CRA Honors Undergraduate Computing Research

Harvey Mudd College students Priya Donti ’15 and Rowan Zellers ’16 were selected as finalists for the Computing Research Association’s Outstanding Undergraduate Researcher Awards for 2015.

Sponsored this year by Microsoft Research, the prestigious program recognizes undergraduates at North American universities who demonstrate outstanding potential in an area of computing research.

Priya Donti

The CRA recognized Donti for artificial intelligence research conducted under Harvey Mudd Assistant Professor of Computer Science James Boerkoel.

Donti led research on the Productivity and Wellness Pal (PaWPal), a smart-phone based assistant that actively elicits an individual’s constraints, preferences and goals in order to nudge them to behave more constructively. The team surveyed volunteer Harvey Mudd students about their current experiences randomly throughout the day, giving researchers an accurate picture of participants’ moment-to-moment productivity and wellness. They found that student effectiveness at particular activities was correlated with any number of contextual features—such as time of day and location—leading to the hypothesis that PaWPal might offer meaningful guidance on when, where and how students most fruitfully spend their time.

“Between coursework and extracurricular activities, students struggle to balance their productivity with personal wellness,” said Donti. “My research attempts to address this problem by using AI to learn about students’ behavior and inform them as to how they might best use their time. I hope that through further research, we can make PaWPal an intelligent and friendly tool that meaningfully improves students’ health and happiness.”

The research abstract for the project—co-authored with seniors Jacob Rosenbloom and Alex Gruver—was accepted to the Association for the Advancement of Artificial Intelligence Student Abstract and Poster Program. The team will present its findings at the highly selective international conference later this month in Austin, Texas.

Boerkoel, who nominated Donti, calls the recognition well deserved. “Priya is the rare complete package—a natural leader whose curiosity, vision and passion for research point toward her potential to have a successful research career with broad social impact. Her vision fundamentally shifted the trajectory of this research project into one that I believe has the potential to revolutionize the field of socially aware computing.”

A Harvey Mudd President’s Scholar, Donti is currently a staff member in the Harvey Mudd Writing Center and a grutor for the Computer Science Department. She is also a leadership board member and former co-president of Engineers for a Sustainable World/Mudders Organizing for Sustainability Solutions (ESW/MOSS). She formerly tutored for Homework Hotline and co-led Science Bus, a program that brings hands-on science lessons to elementary school students. Donti plans to pursue a career that involves both environmental sustainability and community engagement.

Rowan Zellers

The CRA recognized Zellers for two research projects. Most recently, Zellers conducted research on multimodal machine learning at USC’s Institute for Creative Technologies under Assistant Research Professor Louis-Phillipe Morency. The project involved multimodal sentiment analysis: the automatic classification of an opinion as positive or negative. Viewing YouTube videos of people reviewing movies, the team used vision, audio and speech recognition software to track responses, then split the videos into utterances and trained support vector machine models on them to provide sentiment annotations. Zellers designed a pipeline for expanding the YouTube dataset and worked on moving toward support vector regression, which would allow predicting a value for sentiment polarity, rather than just as positive or negative. With the expanded dataset, the team hopes to accrue sufficient data to rerun the trials and compare the results.

“This kind of sentiment analysis with regression has not yet been done on a multimodal scale,” said Zellers, “so my research provides a step forward for the use of machine learning to understand multimodal communication. In doing so, this technology can be applied everywhere in creating better human-computer interfaces.”

Additionally, the CRA recognized Zellers’ summer 2013 research, conducted under the advisement of professors Jacqueline Dresch and Robert Drewell. This project applied machine learning to a computational biology problem in order to see how transcription factors bind to DNA in fruit flies, with the hopes of better understanding embryonic cell development in fruit flies and, eventually, humans.

“Rowan is a brilliant undergraduate student who already shows great potential for research,” said Morency, who along with Dresch submitted Zellers’ nomination. “He understands new research concepts quickly, brings new ideas to the table and learns from both positive and negative results.”

A 2014 Goldwater Scholarship runner-up, Zellers is a joint major in computer science and mathematics with an interest in machine learning—the study of how computers digest the patterns that underlie massive data sets. Zellers plans to get a PhD in a mathematics or computers science field related to machine learning and artificial intelligence.