Showing papers on "Abductive reasoning published in 1993"

Approaches to abductive reasoning: an overview

Abstract: Abduction is a form of non-monotonic reasoning that has gained increasing interest in the last few years. The key idea behind it can be represented by the following inference rule $$O = \mathop C\limits_| - N = \mathop P\limits_|^| - O - \mathop C\limits_|^| - .$$ i.e., from an occurrence of ω and the rule “ϕ implies ω”, infer an occurrence of ϕ as aplausible hypothesis or explanation for ω. Thus, in contrast to deduction, abduction is as well as induction a form of “defeasible” inference, i.e., the formulae sanctioned are plausible and submitted to verification. In this paper, a formal description of current approaches is given. The underlying reasoning process is treated independently and divided into two parts. This includes a description ofmethods for hypotheses generation andmethods for finding the best explanations among a set of possible ones. Furthermore, the complexity of the abductive task is surveyed in connection with its relationship to default reasoning. We conclude with the presentation of applications of the discussed approaches focusing on plan recognition and plan generation.

Non-omniscient belief as context-based reasoning

Abstract: This paper describes a general framework for the formalization of monotonic reasoning about belief in a multiagent environment. The agents* beliefs are modeled as logical theories. The reasoning about their beliefs is formalized in still another theory, which we call the theory of the computer. The framework is used to model non-omniscient belief and shown to have many advantages. For instance, it allows for an exhaustive classification of the "basic" forms of non logical omniscience and for their "composi-tion" into the structure of the system modeling multiagent omniscient belief. 1 The approach This paper describes a general framework for the formal-ization of monotonic reasoning about belief in a multia-gent environment. The most common solution is to take a first order (propositional) theory, to extend it using a set of modal operators, and to take as meaning that an agent believes A (see for instance [Halpern and Moses, 1985]). There is only one theory of the world, however this theory proves facts about the agents' beliefs. According to a first interpretation, this theory is taken to model things how they really are. It is therefore a finite (and possibly incomplete) presentation of what is true in the world, and the fact that B i A is a theorem means that it is, in fact, the case that a i believes A. According to another interpretation, this theory is taken to be the perspective that a generic reasoner has of the world. It is therefore a finite presentation of the reasoner's beliefs, and the fact that A is a theorem means that the reasoner believes that believes A. Once one accepts the second interpretation (as we do), a mechanized theory is naturally taken as representing the beliefs of the computer where it is implemented. Moreover, in the case of multiagent belief, a further step is to have, together with the theory of the computer, one theory (at least, see later) for each agent. *Alessandro Cimatti and Kurt Konolige have provided very useful feedback and suggestions. The work at IRST has been done as part of the MAIA project. This paper is a short version of the IRST technical report #9206-03. The theory of the computer plays the same role as the unique theory in the modal logics approach. The agents' theories are the (mental) representations that the computer has of the agents themselves. The computer has beliefs about the beliefs of the …

Monological reason-based logic: a low level integration of rule-based reasoning and case-based reasoning

Abstract: This paper contains an informal introduction to a theory about legal reasoning (reason-based logic) that takes the notion of a reason to be central. Arguing for a conclusion comes down to first collecting the reasons that plead for and against the conclusion, and second weighing them. The paper describes how we can establish the presence of a reason and how we can argue whether the reasons for or the reasons against the conclusion prevail. It also addresses the topic of meta-level reasoning about the use of rules in concrete cases. It is shown how both rule-based reasoning and case-based reasoning are naturally incorporated in the theory of reason-based logic.

Non-monotonic reasoning with logic programming

Abstract: Our purpose is to exhibit a modular systematic method of representing non-monotonic reasoning problems with the Well-Founded Semantics WFS of extended logic programs augmented with eXplicit negation (WFSX), augmented by its Contradiction Removal Semantics (CRSX) when needed. We apply this semantics, and its contradiction removal semantics counterpart, to represent non-monotonic reasoning problems. We show how to cast in the language of logic programs extended with explicit negation such forms of non-monotonic reasoning as defeasible reasoning, abductive reasoning, and hypothetical reasoning and apply them to such different domains of knowledge representation as hierarchies and reasoning about actions. We then abstract a modular systematic method of representing non-monotonic problems in a logic programming semantics comprising two forms of negation avoiding some drawbacks of other proposals, with which we relate our work.

Case-based reasoning: An introduction

Abstract: Case-based reasoning is a method of solving a current problem by studying the solutions to previous, similar problems. This article presents an introduction to the case-based reasoning process, including an example of the creation and consultation use of the case base. Construction tools for case-based reasoning are identified, and key concepts in case-based reasoning are discussed.

