Imagine asking a machine "can you pass me that?" and having it understand that "that" refers to an object discussed three turns earlier. What seems basic in any AI assistant today was, for decades, an unsolved problem. Barbara Grosz was the first to take it seriously and build a formal theory to understand it. As a graduate student at Berkeley in the early 1970s, she began designing systems that processed dialogue between humans and computers, at a time when almost no one in the field thought that was worth pursuing.
Her most-cited contribution came in 1986: together with Candace Sidner, she published "Attention, Intentions, and the Structure of Discourse" in Computational Linguistics, the paper that founded the field of computational discourse modeling. The central idea was simple but revolutionary: a conversation is not just a list of sentences; it has a structure of purposes, a flow of attention, and a state that shifts with every turn. A few years later, in 1995, she formalized Centering Theory alongside Aravind Joshi and Scott Weinstein, a mathematical framework for explaining how we choose words when referring back to something already mentioned. That work is still at the heart of reference resolution in modern language models.
Grosz spent 38 years at Harvard, where she became the first woman to serve as AAAI President (1993 to 1995), and where she co-founded the Embedded EthiCS program in 2017 alongside philosopher Alison Simmons: instead of teaching computing ethics in standalone courses, they embedded it inside core computer science classes. In 2015 she received the IJCAI Award for Research Excellence, and in 2017 the ACL Lifetime Achievement Award. She has been a member of the National Academy of Engineering since 2008, honors that speak to decades of sustained work.
What I find most valuable about Grosz is not just the theory: it is the consistency. She bet on understanding human language in its real complexity, not in a simplified version, long before that was fashionable. And when she reached institutional power, she used it to open ethics to all her students, not just those who chose that path. Every time you ask Claude something and it remembers what you were talking about before, a little piece of Barbara Grosz's work is operating underneath.
Official links for Barbara Grosz, The pioneer who taught machines to follow a conversation
More figures who shaped AI in AI Legends, or back to the news.