, and the Big Picture team. She and her colleagues make sure people at Google think about fairness and values–and putting Google’s AI Principles into practice–when they work on artificial intelligence. Her team recently launched a series of “AI Explorables," a collection of interactive articles to better explain machine learning to everyone.
When she’s not looking into the big questions around emerging technology, she’s also an artist, known for her artistic collaborations with Wattenberg. Their data visualization art is a part of the permanent collection of the Museum of Modern Art in New York.
I recently sat down with Fernanda via Google Meet to talk about her role and the importance of putting people first when it comes to AI.
How would you explain your job to someone who isn't in tech?
As a research scientist, I try to make sure that machine learning (ML) systems can be better understood by people, to help people have the right level of trust in these systems. One of the main ways in which our work makes its way to the public is through the People + AI Guidebook, a set of principles and guidelines for user experience (UX) designers, product managers and engineering teams to create products that are easier to understand from a user’s perspective.
What is a key challenge that you’re focusing on in your research?
My team builds data visualization tools that help people building AI systems to consider issues like fairness proactively, so that their products can work better for more people. Here’s a generic example: Let’s imagine it's time for your coffee break and you use an app that uses machine learning for recommendations of coffee places near you at that moment. Your coffee app provides 10 recommendations for cafes in your area, and they’re all well-rated. From an accuracy perspective, the app performed its job: It offered information on a certain number of cafes near you. But it didn’t account for unintended unfair bias. For example: Did you get recommendations only for large businesses? Did the recommendations include only chain coffee shops? Or did they also include small, locally owned shops? How about places with international styles of coffee that might be nearby?
The tools our team makes help ensure that the recommendations people get aren’t unfairly biased. By making these biases easy to spot with engaging visualizations of the data, we can help identify what might be improved.
What inspired you to join Google?
It’s so interesting to consider this because my story comes out of repeated failures, actually! When I was a student in Brazil, where I was born and grew up, I failed repeatedly in figuring out what I wanted to do. After spending three years studying for different things—chemical engineering, linguistics, education—someone said to me, “You should try to get a scholarship to go to the U.S.” I asked them why I should leave my country to study somewhere when I wasn’t even sure of my major. “That's the thing,” they said. “In the U.S. you can be undecided and change majors.” I loved it!
So I went to the U.S. and by the time I was graduating, I decided I loved design but I didn't want to be a traditional graphic designer for the rest of my life. That’s when I heard about the Media Lab at MIT and ended up doing a master's degree and PhD in data visualization there. That’s what led me to IBM, where I met Martin M. Wattenberg. Martin has been my working partner for 15 years now; we created a startup after IBM and then Google hired us. In joining, I knew it was our chance to work on products that have the possibility of affecting the world and regular people at scale.
Two years ago, we shared our seven AI Principles to guide our work. How do you apply them to your everyday research?
One recent example is from our work with the Google Flights team. They offered users alerts about the “right time to buy tickets,” but users were asking themselves, Hmm, how do I trust this alert? So the designers used our PAIR Guidebook to underscore the importance of AI explainability in their discussions with the engineering team. Together, they redesigned the feature to show users how the price for a flight has changed over the past few months and notify them when prices may go up or won’t get any lower. When it launched, people saw our price history graph and responded very well to it. By using our PAIR Guidebook, the team learned that how you explain your technology can significantly shape the user’s trust in your system.
Historically, ML has been evaluated along the lines of mathematical metrics for accuracy—but that’s not enough. Once systems touch real lives, there’s so much more you have to think about, such as fairness, transparency, bias and explainability—making sure people understand why an algorithm does what it does. These are the challenges that inspire me to stay at Google after more than 10 years.
What’s been one of the most rewarding moments of your career?
Whenever we talk to students and there are women and minorities who are excited about working in tech, that’s incredibly inspiring to me. I want them to know they belong in tech, they have a place here.
Also, working with my team on a Google Doodle about the composer Johann Sebastian Bach last year was so rewarding. It was the very first time Google used AI for a Doodle and it was thrilling to tell my family in Brazil, look, there’s an AI Doodle that uses our tech!
How should aspiring AI thinkers and future technologists prepare for a career in this field?
Try to be deep in your field of interest. If it’s AI, there are so many different aspects to this technology, so try to make sure you learn about them. AI isn’t just about technology. It’s always useful to be looking at the applications of the technology, how it impacts real people in real situations.