Anwesha Das

Anwesha Das (she/her)

Hi! 🙂 I’m a PhD student in Computer Science at Saarland University, supervised by Prof. Vera Demberg, and part of TRR 248 (CPEC). I began my PhD in 2024, but I had been in SaarbrĂĽcken for a while by then, finishing up my Master’s in DSAI, and somewhere between my years as a research assistant and a thesis with Prof. Anna Maria Feit, I happened to discover joy in the process of science, and stayed. Before that, I worked at Deloitte US-India, having completed my undergraduate at St. Xavier’s College, Mumbai.

Otherwise and elsewhere, I can be found skipping between literary fiction and high-fantasy anime and more, which have a page of their own.

Research

I study what makes a machine (any AI system) good to work with. A large part of it, I think, is how the machine communicates with its human partner. My work draws on 1 pragmatics in human–AI interaction and 2 computational models of communication and attention, in the hope of 3 more rational human–AI cooperation.

Much of my research so far has involved working with the Rational Speech Act (RSA) framework, a Bayesian model of pragmatic reasoning. My models extend it to dynamic environments, where a speaker has to plan a whole sequence of messages. I also test them in behavioural experiments and see whether people can recover meaning successfully in time-critical settings.

For now, my machine speakers simply assume a listener’s epistemic state. I would like them to learn from the context and adapt, perhaps from what a person does, or where they look. There are other (some related, some unrelated) directions that I am also interested in, and I would be glad of company in my quest to understand them: attention-aware interfaces, whether everyday AI assistants reason about their listener and vice versa, and what about the experiences of people who a system designed for the average, neurotypical person leaves out.

Are you looking for collaborators?

Yes! I am fascinated by junctions, by crossings where HCI, language, and AI run into one another. If your work touches on pragmatic communication, on attention, or on how people and AI systems come to understand each other, I would love to hear from you at adas [at] lst.uni-saarland.de.

Publications

Abstracts, figures and BibTeX on the publications page

Teaching