GALGO: a genetic algorithm decision support tool for complex uncertain systems modeled with Bayesian belief networks

Abstract: Bayesian belief networks can be used to represent and to reason about complex systems with uncertain, incomplete and conflicting information. Belief networks are graphs encoding and quantifying probabilistic variables. One type of reasoning of interest in diagnosis is called abductive inference (determination of the global most probable system description given the values of any partial subset of variables). In some cases, abductive inference can be performed with exact algorithms using distributed network computations but it is an NP-hard problem and complexity increases drastically with the lxesence of undirected cycles, number of discrete states per variable, and number of variables in the network. This paper describes an approximate method based on genetic algorithms to perform abductive inference in large, multiply connected networks for which complexity is a concern when using most exact methods and for which systematic search methods are not feasible. The theoretical adequacy of the method is discussed and preliminary experimental results are presented.

Logic-based plan recognition for intelligent help systems

Abstract: Intelligent help systems aim at providing optimal help to the users of complex software application systems. In this context plan recognition is essential for a cooperative system behavior in that it allows to predict the user's future actions, to determine suboptimal action sequences or even serves as a basis for user-adapted tutoring or learning components. In this paper a new approach to incremental plan recognition based on a modal temporal logic is described. This logic allows for an abstract representation of plans including control structures such as loops and conditionals which makes it particularly well-suited for the above-mentioned tasks in command-language environments. There are two distinct phases: With a generalized abductive reasoning mechanism the set of valid plan hypotheses is determined in each recognition step. A probabilistic selection, based on Dempster-Shafer Theory, then serves to determine the "best" hypotheses in order to be able to provide help whenever required.

Bottom-up abduction by model generation

Abstract: We investigate two realizations of parallel abductive reasoning systems using the model generation theorem prover MGTP. The first one, called the MGTP + MGTP method, is a co-operative problem-solving architecture in which model generation and consistency checking communicate with each other. There, parallelism is exploited by checking consistencies in parallel. However, since this system consists of two different components, the possibilities for parallelization are limited. In contrast, the other method, called the Skip method, does not separate the inference engine from consistency checking, but realizes both functions in only one MGTP that is used as a generate-and-test mechanism. In this method, multiple models can be kept in distributed memories, thus a great amount of parallelism can be obtained. We also attempt the upside-down meta-interpretation approach for abduction, in which top-down reasoning is simulated by a bottom-up reasoner.

Structural and probabilistic knowledge for abductive reasoning

Abstract: Different ways of representing probabilistic relationships among the attributes of a domain ar examined, and it is shown that the nature of domain relationships used in a representation affects the types of reasoning objectives that can be achieved. Two well-known formalisms for representing the probabilistic among attributes of a domain. These are the dependence tree formalism presented by C.K. Chow and C.N. Liu (1968) and the Bayesian networks methodology presented by J. Pearl (1986). An example is used to illustrate the nature of the relationships and the difference in the types of reasoning performed by these two representations. An abductive type of reasoning objective that requires use of the known qualitative relationships of the domain is demonstrated. A suitable way to represent such qualitative relationships along with the probabilistic knowledge is given, and how an explanation for a set of observed events may be constituted is discussed. An algorithm for learning the qualitative relationships from empirical data using an algorithm based on the minimization of conditional entropy is presented. >

Reasoning: A Practical Guide for Canadian Students

Abstract: Anyone tasked with the joyous labor of teaching informal logic or critical thinking should take a serious look at this text, which fulfills its promise of offering a practical guide. It begins with an anatomy of the reasoning process, then proceeds to various sources for reasoning such as the media, opinion polls, and experimental research. Part I, Claims, gets students comfortable with the notions of claim, assert, and imply, as well as with the indirect communication forms of innuendo and irony. Then a discussion of the credibility of sources paves the way for the next section, Information and the Media. In that section, the discussion on bias offers a crucial distinction between bias defined as an unfair view and bias as a position out of which a person reasons (the latter for example, "It's my bias that people are more equal than not"). Part I ends with sections on news media and offers a news media checklist of over a dozen questions that, as the authors explain, "an active, critical media comsumer will ask and try to answer" (p.57). Part II, Inferences, shows students that this process is central to reasoning, and expands the student's vocabulary and understanding of propositions, premisses, grounds and conclusions, with lively examples about pending summer jobs and compact disk players that students will like. There are two very useful tables. One, called Role Indicator Words, offers terms such as "therefore," "although," and "must," with an account of their logical relation to the whole idea which is being expressed. The other, Assertion Qualifiers, points to adverbial phrases such as "certainly," "maybe," and to parenthetical phrases such as "I suspect," "as far as I know," to alert students that a writer is ascribing a belief that he or she holds towards a proposition. The next section shows students how inferences can have the logical properties of compound propositions, entailment, equivalence, incompatability or consistency, or contradiction. Then Evaluating Inferences offers several handy ideas, such as the notion that support for an

