Swartout, W. 1983. XPLAIN: A system for Creating and Explaining Expert Consulting Programs. Artificial Intelligence. 21:285-325.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....As the power of computers is extended through the use of artificial intelligence (AI) people will come to rely on computers for advice in a wide range of expert domains. Experience indicates that it is not enough for expert programs to arrive at an answer; they must also provide an explanation [ Swartout, 1983 ] Designers of complex computer systems cannot expect the diverse user community to be familiar with various coding formats or representations used internally by the software. However, it is reasonable to assume that users will be able to interpret and understand natural language output. It is ....

W.R. Swartout. XPLAIN: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21(3):285--325, September 1983.

....were in the form of traces of the system s execution. The rules of the system encoded the reasoning process but not the rationale behind the process. The execution trace could not be used to provide justifications of system s actions. To overcome this kind of limitation, Swartout s XPLAIN [3] explicitly represented general problem solving knowledge using domain models. The explanation routines examined these models to produce a general description of the system s reasoning. Early knowledge based systems, such as MYCIN and XPLAIN, require that knowledge be expressed at such low levels ....

Swartout WR. XPLAIN: A system for creating and explaining expert consulting programs, Artificial Intelligence, 1983; 21(3), 285--325.

....additional information from the user. For some authors, such as Pearl [41] the best explanation is the most probable assignment of values to a set of variables this kind of explanation is also called abduction [10, 25, 46, 49] Other systems try to o er a simple but comprehensible argument [3, 9, 32, 36, 44, 50, 53, 57, 59]. Finally, in the most sophisticated methods, explanation constitutes an intelligent dialogue in natural language with the system user by way of interactive methods [2, 5, 6, 7, 8, 39] However, in this paper we will only describe explanation methods for Bayesian networks. 2 Properties of ....

W. Swartout. XPLAIN: A system for creating and explaining expert consulting programs. Articial Intelligence, 21:285325, 1983.

....identified the ability of an expert system to explain its reasoning strategies and results to users as an important factor in its acceptance. Researchers have constructed systems that give explanations of logical reasoning for applications spanning the range from blocks world [158] to medicine [136, 12, 143, 153, 112]. Unfortunately, relatively little has been done on the explanation of decision theoretic inference. We shall review some of the ongoing research on techniques for justifying the results of decision theoretic reasoning strategies. Evidence Weights One approach to explaining probabilistic ....

W. Swartout. XPLAIN: A system for creating and explaining expert consulting systems. Artificial Intelligence, 21:285--325, September 1983.

....generation system requires sophisticated semantic information and extensive domain knowledge, which is impossible to provide by a summarizer which works in a general domain and relies on text analysis tools to extract information. All traditional generation systems work in a specific domain [Swartout, 1983, McKeown, 1985, McCoy, 1986, Meteer et al. 1987, Miller and Rennels, 1988, Meteer, 1989, Elhadad, 1992, Robin, 1994, McKeown et al. 1994] Since there are usually limited types of messages in a specific domain, generation systems can rely on the deep understanding of the semantic input and ....

Swartout, W. (1983). Xplain: a system for creating and explaining expert consulting systems. Artificial Intelligence, 21(3):285--325.

....and a reactive approach to explanation is not sufficient (Moore and Swartout, 1990) The following step was to tailor explanation to users needs. Among the various interesting attempts, there are: 1) Adapting the degree of details to the users knowledge (Wallis Shortliffe, 1984; Swartout, 1983); 2) Adapting the explanation types to the users knowledge through a user model (Paris, 1990; Moore and Swartout, 1990) 3) Adapting the explanation types to the users goals (McKeown et al. 1985; Van Beek, 1987) In most of these attempts, explanations are conceived as texts that have to be ....

Swartout, W.R. (1983) XPLAIN: a system for creating and explaining expert consulting programs, Artificial Intelligence, 21(3), 285-325.

....domain. These justifications, the structural knowledge (for instance hierarchic relations between rules) and the strategic knowledge (e.g. If there are non usual causes, treat them ) need to be made explicit for improving the possibility of understanding and modification of a system [see also Swartout, 1983; Hasling et al. 1984; Chandrasekaran et al. 1989; Wick Thompson, 1989] 2.2 Tailored explanations Three issues must be considered in building a system that is able to tailor its explanations: What must we tailored Which is the information to use and how to get it Which tailoring technique ....

