This interview is part of our AI in Your Community series.
Emily Mower Provost
Human emotion affects government policies, shapes culture, and even biases technology. Simultaneously, it is one of the most challenging variables to research and quantify. In my latest interview in our AI in Your Community series, I spoke with Assistant Professor in the University of Michigan Computer Science and Engineering Department, Emily Mower Provost. Emily’s work uses machine learning and signal processing to look at how people communicate — particularly, through speech — in order to understand human behavior. Programming methods to recognize mood for individuals with bipolar disorder is just one example of her work that not only provides insights into human behavior, but tools that people can use to give themselves ownership over their wellness.
In our conversation, Emily describes the problems worth solving and persevering through and where AI fits into the equation.
Tara Chklovski: What inspires you?
Emily Mower Provost: The idea that we can use technology to understand people. Technology gives us new ways to learn about how humans communicate. It allows us to tackle the unknown. Advances in technology allow us to focus on problems that were previously deemed too difficult to handle. For example, when we talk about emotion, maybe someone’s a little bit angry, but also a little bit frustrated, sad, or surprised; there is an ambiguity. Ambiguity leads innovation.
TC: Can you give an example of how technology is addressing some of these ambiguous problems?
EMP: Here at the Depression Center at the University of Michigan, we’ve been really interested in designing technology for people with bipolar disorder. The goal is to provide warnings if somebody is about to transition into a state that might be unhealthy, because the disorder is characterized by uncomfortable highs of both ecstatic and low energy. The challenge is that everyone expresses their illness differently — that’s where the ambiguity enters in. So, we’re trying to figure out how we can design personalized technology that will give people a way of tracking this but that is tied to their own symptomatology.
TC: Where would we see this technology?
EMP: On individual cell phones. So, we would record speech as people go about their daily lives. The way we see this working is that the technology’s artificial intelligence would act as a real-time monitor for predicting states of mania and depression that won’t require people to do anything; it would be totally passive. It will run completely on a person’s phone, allowing them to have more insight into their own health.
TC: Very cool concept. What advice would you give the many children and families who don’t know about AI?
EMP: One of the greatest places they can check out is Curiosity Machine. I suggest also reading all that you can about AI and working on your math skills! Math is incredibly important in this field. Probability and linear algebra allow us to understand the validity of our assumptions and if our systems are improving.
TC: How will AI strengthen communities and societies?
EMP: AI can enable personalized medicine. At Michigan and around the country, an initiative called Precision Health attempts to find healthcare technology that’s more targeted to the patient. One of our frequent challenges is that we don’t understand why people respond to a medication in a certain way or why particular therapies are effective. There’s this fascinating work at the Depression Center where researchers are trying to create stem cells that can stimulate neural activity for individuals with bipolar disorder. If they are able to accomplish this, then AI may help identify types of medication that would be uniquely effective for a single person’s brain chemistry.
For our work, giving people ownership over their own health is hugely important. Allowing people to understand when they’re healthy versus when they’re not, and how to treat their conditions is a meaningful solution to problems of ambiguity in healthcare.
TC: As we are entering the first year of the Curiosity Machine AI Family Challenge, children participants and their families will need to find problems to solve with AI. What advice would you give them about problem identification?
EMP: Try to figure out what is needed. Observe, observe, observe. Look for where there’s a gap, because where there’s a gap, there’s an opportunity to innovate. Once you identify this gap, theorize what’s needed to actually fill it. In addition to observing, talk to people who might be experts in the space you’re observing. It’s easy to design technology that you think your community needs without understanding what the rest of the community actually needs.
Another factor to think about is to invest time in solving a problem that really, really interests you. In engineering, you’re going to find times when your tinkering works well and then other times, your tinkering really doesn’t work well. If the problem you’re trying to solve is not something that you are fundamentally motivated by, then it’s going to be harder to persist through those times when your engineering doesn’t work.
TC: How do you overcome these challenges in your engineering?
EMP: The ambiguity and unknown of engineering new technology is really exciting and motivating. Again, you need to find this intrinsic motivation to pick yourself back up when your attempts at engineering don’t work well. Perseverance is critical.
Two or three years ago, I was frustrated because things we were working on weren’t getting accepted into the places that I hoped they would. I said to myself, “Okay. Why am I even doing this?” So I went to my whiteboard and I wrote “emotion” in a circle right in the middle and started writing lines that were connecting to things that emotion allows us to do. All of a sudden, I realized that my work is not abstract. It has real value. After I read a few papers I thought, “Well, if I’m going to look at psychology, why would these technologies be useful to people in psychology? Why are they useful to people in human-robot interactions?” You need to step back and ask “Why does my work matter?” to reaffirm that your failures are worth it for meaningful engineering.
TC: Your responses are definitely helpful for participants to consider when they are thinking about how to see the world in a different light, augmented by AI, in our AI Family Challenge.
EMP: That’s what is so exciting about right now. All of a sudden, there are changes in computation power, in algorithms, in our understanding of how to work through this data and how to effectively interact with people who might have different backgrounds. Now, amazing collaborative potentials exist that open the door for educating communities about AI and problem solving. Now is the time to starting asking these questions.
