Papers by Peter Haddawy
Proceedings / the ... Annual Symposium on Computer Application [sic] in Medical Care. Symposium on Computer Applications in Medical Care, 1994
We present a system that generates explanations and tutorial problems from the probabilistic info... more We present a system that generates explanations and tutorial problems from the probabilistic information contained in Bayesian belief networks. BANTER is a tool for high-level interaction with any Bayesian network whose nodes can be classified as hypotheses, observations, and diagnostic procedures. Users need no knowledge of Bayesian networks, only familiarity with the particular domain and an elementary understanding of probability. Users can query the knowledge base, identify optimal diagnostic procedures, and request explanations. We describe BANTER's algorithms and illustrate its application to an existing medical model.
Theoretical Computer Science, 1997
We de ne a language for representing context-sensitive probabilistic knowledge. A knowledge base ... more We de ne a language for representing context-sensitive probabilistic knowledge. A knowledge base consists of a set of universally quanti ed probability sentences that include context constraints, which allow inference to be focused on only the relevant portions of the probabilistic knowledge. We provide a declarative semantics for our language. We present a query answering procedure which takes a query Q and a set of evidence E and constructs a Bayesian network to compute P(QjE). The posterior probability is then computed using any of a number of Bayesian network inference algorithms. We use the declarative semantics to prove the query procedure sound and complete. We use concepts from logic programming to justify our approach.
The Decision-Theoretic Video Advisor
We describe ongoing work toward development of a decision-theoretic agent to help users choose vi... more We describe ongoing work toward development of a decision-theoretic agent to help users choose videos based on their preferences. The DIVA (Decision-Theoretic Interactive Video Advisor) system elicits user preferences using a case-based technique. Hard constraints are used ...
Computational Intelligence, 1998
AI planning agents are goal-directed: success is measured in terms of whether or not an input goa... more AI planning agents are goal-directed: success is measured in terms of whether or not an input goal is satis ed, and the agent's computational processes are driven by those goals. A decision-theoretic agent, on the other hand, has no explicit goals| success is measured in terms of its preferences or a utility function that respects those preferences.
Decision-theoretic Refinement Planning Using Inheritance Abstraction
... 1 Introduction Given a probabilistic model of the world and of avail-able actions and a utili... more ... 1 Introduction Given a probabilistic model of the world and of avail-able actions and a utility function representing the planner's objectives we wish to nd the plan that max-imizes expected utility. ... The system rea-sons with a probabilistic temporal world model. ...
This paper discusses techniques for perform ing efficient decision-theoretic planning. We give an... more This paper discusses techniques for perform ing efficient decision-theoretic planning. We give an overview of the DRIPS decision theoretic refinement planning system, which uses abstraction to efficiently identify opti mal plans. We present techniques for au tomatically generating search control infor mation, which can significantly improve the planner's performance. We evaluate the effi ciency of DRIPS both with and without the search control rules on a complex medical planning problem and compare its perfor mance to that of a branch-and-bound deci sion tree algorithm.
Decision-theoretic re nement planning is a new technique for nding optimal courses of action. The... more Decision-theoretic re nement planning is a new technique for nding optimal courses of action. The authors sought to determine whether this technique could identify optimal strategies for medical diagnosis and therapy. An existing model of acute deep venous thrombosis of the lower extremities was encoded for analysis by the drips decision-theoretic re nement planning system. The encoding represented 6,206 possible plans. The drips planner used Arti cial Intelligence techniques to eliminate 5,150 plans (83%) from consideration without examining them explicitly. drips identi ed the ve strategies that minimized cost and mortality. We conclude that decision-theoretic planning is useful for examining large medical decision problems.
Representations for Decision-Theoretic Planning: Utility Functions for Deadline Goals
Issues in Decision-Theoretic Planning: Symbolic Goals and Numeric Utilities

User Modeling and User-adapted Interaction, 2006
Today a great many medical schools have turned to a problem-based learning (PBL) approach to teac... more Today a great many medical schools have turned to a problem-based learning (PBL) approach to teaching as an alternative to traditional didactic medical education to teach clinical-reasoning skills at the early stages of medical education. While PBL has many strengths, effective PBL tutoring is time-intensive and requires the tutor to provide a high degree of personal attention to the students, which is difficult in the current academic environment of increasing demands on faculty time. This paper describes the student modeling approach used in the COMET intelligent tutoring system for collaborative medical PBL. To generate appropriate tutorial actions, COMET uses a model of each student’s clinical reasoning for the problem domain. In addition, since problem solving in group PBL is a collaborative process, COMET uses a group model that enables it to do things like focus the group discussion, promote collaboration, and suggest peer helpers. Bayesian networks are used to model individual student knowledge and activity, as well as that of the group. The validity of the modeling approach has been tested with student models in the areas of head injury, stroke, and heart attack. Receiver operating characteristic (ROC) curve analysis shows that the models are highly accurate in predicting individual student actions. Comparison with human tutors shows that the focus of group activity determined by the model agrees with that suggested by the majority of the human tutors with a high degree of statistical agreement (McNemar test, p = 0.774, Kappa = 0.823).
Since problem solving in group problem-based learning is a collaborative process, modeling indivi... more Since problem solving in group problem-based learning is a collaborative process, modeling individuals and the group is necessary if we wish to develop an intelligent tutoring system that can do things like focus the group discussion, promote collaboration, or suggest peer helpers. We have used Bayesian networks to model individual student knowledge and activity, as well as that of the group. The validity of the approach has been tested with student models in the areas of head injury, stroke and heart attack. Receiver operating characteristic (ROC) curve analysis shows that, the models are highly accurate in predicting individual student actions. Comparison with human tutors shows that group activity determined by the model agrees with that suggested by the majority of the human tutors with a high degree of statistical agreement (McNemar test, p = 0.774, Kappa = 0.823).

