How did you establish and sustain this cross-disciplinary collaboration?
I think if you want to establish a diverse research program, then you need to collaborate with professors and researchers in other disciplines.
It’s a challenge to do this. You have your own research program. You’re trying to move that forward. Then how do you bring others into the fold when they have their own research programs?
I have found the best way to do that is to share a doctoral student. Many times I will pay for that doctoral student from a grant, and I will work with that student — let’s say, for example in computer science. That’s how we got this work initiated.
So we funded a doctoral student, and his co-adviser was a computer scientist. And we had weekly meetings. There was probably a two or three year process where we learned the same vocabulary. I think you need to bring on individuals who are willing to go through that process with you.
It is a process of learning their world, understanding how they frame problems. It really changed my thinking, fundamentally. And I’ve started to learn their vocabulary, as well. In meetings now, even if it’s a very technical problem, some of the issues can come from me because I have enough understanding that I can challenge them a little bit. And it’s great to hear my computer science colleague now talk about phonemes and challenges with co-articulation. It’s been a very fun process, but it does take dedication. It’s something that happens over time.
Then the doctoral student moves the research forward incrementally. And every week, we get the team together and we solve problems. So, I’ve found that to be an effective way to do that. To ensure that everybody is invested — when you have a student involved, everyone is invested. Everyone is going to help the student move forward.
How has collaboration with computer scientists changed your research?
The amount of work that’s involved with data reduction for studying physiologic processes is enormous. So it limits the amount of work you can accomplish for a given study. What I’ve been trying to do over the past decade is develop computer tools that can establish a workflow, so that we can up the numbers in our studies.
Every time I approach a problem now, I think: How can we automatize this? How can we make this more efficient so we can have more data and we can have more confidence in our findings?
Now that I’m working with computer science, I see all sorts of avenues and ways to do this. To use algorithms to give you measurement error. To extract variables automatically. Being able to test — the issue is, you have 60 variables that can be addressed, and that’s overwhelming. But now I’m not as overwhelmed by it. Now I understand that’s part of the process. And there are ways to associate those variables and find the ones that are the most meaningful, relative to the problem you’re working on.
I think it removed a lot of barriers for me, in terms of, “Okay, if this doesn’t work, what’s the next step?” Now I understand there’s that iterative process between hypothesis-driven questions and data-driven questions. You always have the next step to go to.