....to each rule and concept in the knowledge base and a measure of importance of each concept. User s needs and the appropriate concepts or rules are then matched. XPLAIN tailors its answers to users by varying the number of steps included in an explanation depending on the type of user [Swartout, 1983]. Viewpoint markers are attached to the steps in a prototype method, indicating which type of user the step should be explained to. Adapting the explanation types to the users knowledge . TAILOR adapts the explanations according to a user model with stereotypes and individual features [Paris, ....

Swartout W.R. (1983) XPLAIN: a System for Creating and Explaining Expert Consulting Programs. Artificial Intelligence, 21(3), 285-325.

....in terms of symptoms. ffl Learning of and from explanations. Experts and end users require a system to give adequate explanations of its problem solving behavior. Therefore explanations are a major goal in knowledge engineering. Swartout and his colleagues (e.g. Neches et al. 1985, Swartout, 1983] take explanations (and requirements for explanations) as the prime input to the knowledge acquisition process. Porter and Murray [Murray and Porter, 1988] view a system s attempt to produce a sufficiently strong explanation for a new piece of information as the basic method for learning by ....

Swartout, W. (1983). XPLAIN: A system for creating and explaining expert consulting systems. Artificial Intelligence, 20, 285--325.

....used but turned out not to be relevant, and why the system did not reach a conclusion that the questioner might have considered plausible. Our intuition has always been strong that such abilities of programs to explain and justify their behavior would turn out to be critical to their acceptance [1, 27], and more recent surveys of physicians attitudes support this intuition [30] 4.2 Frame Matching For diagnostic reasoning, the simplest heuristic is: it is what it looks like. Figure 11 suggests what happens when a description of the facts about a case is overlaid onto a description of an ....

William R. Swartout. Xplain: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21:285--325, 1983.

....when told that chloroplasts contain chlorophyll, the learner is able to explain why leaves are green and can perform photosynthesis [Mur90] Previously, these beliefs were known, but the system was unable to explain them. The advantages of possessing explanations of beliefs is well recognized [Swa83, SWMB85, SS77] yet learning behaviors that acquire explanations of existing beliefs have not been widely investigated in machine learning research. The REACT model does not commit to a particular set of conditions that trigger adaptation. These are specified as the admissibility criteria, ....

....a system to strengthen an imperfect theory by connecting unexplained rules to the underlying principles and tacit assumptions that justify their use. By identifying underlying principles and assumptions, explanations of rules enable the system to justify and qualify its conclusions to the user [Swa83] to guide knowledge refinement [SWMB85] and, in the case of default reasoning when assumptions are not met, to improve problem solving [SS77] An imperfect theory is strengthened when new information enables previously unsupported rules to be explained. Initially, a novice s knowledge includes ....

[Article contains additional citation context not shown here]

W. Swartout. XPLAIN: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21:285--325, 1983. 329

....includes facilities to determine what information to include. Given that this information does not directly mirror the user s question, the system also needs to determine how to phrase the information in language. Similarly, expert system explanation is another application where it has been shown [129] that a simple translation of the underlying inference trace (as is often done using templates [111] does not produce a satisfactory explanation of the system s reasoning. Finally, in machine translation, where the content of the generated text is determined by parsing the source language, ....

W.R. Swartout. XPLAIN: a system for creating and explaining expert consulting systems. Artificial Intelligence, 3(2):285--325, 1983.

....includes facilities to determine what information to include. Given that this information does not directly mirror the user s question, the system also needs to determine how to phrase the information in language. Similarly, expert system explanation is another application where it has been shown [124] that a simple translation of the underlying inference trace (as is often done using templates [109] does not produce a satisfactory explanation of the system s reasoning. Finally, in machine translation, where the content of the generated text is determined by parsing the source language, ....

W.R. Swartout. XPLAIN: a system for creating and explaining expert consulting systems. Artificial Intelligence, 3(2):285--325, 1983. 55

....and executed for evaluation. The user interface also supports, in addition to the graphical representations, a natural language processor. This allows the user to query the system using English like sentences. A future aim would be to extend this part into a complete explanation system [Dhar,87, Swartout,83 ] Chapter 7 presents a more detailed description of the user interface and its different forms. 4.2.6. Results Presentation The fourth and last module is the output presentation. There has been much work in the past to deal with presentation of data, for example Kiviat graphs [Morris,74] which ....