Combining heuristic reasoning with causal reasoning in diagnostic problem solving

Abstract: In this paper we discuss two different approaches for the combination of heuristic and causal reasoning in diagnostic problem solving. In particular, we first present the two-level architecture CHECK which exploits both experiential knowledge and a deeper form of knowledge. While the former is represented by means of a frame-based formalism, the latter is based on a causal network representation. The co-operation of reasoning at the two levels is discussed: the results of the heuristic level are used to focus reasoning at the causal level. Diagnostic problem solving at the causal level has been logically characterized as a form of abductive reasoning. Because of some difficulties of the CHECK approach (mainly regarding the possible lack of consistency of two independently acquired knowledge bases) we investigated an alternative approach, represented by the AID architecture, which mainly relies on a causal representation of knowledge. In AID the abductive formalization of diagnosis plays a major role, and the reasoning process is focused by operational knowledge that is automatically synthesized from the causal model.

Towards a Logical Theory of Practical Reasoning

Abstract: From the very beginning, logicians have counted practical (or action-oriented) reasoning as well as theoretical (or belief-oriented) reasoning as part of their subject. However, despite a tradition that continues to the present, logicians have not produced formalisms that could be considered of any use in designing an agent that needs to act intelligently, or in helping an intelligent agent to evaluate its reasoning about action. In contrast, the decision-theoretic paradigm that grew out of the economic tradition is widely applied in many areas, including AI, and has dominated recent philosophical thinking about practical reasoning, without serious competition from the logical tradition. This lack of progress is largely due to the unavailability of qualitative mechanisms for dealing with the appropriate inference procedures. To handle even the simplest cases of practical reasoning, it is essential to deliver a reasoning mechanism that allows practical conclusions to be nonmonotonic in the agent’s beliefs. (If have decided to drive my car to work rather than to ride my bicycle, I may well want to to withdraw this conclusion on learning that the car has a fiat tire.) And, until recently, the only way to formalize inference procedures with these characteristics has been to use probability functions. Therefore, quantitative utility theory has remained about the only game in town.

An abductive expert system for interpretation of real-time data (of substations)

Abstract: DINT, a data interpretation expert system which can provide an explanation of the real-time operating state of a substation to its operators, is described. DINT is based on a generalized version of the set covering model for diagnostic problem solving. This abductive reasoning model is an object-based, frame-like knowledge representation paradigm. Test cases illustrate the application of the model to the real-time operation of a substation. >

Pleasure in Practical Reasoning

Abstract: The A. argues that practical reasoning is no worse off than theoretical reasoning, as far as the arbitrariness of its premises goes. Philosophers, unless they are skeptics, are generally not worried about theoretical reasoning being ungrounded. The A. will argue that practical reasoning can avail itself of an analog of belief formation underwritten by observational circumstances, so that practical reasoning has no more cause for embarrassment than theoretical reasoning

Obvious abduction

Abstract: Abduction, or the inference to the best explanation, is a pervasive phenomenon, both in science and in everyday life. Recently, there is a growing awareness that key tasks in many different areas of AI can be cast as abduction. This research is motivated by an apparent paradox. On the one hand, most computation models of abduction have proven to be intractable. On the other hand, humans are capable of making certain kinds of abductive inferences in a flash. To explain this paradox, we propose a theory of abduction that purports to cover the kinds of abductive inferences humans make efficiently. We use the term obvious abduction to loosely refer to these kinds of abductive inferences. The main contributions are as follows: Generality. The theory we propose is applicable across several application domains such as diagnosis, plan recognition, and natural language parsing. Such a unified theory will not only facilitate more accurate characterization and understanding of abductive reasoning, but also foster cross fertilization among different applications of abductive reasoning. Efficiency. The complexity of our abduction algorithm is polynomial in the size of the knowledge base, and exponential in the number of observations to be explained. Therefore, abduction is relatively efficient when the number of observations is small. Probabilistic justification. The knowledge representation scheme in obvious abduction allows probabilistic/statical knowledge to be represented. The inference algorithm is able to compute the probability of explanations and the most probable explanation is preferred.

Incorporating background knowledge and structured explananda in abductive reasoning: a framework

Abstract: Abductive reasoning plays a dominant role in a wide variety of cognitive tasks, including diagnosis, language understanding, and learning. The author's work is aimed at developing a computational model of abductive reasoning keeping in view some of the important aspects that cut across specific instances of abductive reasoning. A significant portion of the paper is devoted to the discussion of two of those aspects, namely, those of background knowledge and structured explananda. The former aspect concerns the effect the agent's memory of specific facts has on abductive reasoning. The latter aspect corresponds to dealing with explananda having conceptual structure and contrasts with the approach of taking explananda to be atomic. The paper discusses various issues related to these two aspects and develops an algorithm for abductive reasoning incorporating those aspects. >

