Paul Cohen

Not to be confused with Paul J. Cohen (1934–2007), the mathematician who proved the independence of the Continuum Hypothesis and won the 1966 Fields Medal. This entry covers Paul R. Cohen, the artificial intelligence researcher.

Paul R. Cohen is a professor of computer science whose career spans academic AI research, cognitive science, and government program management. He chaired the Computer Science department at the University of Arizona, later served as founding Dean of the School of Computing and Information at the University of Pittsburgh (2017–2020), and continues as a professor there.

From 2013 to 2017, Cohen worked as a DARPA program manager in the Information Innovation Office, where he designed and led programs including Big Mechanism, Communicating with Computers, and World Modelers — efforts aimed at getting machines to read, reason about, and model complex biological and social systems.

What is Paul Cohen known for in AI?

Cohen's research centers on empirical methods for evaluating and understanding AI systems, an area he helped formalize with his book Empirical Methods for Artificial Intelligence. His broader work addresses cognitive architectures, planning, and how AI systems can represent and reason about the kind of everyday, commonsense knowledge that underlies concepts like Open Mind Common Sense.

He has also worked on modeling complicated systems — from cellular biology to social dynamics — using computational and statistical methods, reflecting an interdisciplinary approach that bridges AI, cognitive science, and domain science.

What else has Paul Cohen been involved in?

Cohen manages the Harold Cohen Trust, which stewards the legacy of the computer artist Harold Cohen and his AARON system, one of the earliest programs for generating original artwork. In 2025, Paul Cohen became CEO of Causerie.AI Inc.

Why does Paul Cohen matter to AI practitioners?

Cohen's emphasis on empirical rigor pushed AI research toward measurable, reproducible evaluation rather than anecdotal demonstration — a discipline that underpins modern benchmarking practices. His DARPA-era programs also helped shape funding and research directions in machine reading and large-scale system modeling, work that remains relevant to teams building AI systems that reason over complex, real-world domains.