....or recommend other tools. The user interface also supports, in addition to the graphical representations, a natural language processor. This allows the user to query the system using English like sentences. A future aim would be to extend this part into a complete explanation system [Dhar,87, Swartout,83 ] A topic which is of interest to this work and which is seeing an increase in importance is that of qualitative reasoning [de Kleer,84] Qualitative reasoning, in our context, can be interpreted to be the reaching of qualitative decisions, interpretations from systems equations that are ....

Swartout, W. R., "XPLAIN: a System for Creating and Explaining Expert Consulting Programs," Artificial Intelligence, 21, pp. 285-325, (1983). 208

....as those the reporting of stock market behaviour or weather forecasting) numerical input may be available from independent sources. But experience shows that an application that wants to use NLG often needs to be built with this in mind if it is to produce material from which NLG is possible (Swartout 1983; Mellish and Evans 1989) 4.2. What Should the Output be The second problem with evaluating NLG is that of assessing the output. There is no agreed objective criterion for comparing the goodness of texts. Other possible notions like appropriateness and correctness depend on the task ....

Swartout, W. R. (1983) "XPLAIN: A System for Creating and Explaining Expert Consulting Programs", Artificial Intelligence Vol 21.

....efforts, the extent to which the systems have modeled user specific information such as the user s knowledge and the dialogue history has increased greatly over time. This section briefly overviews early explanation generation systems in order to illustrate this progression. 2.1. 1 XPLAIN Williams Swartout gave his XPLAIN system (Swartout, 1983) the task of explaining how an expert consulting system arrived at conclusions or why it asked the user certain questions. Its primary goal was to allow a user to understand the reasoning behind an expert system s actions in order to ensure that the user had faith in the recommendations made by ....

William R. Swartout. 1983. XPLAIN: A system for creating and explaining expert consulting systems. Artificial Intelligence, 21(3):285--325.

....and pose questions about specific segments of the document to determine how that region of text follows from the case facts. Results from key projects in the expert systems community suggest that explanation is a key functionality for successful deployment and adaptation for everyday practice [Swa83]. Users are typically unwilling to accept the conclusions drawn by expert systems unless the systems are able to justify their reasoning. Similarly, we believe that for many classes of documents, particularly those that are the product of relatively complex legal reasoning, attorneys will be much ....

William R. Swartout. XPLAIN: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21:285--325, 1983.

....state that for learning tasks where comprehensibility is not crucial, voting methods are extremely useful. However, as many authors have pointed out, problem domains, such as credit approval and medical diagnosis, do require interpretable as well as accurate classification methods. For instance, Swartout [1983] commented that trust in a system is developed not only by the quality of the results but also by clear description of how they were derived. In addition to providing diagnoses or prescriptions, a consultant program must be able to explain what it is doing and why it is doing it. In this ....

Swartout, W. [1983] XPLAIN: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21, 285-325.

....plans. The different perspectives additionally provide a finer grained analysis than their two part separation of a person s understanding into situation and problem (mental) models. Aside from systems in the software domain, perspectives have played a role in explanation in expert systems [Swartout 83, BatemanParis 89] and intelligent tutoring systems [Souther et al. 89] The role of perspectives in explaining results of expert system deductions has focused on providing alternative texts depending on the level of sophistication of the user and not, as in EXPLAINER, on the need to interpret the ....

W. Swartout. XPLAIN: A System for Creating and Explaining Expert Consulting Programs. Artificial Intelligence, 21:285--325, 1983.

....(3) 4 THE EXPLANATIONS 4. 1 GENERAL INTRODUCTION The first researches in the field of explanations were closely tied to knowledge based systems, Clancey used the system MYCIN [4] to construct the explicative systems GUIDON [3] then NEOMYCIN [2] Similarly, Swartout developed the system XPLAIN [10] based on a system of prescription of digitalis: Therapy Digitalis 1 2 3 4 5 6 7 C A B Figure 2: A Configuration of Goban Advisor. These systems are based on the explanation of knowledge and particularly the decomposition of the task into sub tasks. But our approach is different and maybe ....

W. R. Swartout (1983). XPLAIN: a System for Creating and Explaining Expert Consulting Programs. Artificial Intelligence 21(3).

