I'm a PhD student in Computer Science at the University of Michigan, where I work at the intersection of human-centered AI, HCI, affective computing, and computational mental health. My research broadly asks how computational systems represent people's experiences and behaviors, how those representations relate to people's own understanding of themselves, and how we can design intelligent systems that are useful and understandable to the people interacting with them.
I graduated from the University of Michigan in 2020 with a B.S. in Computer Science and Biopsychology, Cognition, and Neuroscience. After graduating, I worked as a software engineer at Microsoft, first on Azure and later on a research incubation team developing healthcare technology. I returned to the University of Michigan for graduate school and am now pursuing my PhD in Computer Science. My work combines computational and human-centered approaches. I'm interested in how AI systems can model complex human states, and in how people understand, interpret, and interact with the representations those systems produce, as well as with the systems themselves. I am particularly interested in the intersection of AI and mental health, and in how we can design systems that are useful, understandable, and ethical.
I study how people experience passive mental health monitoring in everyday life. In a recent longitudinal study, I examined how people with bipolar disorder made sense of six months of ambient-audio monitoring, including how they interpreted their data, provided context that monitoring systems could not capture, and navigated questions around privacy, control, and bystanders. This work motivates broader questions about when intelligent systems should involve users in interpreting, correcting, or supplementing computational representations of their experiences.
I also study how speech and affective signals can be used to understand changes in mental health over time. My current work uses longitudinal ambient speech, self-report, and clinical assessments to investigate relationships between everyday behavior, emotion, and symptoms of bipolar disorder. I'm particularly interested in within-person variation, and understanding how models can account for individual differences in behavior and emotion.
Another line of my work examines the relationship between computational representations of emotion and people's own reports of their emotional experiences. My recent work has studied the generalizability of emotion models trained on third-party annotations to self-reported emotion, as well as questions around variability and reliability in emotion annotation.
What You Feel Is Not What They See: On Predicting Self-Reported
Emotion from Third-Party Observer Labels
Yara El-Tawil, Aneesha Sampath, Emily Mower Provost
IEEE International Conference on Acoustics, Speech and Signal
Processing (ICASSP), 2026
Paper
Rethinking Emotion Annotations in the Era of Large Language Models
Minxue Niu, Yara El-Tawil, Amrit Romana,
Emily Mower Provost
IEEE Transactions on Affective Computing, 2025
Paper
The Whole Is Bigger Than the Sum of Its Parts:
Modeling Individual Annotators to Capture Emotional Variability
James Tavernor, Yara El-Tawil, Emily Mower Provost
Interspeech, 2024
Paper