This paper describes COMET, a collaborative intelligent tutoring system for medical problem-based... more This paper describes COMET, a collaborative intelligent tutoring system for medical problem-based learning. The system uses Bayesian networks to model individual student knowledge and activity, as well as that of the group. It incorporates a multi-modal interface that integrates text and graphics so as to provide a rich communication channel between the students and the system, as well as among students in the group. Students can sketch directly on medical images, search for medical concepts, and sketch hypotheses on a shared workspace. The prototype system incorporates substantial domain knowledge in the area of head injury diagnosis. A major challenge in building COMET has been to develop algorithms for generating tutoring hints. Tutoring in PBL is particularly challenging since the tutor should provide as little guidance as possible while at the same time not allowing the students to get lost. From studies of PBL sessions at a local medical school, we have identified and implemented eight commonly used hinting strategies. We compared the tutoring hints generated by COMET with those of experienced human tutors. Our results show that COMET's hints agree with the hints of the majority of the human tutors with a high degree of statistical agreement (McNemar test, p = 0.652, Kappa = 0.773).
Artificial Intelligence in Medicine, 2006
6 S. Suebnukarn, P. Haddawy (McNemar test, p = 0.652, k = 0.773). The focus of group activity cho... more 6 S. Suebnukarn, P. Haddawy (McNemar test, p = 0.652, k = 0.773). The focus of group activity chosen by COMET agrees with that chosen by the majority of the human tutors with a high degree of statistical agreement (McNemar test, p = 0.774, k = 0.823). Conclusion: Bayesian network clinical reasoning models can be combined with generic tutoring strategies to successfully emulate human tutor hints in group medical PBL. #

Artificial Intelligence, 1994
This paper proposes and investigates an approach to deduction in probabilistic logic, using as it... more This paper proposes and investigates an approach to deduction in probabilistic logic, using as its medium a language that generalizes the propositional version of Nilsson's probabilistic logic by incorporating conditional probabilities. Unlike many other approaches to deduction in probabilistic logic, this approach is based on inference rules and therefore can produce proofs to explain how conclusions are drawn. We show how these rules can be incorporated into an anytime deduction procedure that proceeds by computing increasingly narrow probability intervals that contain the tightest entailed probability interval. Since the procedure can be stopped at any time to yield partial information concerning the probability range of any entailed sentence, one can make a tradeo between precision and computation time. The deduction method presented here contrasts with other methods whose ability to perform logical reasoning is either limited or requires nding all truth assignments consistent with the given sentences.
Decision-Theoretic Refinement Planning: Principles and Application
Abstract We present a general theory of action abstraction for reducing the complexity of decisio... more Abstract We present a general theory of action abstraction for reducing the complexity of decision-theoretic planning. We develop projection rules for abstract actions and prove our abstraction techniques to be correct. We present a planning algorithm that uses the abstraction theory ...
This paper discusses the problem of abstracting conditional probabilistic actions. We identify tw... more This paper discusses the problem of abstracting conditional probabilistic actions. We identify two distinct types of abstraction: intra-action abstraction and inter-action abstraction. We de ne what it means for the abstraction of an action to be correct and then derive two methods of intra-action abstraction and two methods of inter-action abstraction which are correct according to this criterion. We illustrate the developed techniques by applying them to actions described with the temporal action representation used in the drips decision-theoretic planner and we describe how the planner uses abstraction to reduce the complexity of planning.
Generating Bayesian Networks from Probablity Logic Knowledge Bases
We present a probabilistic logic programming framework that allows the representation of conditio... more We present a probabilistic logic programming framework that allows the representation of conditional probabilities. While conditional probabilities are the most commonly used method for representing uncertainty in probabilistic expert systems, they have been largely neglected by work in quantitative logic programming. We define a fixpoint theory, declarative semantics, and proof procedure for the new class of probabilistic logic programs. Compared to other approaches to quantitative logic programming, we provide a true probabilistic framework with potential applications in probabilistic expert systems and decision support systems. We also discuss the relationship between such programs and Bayesian networks, thus moving toward a unification of two major approaches to automated reasoning.
Thesis Chapters by Peter Haddawy

This thesis addresses the problem of matching buyers and sellers in barter trade exchange e-marke... more This thesis addresses the problem of matching buyers and sellers in barter trade exchange e-marketplaces. A barter trade exchange is a collection of businesses that buy and sell products among themselves. The collection of businesses is viewed as a micro-economy, so that matching is viewed from an economic perspective. An optimal matching seeks to maximize trade volume and to ensure that all companies share in the trade. The matching problem is given a formal representation and an efficient heuristic search algorithm is developed to solve it. The quality of solution of the heuristic search algorithm is evaluated by comparing it to the optimal solution obtained by exhaustive search on a large set of problems. The algorithm is shown to be fast enough to deal with very large real-world problems. The developed technique has the potential to greatly benefit the barter trade exchange industry as the size of trade exchanges grows.
Uploads
Papers by Peter Haddawy
Thesis Chapters by Peter Haddawy