....5. Techniques and Program Tracing In the field of explanation for knowledge based systems it is commonly argued that traditional execution traces, whilst perhaps adequate to explain what a program does, are inadequate to justify to users that such behaviour is reasonable. Systems such as XPLAIN (Swartout, 1983) attempt to remedy this deficiency by constructing the initial program within an environment which forces the system designer to add justifications of design choices when developing the program. This extra information may then be used to justify, in terms of the programmer s design plan, why a ....

Swartout, W.R. 1983. XPLAIN: A System for Creating and Explaining Expert Consulting Systems.

....form directly supported by the language. Unfortunately, different users may invent different ad hoc mechanisms for different applications, even when the target knowledge and inferences are identical. An example of this difficulty is provided by the nikl reimplementations of XPLAIN and ABEL. XPLAIN [40] generated causal explanations of digitalis therapies, and ABEL [33] generated causal explanations of acid base electrolyte disorders. Both systems were originally implemented using XLMS [14] Each used the same notion of causality in its knowledge base, and the original knowledge bases were ....

W. R. Swartout. Xplain: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21:285--325, 1983.

....for performing gait analysis. 7 Other Approaches There have been several other efforts to integrate associational and qualitative models, which can be broadly divided into three classes. The first approach compiles qualitative models into associational models (Chandrasekaran and Mittal, 1983; Swartout, 1983). Because runtime problem solving relies solely on an associational model, it has the problem of compiling associations for a combinatorial number of situations. The second approach uses an associational model to index into qualitative models, but does not use the associational model for further ....

Swartout, W. R. (1983). XPLAIN: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21(3):285--325.

....included them in the word count. 2 1 Introduction Ideally, a language generation system should select words that its user knows, or at least, can deduce meaning for in context. While this would seem to involve simply substituting a known word for an unknown word (as is done, for example, in [Swartout 83] in many cases avoiding an unknown word requires entirely rephrasing the original sentence. For example, in our domain of equipment maintenance and repair, if the user does not know the word polarity a sentence like Check the polarity. will be rephrased as Make sure the plus on the battery ....

Swartout, W.R. XPLAIN: a system for creating and explaining expert consulting systems. Artificial Intelligence 21(3):285-325, 1983.

.... representations of graphical weather maps (see, for example, Goldberg, Driedgar, and Kittredge [1994] summarise statistical data extracted from a database or spreadsheet (Iordanskaja, Kim, Kittredge, Lavoie, and Polgu ere [1992] describe a chain of reasoning carried out by an expert system (Swartout [1983]) and produce on line technical documentation (Rosner and Stede [1994] 3 Dynamic hypertext: the integration of NLG and hypertext Dynamic hypertext is the name we give to the result of merging nlg techniques with hypertextual information delivery: the nlg system creates documents which are nodes ....

Swartout, W. (1983) XPLAIN: A system for creating and explaining expert consulting systems. Artificial Intelligence 21(3):285--325.

....of the migraine treatment migraine pharmacological treatment migraine trigger factor modification migraine abortive tretment dietary control . INDERAL CAFERGOT ELAVIL . migraine prophylactic treatment Figure 4: System s Model of User s Goals response. For example, XPLAIN [18] is a medical expert system capable of justifying its recommendations regarding the dosage of the drug Digitalis. By keeping track of the last few justifications produced and referring to an explicit representation of causal knowledge in the domain, XPLAIN is able to suggest simple analogies with ....

Swartout, W. R. XPLAIN: A system for creating and explaining expert consulting systems. Artificial Intelligence 21, 3 (September 1983), 285--325.

....system to select views, and (b) translate the views to natural language (Figure 4) it would be well on its way to producing coherent explanations. As a building block for the Knight explanation system, we designed and implemented a robust KB accessing system (Figure 5) that extracts views [5, 66, 57, 28, 50, 44, 64, 3, 1, 39, 59, 63] of concepts represented in a knowledge base. Each view is a coherent subgraph of the knowledge base describing the structure and function of objects, the change made to objects by processes, and the temporal attributes and temporal decompositions of processes. Each of the nine accessors in our ....

....a user s current goal from his most recent utterances and uses this goal to select a hierarchy from the multiple hierarchy knowledge base. The selected view controls the content of the explanation and the reasoning that produced that content. In a similar vein, viewpoints in Swartout s Xplain [64] are annotations that indicate when to include a piece of knowledge in an explanation. It is preferable to construct (i.e. extract) views at runtime rather than encoding them in a knowledge base. If a KB accessing system could dynamically construct views, the discourseknowledge engineer would be ....

