Judea Pearl

Judea Pearl's work has transformed artificial intelligence (AI) by creating a representational and computational foundation for the processing of information under uncertainty. Pearl's work went beyond both the logic-based theoretical orientation of AI and its rule-based technology for expert systems. He identified uncertainty as a core problem faced by intelligent systems and developed an algorithmic interpretation of probability theory as an effective foundation for the representation and acquisition of knowledge. Focusing on conditional independence as an organizing principle for capturing structural aspects of probability distributions, Pearl showed how graph theory can be used to characterize conditional independence, and invented message-passing algorithms that exploit graphical structure to perform probabilistic reasoning effectively. This breakthrough has had major impact on a wide variety of fields where the restriction to simplified models had severely limited the scope of probabilistic methods; examples include natural language processing, speech processing, computer vision, robotics, computational biology, and error-control coding. Equally significant is Pearl's work on causal reasoning, where he developed a graph-based calculus of interventions that makes it possible to derive causal knowledge from the combined effects of actions and observations. This work has been transformative within AI and computer science, and has had major impact on allied disciplines of economics, philosophy, psychology, sociology, and statistics.

Judea Pearl realized the overwhelming prevalence of uncertain information in real-world systems and developed a theoretical and algorithmic foundation for Artificial Intelligence based on probability theory. His influence outside AI has also been great. He forged links between computer science and statistics where none existed previously; his models are used to describe everything from the effects of diseases to the likely behavior of terrorists; and his ideas have revolutionized the understanding of causality in statistics, psychology, medicine, and the social sciences.

Pearl has made pioneering contributions in several areas:

• His 1984 book, Heuristics, tied together a broad array of work from several fields concerned with combinatorial problem solving and optimization. It contained much original work, including the use of formal tree models to analyze the relationship of search depth to decision quality; the first theorems relating the quality of heuristics and the complexity of search; and the discovery and analysis of admissible heuristics, which has led to marked improvements in the capability of modern planning software.
• His 1988 book, Probabilistic Reasoning in Intelligent Systems, has been influential in shaping both the theory and the practice of knowledge-based systems. Pearl's methods helped establish probabilistic networks for representing uncertainty in computer systems. His work demonstrated that networks of cause-effect relationships can represent the qualitative aspects of probabilistic knowledge and that most computations can be performed in a distributed fashion within the topology of these networks. The effects of Pearl's innovations are apparent in both academia and industry. His work is among the most cited in the computer science literature and is largely responsible for creating the field of Uncertainty in AI.
• His 2000 book, Causality: Models, Reasoning, and Inference, introduces causal networks, which refine the idea of Bayesian networks by modeling not just the uninterrupted operation of a stochastic system, but also the effects of all possible interventions. His contributions have had a major impact on the way causality is understood and measured in many scientific disciplines, most notably philosophy, psychology, statistics, econometrics, epidemiology, and social science. It is hard to identify a body of AI research that has been as influential on these related disciplines as has been Pearl's work on causality, culminating with his recent theories of causal discovery, counterfactuals, and interventions.