Reasoning about Coherent and Cooperative System Responses

Abstract: This paper discusses the planning of system responses in information-seeking dialogues. Many dialogue systems are capable of answering single questions or carrying out dialogues which have fairly fixed structures, but they show little or no capability to continue the dialogue in an intelligent way, if something unexpected takes place. Our system aims to be flexible and overcome the shortcomings in its knowledge base by contextual reasoning that deals with the enablements and requirements for communication. Dialogue is regarded as a negotiation and the most appropriate response in the context is determined by communicative principles that are considered as constraints on cooperative and coherent communication. The prototype system is based on the knowledge base update procedure developed by Guessoum and Lloyd (1990, 1991), and it is a part of the Dialogue Manager in the PLUS system.

Modelling cyclicity and generalized cost-based abduction using linear constraint satisfaction

Abstract: Abductive reasoning (explanation) is a backward-chaining process on a collection of logical rules. Cost-based abduction is a model for abductive reasoning which provides a concrete formulation of the explanation process. Unfortunately, abduction is an NP-hard task. Current approaches for performing abductive reasoning have been based on graph searching heuristics. However, they are very restrictive and still exhibit expected-case exponential growth rates. One particularly stringent restriction can be found in cost-based abduction whereby the knowledge base must be acyclic. The existence of cyclicity results in anamolous behaviour. In this paper, we present an extended model called generalized cost-based abduction for general knowledge bases. We provide two approaches for solving this model by using a recently introduced technique different from graph searching. This technique uses linear constraints to flexibly represent our knowledge. The problem can then be recast into 0-1 integer linear progra...

Fuzzy Logic, Inductive and Analogical Reasoning

Abstract: We consider domains where deductive inference is not immediatly possible and we study how an automated reasoning can yield a decision from a knowledge which is not well clarified We present both fuzzy inductive and analogical reasoning, showing that they can be merged in a fuzzy logic framework

Acquisition and validation of expert knowledge by using causal models

Abstract: In this paper, we present techniques to support knowledge acquisition and validation within the framework of SGES Our approach can be called a causal model-based knowledge acquisition process It applies to a KBS composed of two knowledge bases: the expert knowledge base refering to the heuristic level is represented by production rules and will form the operational knowledge base, and the causal knowledge base composed of causal models When new expert knowledge (a production rule) is acquired, abductive reasoning based on causal models provides justifications which are then analyzed with appropriate criteria These justifications are useful for refining and extending an initial expert knowledge base: they can be used to propose explanations, to comment on rules, to control them, to suggest modifications or other rules Our approach has been applied to the design of a medical diagnostic reasoning system for electromyography Examples in this field are used in the paper

QUIXOTE as a tool for natural language processing

Abstract: The authors developed a language QUIXOTE as a tool to deal with various information in natural language processing. QUIXOTE is a hybrid language of a deductive object-oriented database and constraint logic programming language. The new mechanism of QUIXOTE is a combination of an object-orientation concept such as object identity and the concept of a module that classifies a large knowledge base. In addition, its logical inference system is extended to be able to make restricted abduction. The authors first apply QUIXOTE to the sorted feature structure of constraint-based grammar formalisms. Next, it is shown that QUIXOTE can contribute to the description of situation-based semantics. The authors have implemented a system to make abductive reasoning to clarify hidden information. Also, they resolve the problem of noun phrase reference.

Recognizing Intentions By Fuzzy Abductive Reasoning

Abstract: This paper proposes a method to infer intentions of users of computer systems that have interactive man-machine interface. The main feature of the approach is the employment of fuzzy abduction -- inference procedure to find an appropriate explanation of given fuzzy events. The paper describes a user utterance model expressed by a goal hierarchy whose goals are different le vels of intentions, and introduces fuzzy abduction as a mechanism to infer intentions. Then, it shows an example to recognize user's intentions from his utterance.

Natural properties of abductive hypotheses in three-valued logic

Abstract: This paper shows some interesting properties of Kleene’s threevalued logic in relation to abductive reasoning. A semantical characterization of abductive explanations is proposed, based on the notion of minimal three-valued model. This establishes a relation between the minimization problem in abductive reasoning and three-valued semantics, in the same sense as non-monotonic reasoning deals with minimization in two-valued semantics.

Intelligent Systems: The Weakest Link?

Abstract: This paper investigates the role intelligent systems may play in the design, operation and maintenance of safety critical applications. It questions whether the techniques currently used to construct intelligent knowledge-based systems can produce designs which can meet the performance requirements of such applications and which, in particular, can be trusted. The paper concludes that conventional rule-based systems are essentially ad-hoc and thus not really suitable, however, more sophisticated techniques as embodied in the abductive, qualitative reasoning and inscriptor approaches are seen to point the way to a solution.