Josh Rozner
Graduate Student
[email protected]

James H. Clark Center
318 Campus Drive
Stanford, CA 94305

Short Bio

I am a doctoral candidate at Stanford in the Institute for Computational and Mathematical Engineering, advised by Cory Shain. I received my undergraduate degree from Yale where I strayed from chemistry and physics to computer science and applied math. I’d call myself an engineer by formal training, a humanist by world training, and a linguist by curiosity. I am interested in characterizing the many kinds of processes (human, machine-learned, computational) which we might broadly term inference, reasoning, or computation. These questions are increasingly interesting, relevant, and amenable to study given recent advances in machine-learned capabilities. My present research focuses on compositionality and productive generalization in language, including language acquisition. My approaches are often computational–machines have the nice attribute of being compliant and unprotesting test subjects–but are cognitively motivated, especially by my own foreign language studies, which continue to provide new ideas. I have worked in related areas of causal modeling and causal abstraction, which often inform my work. I sometimes refer to myself as an “accidental linguist,” ending up here as a result of my own adventures in language, with little formal training, much still to learn, and gratitude for the richness of prior work and the intellectual curiosity of the field’s people.