This paper describes an explanation-based learning (EBL) system based on a version of Newell, Shaw and Simon's LOGIC-THEORIST (LT). Results of applying this system to propositional calculus problems from Principia Mathematica are compared with results of applying several other versions of the same performance element to these problems. The primary goal of this study is to characterize and analyze differences between not learning, rote learning (LT's original learning method), and EBL. Another aim is to provide base-line characterizations of the performance of a simple problem solver in the context of the Principa problems, in the hope that these problems can be used as a benchmark for testing improved learning methods, just as problems like chess and the eight puzzle have been used as benchmarks in research on search methods.

This paper presents a knowledge-based learning method and reports on case studies in different domains. The method integrates abduction and learning. Abduction provides an improved method for constructing explanations, in comparison with deductive methods traditionally associated with explanation-based learning. The improvement enlarges the set of examples that can be explained so that one can learn from additional examples using explanation-based macro-learning techniques. Abduction also provides a form of knowledge level learning. The importance of abductive learning is shown by case studies involving over a hundred examples taken from diverse domains requiring logical, physical, and psychological knowledge and reasoning. The case studies are relevant to a wide range of pratical tasks including: natural langauge understanding and plan recognition; qualitative physical reasoning and postdiction; diagnosis and signal interpretation; and decision-making under uncertainty. The descriptions of the case studies show how to set up abduction engines in particular domains and how abduction solves particular tasks. They show how to provide and how to represent the relevant knowledge, including both domain knowledge and meta-knowledge relevant to abduction and search control. The description of each case study includes an example, its explanation, and discussions of what is learned by macro-learning and by abductive inference.

This paper describes an explanation-based learning (EBL) system based on a version of Newell, Shaw and Simon's LOGIC-THEORIST (LT). Results of applying this system to propositional calculus problems from Principia Mathematica are compared with results of applying several other versions of the same performance element to these problems. The primary goal of this study is to characterize and analyze differences between not learning, rote learning (LT's original learning method), and EBL. Another aim is to provide base-line characterizations of the performance of a simple problem solver in the context of the Principia problems, in the hope that these problems can be used as a benchmark for testing improved learning methods, just as problems like chess and the eight puzzle have been used as benchmarks in research on search methods.

This chapter describes an inductive learning method that derives logic programs and invents predicates when needed. The basic idea is to form the least common anti-instance (LCA) of selected seed examples. If the LCA is too general it forms the starting poínt of a general-to-specific search which is guided by various constraints on argument dependencies and critica! terms. A distinguishing feature of the method is its ability to introduce new predicates. Predicate invention involves three steps. First, the need for a new predicate is discovered and the arguments of the new predicate are determíned using the same constraints that guide the search. In the second step, instances of the new predicate are abductively inferred. These instances form the input for the last step where the definition of the new predicate is induced by recursively applying the method again. We also outline how such a system could be more tightly integrated with an abductive learning system.

This chapter describes an inductive learning method that derives logic programs and invents predicates when needed. The basic idea is to form the least common anti-instance (LCA) of selected seed examples. If the LCA is too general it forms the starting poínt of a gneral-to-specific search which is guided by various constraints on argument dependencies and critical terms. A distinguishing feature of the method is its ability to introduce new predicates. Predicate invention involves three steps. First, the need for a new predicate is discovered and the arguments of the new predicate are determíned using the same constraints that guide the search. In the second step, instances of the new predicate are abductively inferred. These instances form the input for the last step where the definition of the new predicate is induced by recursively applying the method again. We also outline how such a system could be more tightly integrated with an abductive learning system.

This paper reports on one aspect of a collaborative project exploring applications of AI to Nondestructive Evaluation (NDE). The goal of NDE is to determine whether parts under evaluation are "good" or "defective" without damaging the parts. Piezoelectric probes are often used to produce ultrasonic signals that indicate the existence of cracks and other defects inside solid materials. This paper shows how knowledge-intensive explanatory reasoning can be used to construct an interpretation of the signals, simultaneously classifying the material under evaluation.

Emotions and cognition are inextricably intertwined. Feelings influence thoughts and actions, which in turn give rise to new emotional reactions. We claim that people infer emotional states in others using common-sense psychological theories of the interactions between emotions, cognition, and action. We present a situation calculus theory of emotion elicitation representing knowledge underlying common-sense causal reasoning involving emotions. We show how the theory can be used to construct explanations for emotional states. The method for constructing explanations is based on the notion of abduction. This method has been implemented in a computer program called AbMaL. The results of computational experiments using AbMaL to construct explanations of examples based on cases taken from a diary study of emotions indicate that the abductive approach to explanatory reasoning about emotions offers significant advantages. We found that the majority of the diary study examples cannot be explained using deduction alone, but they can be explained by making abductive inferences. These inferences provide useful information relevant to emotional states.

Emotions and cognition are inextricably intertwined. Feelings influence thoughts and actions, which in turn give rise to new emotional reactions. We claim that people infer emotional states in others using common-sense psychological theories of the interactions between emotions, cognition, and action. We present a situation calculus theory of emotion elicitation representating knowledge underlying common-sense causal reasoning involving emotions. We show how the theory can be used to construct explanations of emotional states. The method for constructing explanations is based on the notion of abduction. This method has been implemented in a computer program called AbMaL. The results of computational experiments using AbMaL to construct explanations of examples based on cases taken from a diary study of emotions indicate that the abductive approach to explanatory reasoning about emotions offers significant advantages. We found that the majority of the diary study examples cannot be explained using deduction alone, but they can be explained by making abductive inferences. These inferences provide useful information relevant to emotional states.

Diagnostic expert systems constructed using traditional knowledge-engineering techniques identify malfunctioning components using rules that associate symptoms with diagnoses. Model-based diagnosis (MBD) systems use models of devices to find faults given observations of abnormal behavior. These approaches to diagnosis are complementary. We consider hybrid diagnosis systems that include both associational and model-based diagnostic components. We present results on explanation-based learning (EBL) methods aimed at improving the performance of hybrid diagnostic problem solvers. We describe two architectures called EBL_IA and EBL(p). EBL_IA is a form fo "learning in advance" that pre-compiles models into associations. At run-time the diagnostic system is purely associational. In EBL(p), the run-time diagnosis system contains associational, MBD, and EBL components. Learned associational rules are preferred but when they are incomplete they may produce too many incorrect diagnoses. When errors cause performance to dip below a give threshold p, EBL(p) activates MBD and explanation-based "learning while doing". We present results of empirical studies comparing MBD without learning versus EBL_IA and EBL(p). The main conclusions are as follows. EBL_IA is superior when it is feasible but it is not feasible for large devices. EBL(p) can speed-up MBD and scale-up to larger devices in situations where perfect accuracy is not required.