Mower Provost talks about getting awards, doing industry research, understanding human behavior – and Star Wars.

Emily Mower Provost was recently named a 2020 Toyota Scholar Award and is one of the most fascinating faculty members we know. Her research uses machine learning to measure aspects of human behavior like mood and emotion to provide interventions for people with a variety of health-related issues. She’s worked on autism, depression, Parkinson’s disease and Huntington’s disease, in addition to mood and emotion in the context of mobility.

 

We sat down (ok, we google hung out) with her recently to learn a little more about her varied research interests, interdisciplinary collaboration, and some of the ways she works with industry. We also learned she’s really into Star Wars, so that’s cool too.

 

What do you like most about your job?

The amazing thing about a faculty job is you get to pursue whatever you think is interesting or important. You can be constantly in a state of inquiry. Not knowing about a domain isn’t a barrier to entry, you just go back to the beginning, do research, and learn what to do next to investigate a new question. There’s also constant communication in academia, which I love.

 

You tie electrical engineering, computer science and healthcare together.  Those may not be obvious bedfellows. How did this path unfold for you?

When I was getting my PhD I wanted to make prosthetic hands. I was really inspired by Luke’s hand in Star Wars. I thought A) Star Wars is amazing and B) that would be really impactful.

My two potential advisors worked in speech processing and human robot interaction. I wasn’t sure what the overlap looked like but found myself interested in human-robot interaction post stroke rehabilitation, where robots were aware ‘what’ humans were doing but not ‘how’. The ‘how’ is just as important. I wanted to bring human sensing into the loop. 

It became clear that it was too large for one thesis. I narrowed it down to emotion but remained interested in the health applications. I worked in autism research as a postdoc. When I came to U-M I had the opportunity to work with the U-M Depression Center in the context of bipolar disorder using speech data. Eight years later I understand the scope of why this is so big and incredibly impactful, and it has been a really interesting engineering challenge.

Fundamentally what drives me is the idea of not only trying to build state of the art algorithms, but through them, to understand more about human behavior.

 

You’ve won an NSF Career grant, Toyota Scholar Award and Intel Research Fellowship, among others. You’ve gotten lots of recognition for your work. What does that mean to you?

The NSF award still makes me happy. You can apply for it three times, and I got it on my third try. We talk about our accolades, but we don’t talk about the sheer quantity of disappointments. It’s important to talk about both. People see the accomplishments but don’t see the other side of it. We get rejection constantly, so getting this recognition is so affirming.

 

You partner with companies in some of your research. Can you talk about that?

There are a variety of ways industry sponsored research plays an important role on campus. One of the obvious ones is the practical way it helps with necessary day to day administrative expenses. The most exciting part, though, is that it asks us to think about how the work we’re doing can apply in a much shorter term to an actual real world product or problem.

That’s the case with Toyota research where we’re looking at how patterns in emotion indicate emerging risk factors for drivers and IBM’s research in emotion and dialogue systems.

It grounds you. These are less esoteric problems. You can look at what the engineering challenges are that help make an end result realizable. It also helps me think about what I’m doing fits into a broader ecosystem.

A lot of my students go into industry after finishing their PhD. Doing industry sponsored research while they’re here helps them understand what research looks like in industry, what types of problems are valued and what kind of solutions are needed. It’s also a great way for students to get internships.

 

You partner with faculty across campus in your research. Can you speak to the interdisciplinary nature of your work?

What drives me is real world problems and knowing that the algorithms I develop make a difference in people’s lives. In an ideal world, these solutions contribute more to our understanding of people, their interactions, and how it relates to our health.

In interdisciplinary research it’s a framework of ‘this is what I value, this is what you value, and how can we come together and talk about things in a way we all advance’. We need to talk to each other so we can come up with a conversation about things so it means something. 

Interdisciplinary collaboration happens well when you really try to listen and through good communication and back and forth iterations, you begin to understand what the other side needs, values and is missing. Then you can design something that no one has designed before.

 

Many students look at industry research and academic research as very different paths. How do the two work together? 

There’s a spectrum. At the far end in industry, the research you are doing goes into a product in the next 1-5 years. Some industry labs are pure research labs, but fewer though, and those still can ask these big open questions. The cool thing about industry research is it will get used by people in the near term. There’s ownership over what is being done, and from an impact standpoint that’s cool also.

The nice thing about academia is that the answer to your research question might be 20 years off. When you get your head around a problem, then you can go back to industry for refinement and what it means from an enterprise standpoint.

It’s great working on industry sponsored research projects because what you’re developing might end up back in a real product. With some of my research, what we’re hoping to do is develop a real algorithm to detect things like road rage. If your car can detect rage, fatigue, or distraction, your car can decide what to do and safely maneuver the vehicle for you.

It’s a nice example of industry and faculty feeding each other.

 

How has your research shifted in light of COVID-19 and stay at home orders?

We’re lucky in that all the work we do is applied machine learning. We can do absolutely all of it online. This situation is just putting us in different places, so it is also changing our ability to see each other, problem solve and work off each other’s energy.

The human side of our work is very different now. Given the work we do, the whole human communication dynamic, we understand. We are still doing group meetings and one on one meetings. The students are doing game nights and paper reading clubs.

 

What else is on your mind?

I want to make sure Toyota knows how grateful I am for the Toyota Scholar Award. I’m incredibly honored, and it is a privilege to be associated with Toyota in this way.

For industry and companies that work with U-M broadly, I appreciate the opportunity to partner and continue to grow. We are lucky we have this kind of support and engagement.