Despite pandemic distancing, summer research for undergrads stays on course

// Materials Science & Engineering, Chemical & Biological Engineering

Photo of Ben Afflerbach

PhD student Ben Afflerbach helped transform the in-person, lab-based REU program into a virtual primer on data science and machine learning research. Credit: Renee Meiller.

Each summer, dozens of undergraduate students from colleges and universities across the nation, including Puerto Rico, come to the University of Wisconsin-Madison campus to learn what it’s like to work full time in a lab via the college’s Research Experience for Undergraduates (REU) programs in various departments. For most of those students, who come from technical schools, small colleges or universities without research opportunities, it gives them a taste of what graduate student life is like and helps them figure out whether grad school is the right next step in their academic careers.

But due to the COVID-19 pandemic, REU programs at UW-Madison pivoted to distance education in summer 2020. Redesigning the courses, however, was no easy task; typically, students work full time in the lab for 10 weeks assisting in research projects that are already underway. For the distance REU, project leaders had to reimagine the experience, finding new ways to engage and mentor students and provide a glimpse of life in the lab.

To that end, organizers combined several UW-Madison REU programs into the Integrated Chemistry, Chemical Engineering, and Materials Research Experience. Directed by Chemical and Biological Engineering Distinguished Faculty Associate Andrew Greenberg, the integrated program “hosted” 30 students divided into three research tracks.

All students participated in an expanded professional development curriculum, including seminars on topics like searching for graduate schools, writing grant proposals and conducting library research. They also gathered virtually to view talks from UW-Madison researchers who discussed their work and how they operate in the lab.

But the heart of the program was the research component. For the materials research track, offered through the Materials Research Science and Engineering Center (MRSEC) at UW-Madison, the task of leading the research element fell to Benjamin Afflerbach, a fifth-year PhD student in the lab of Harvey D. Spangler Professor of Materials Science and Engineering Dane Morgan.

Afflerbach had supervised REU students in person in the summer of 2019, so he had experience with the program. However, mentoring a group of nine undergrads through an online research project based on data science and machine learning was quite different. “We were fully remote, and the research element was scaled back from 40 hours a week to 10 hours a week,” he says. “It would be hard to expect a student with zero research background and knowledge and only remote guidance to work effectively full time.”

Still, Afflerbach wanted to make sure the research project was meaningful. Instead of giving the students on open-ended research problem like he would have in the lab, he modified structured educational modules previously developed with industry partner Citrine Informatics, which develops an AI platform for materials research. The hands-on learning utilized state-of-the-art open-source software for machine learning and interactive Jupyter Notebooks to allow students with little to no programming experience to engage with machine learning software.

Those platforms, Afflerbach says, allowed students to perform a machine learning workflow on a dataset, eventually leading to fully-developed machine learning models. In particular, the students built models to predict properties of materials. “You can say, ‘I’m interested in materials containing aluminum with a certain target band-gap,’” he says. “You can then predict the band gap for a long list of candidates from your machine learning model and select the materials with those properties. Then you can use those candidates to accelerate research or the discovery of new materials.”

The structured lessons and workflow platforms allowed students to operate more independently and partially worked around one of the most difficult aspects of collaborating online. “A lot of times I’ve found students might be programming and encounter a technical bug,” Afflerbach says. “If we were sitting next to each other in lab, I could just look and say, ‘You’re missing something.’ Now, every time takes an email or Slack message. The back and forth takes a lot more energy and effort.”

Hopefully, the REU program will be back to its usual in-person experience in summer 2021, and students who participated in the 2020 session will be given to option to return. Afflerbach, however, thinks redesigning the curriculum was a helpful exercise and it taught him what works and doesn’t work in his own teaching and mentoring.

He’s transferred the module he created to Nanohub, where motivated students interested in machine learning can access them as a cloud resource and educators pivoting to distance education can readily download them for use as a part of their curriculum.

Author: Jason Daley