During the first few years of my PhD, my research experience specifically focused on the educational sector. On the student side, I worked with the Virtual Learning Lab project. In this project, I developed machine learning models to predict student affect while they interact with an online mathematics learning platform. Besides aiming to improve accuracy, I have investigated how these models generalize between different groups of students [see EDM 2019]. My other work with student affect is through the EMERGE project led by Dr. June Gruber. In this study, I helped develop and implement an experience sampling method to measure first-year students’ affect and its context over several weeks. This gave me experience with the practical side of designing experiments, managing data collection, and later analysis.
In addition to affect, I studied the context through which students are using the platform in order to understand their motivations; that is, are students logging in because they want to improve their math ability, or does their grade depend on them completing a certain assignment? Finally, I mentored a team working to prospectively predict student performance on quizzes using their activity patterns on the platform. This work draws on theoretical modeling of Item Response Theory as well as time-series analysis and modeling through machine learning [see LAK 2021a]. Through this work, I gained experience managing large and messy datasets. This primarily involved working with interaction log and student self-report data modalities.
My work in education also focused on improving teacher outcomes. I previously worked with the Cyber-Enabled Teacher Discourse project to develop an automated approach to teacher feedback [see LAK 2021b] and professional development. In this project, I developed machine learning models to identify key elements of teacher speech associated with greater student engagement and learning. I integrated these models into an automated framework [see CHI 2020] where teachers can record audio of their teaching and receive feedback in the form of a smartphone application. This project gave me experience working with audio processing as well as natural language processing techniques. This project in particular allowed me to work with an interdisciplinary group of researchers specializing in ELA education, sociology, and signal processing.
More recently, I have pivoted outside of the education (that is, traditional school) sector and am starting to think about learning and cognitive modeling in a broader context. For example, my recent work with the Cyber-Physical Systems: Cognitive Autonomy project involves interpretable modeling of human drivers and their decisions in risky environments [see ITSC 2022]. These models can be used to estimate human situational awareness [see CPHS 2022] and future research can investigate triggering interventions if needed. This type of work not only draws on my experience in the cognitive sciences but has also allowed me to collaborate with researchers specializing in formal control systems analysis.
Current Research Interests
Drawing on this background in educational technology and mental modeling, I am investigating how we can use autonomy to teach people complex skills. Specifically, how can we quickly reskill users to work with the robots of the future and personalize training to their current skill level? My current thesis work is building an intelligent tutoring system for teaching low-level UAV control. A large component of building this system is automatically assessing a user’s current skill [see HRI 2023]. Given a skill estimate, we can incorporate pedogogical theory such as deliberate practice and scheduled review as well as principles of gamification to provide adaptive training.
Beyond the educational questions, I know that earning users’ trust and acceptance is vital for the uptake of whatever technology we develop. In particular, I’m interested in how users interpret automated formative feedback, especially if their memory of their performance in the exercise differs than the feedback they receive. Additionally, how do we train users to understand the capabilities and limitations of autonomous systems? These research questions will draw on theory and practice of Human-Robot Interaction, especially focused user studies and analysis of explainability.