Sofia Hirschmann

About

I’m a PhD student in Computer Science at Northeastern University, advised by Stacy Marsella in the Cognitive Embodied Social Agents Research (CESAR) Lab. My work sits at the intersection of cognitive science, human-AI interaction, and computational modeling, with a focus on how people and intelligent systems coordinate, adapt, and make decisions together.

The question running through all of my research is how an intelligent system can build and maintain an accurate model of the person it’s interacting with, including their goals, capabilities, beliefs, and emotional state, rather than falling back on rigid assumptions. I approach this through the lens of Theory of Mind: giving systems the ability to reason about what people intend and why, so they can adapt to people rather than the other way around.

I explore these ideas across various settings, cooperative and adversarial alike. I’ve studied how robots can coordinate with human teammates by actively learning what their partners are capable of rather than assuming it; how cognitive biases can be modeled to anticipate the decisions of network attackers; how naming our emotions reshapes the choices we make under pressure; and how multi-agent language models can give people a space to practice social skills through improvised role-play. What ties these together is a single goal: systems that reason about people as people.

I also carry this work into practice as a founding engineer at Elevare, a maternal health platform, where I build the app’s core infrastructure and a model of each mother that develops from her engagement and is grounded in clinical evidence, so the system can offer relevant, personalized support and escalate to human providers when a concern arises. More broadly, I want technology that understands people on their own terms and makes us more capable, not less.