W. R. Swartout. XPLAIN: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21:285--325, 1983.

....they avoid a direct implementation and that they support the reusability of model components. They di#er in the perspective of model development. Generalizable subtasks (generic task architecture [ Chandrasekaran, 1988 ] explainability of the problem solving process (explainable expert systems [ Swartout, 1983 ] and the distinction of di#erent knowledge types (KADS [ Wielinga et al. 1992 ] are three well known approaches. 2 3 OBJECTIVES Current knowledge based systems, however, lack a series of abilities usually attributed to human experts: improving problem solving capabilities with an ....

W.R. Swartout. XPLAIN: A system for creating and explaining expert consulting systems. Artificial Intelligence, 20:285--325, 1983.

....three problems with 1 We use the syntax of LOOM knowledge representation system[6] to define concepts and relations in this paper. rule based systems that critics have identified as hindering system maintenance and limiting the ability to generate high quality explanations and justifications [8, 9]. First, rules fail to explicitly separate different kinds of knowledge; different clauses in the same rule may implicitly serve to represent contexts, affect control, or capture structural knowledge [10, 11] Because the intent behind them is unclear, it is hard to explain rules and difficult to ....

W. Swartout, "XPLAIN: A system for creating and explaining expert consulting systems," Artificial Intelligence, vol. 21, no. 3, pp. 285--325, September 1983.

....Swartout and Moore s terms, it may handle the problem of inadequacy of low level rules with abstraction, but it still may not distinguish different roles of knowledge and it still may have insufficient knowledge in terms of justifications or background terminology. The work beginning with xplain [115] and continued in the Explainable Expert System (ees) 96] addresses the last two issues by containing a terminology and capturing design information. There is definitional knowledge about the domain and some factual information about a particular event (both included in xplain as factual ....

W. R. Swartout. XPLAIN: A system for creating and explaining expert consulting systems. In Artificial Intelligence, 21(3):285--325, September, 1983.

.... Some of the major issues pertaining to comprehensibility that we have considered in our work to date are the following: ffl Concept descriptions and explanations: Much research on the topic of comprehensibility in AI has focused on generating explanations of a system s behavior (Shortliffe, 1976; Swartout, 1983; Paris, 1987; Moore Swartout, 1989) The explanation task in this body of work has involved both describing why a given instance is classified the way that it is, and describing the defining criteria for a particular concept. In contrast to providing explanations of individual instances, our ....

Swartout, W. (1983). XPLAIN: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21(3):285--325.

.... and the application of machine learning technology [Vla91a] Two learning strategies have been adapted to the context of requirements acquisition: learning by instruction, where the learner conducts the acquisition process by using meta knowledge about the kind of knowledge to be acquired [Dav82] [Swa83], Ben85] and learning by analogy, where the learner retrieves knowledge about some similar system to map it to the system being learned [Hal89] The overall approach taken in KAOS has three components: i) a conceptual model for acquiring and structuring requirements models, with an associated ....

....( libdb: LibraryDatabase, lib: Library) Representation(libdb, lib) libdb.available = lib.available libdb.checkedOut = lib.checkedOut libdb.lost = lib.lost ] For more details, see [Vla91c] 3. A Goal directed Acquisition Strategy In a learning by instruction framework ( Dav82] [Swa83], Ben85] requirements about the composite system are acquired as domain specific instances of elements of the conceptual metamodel. Such instances must satisfy the meta constraints specified once for all like, e.g. the cardinality constraints on meta relationships (see Figure 2) or the ....

W. Swartout, "XPLAIN: A System for Creating and Explaining Expert Consulting Programs", Artificial Intelligence, vol. 21, 1983, 285-325. - 36 -

....explanations in that setting (Buchanan and Shortliffe, 1984; Barr and Feigenbaum, 1982; Coyne, 1990; Chandrasakaran, et al. 1989) Users of these early systems found that understanding why the system had reached a particular conclusion or decision was as important as reaching the decision. Swartout (1983) comments that trust in a system is developed not only by the quality of the results but also by clear description of how they were derived. In addition to providing diagnoses or prescriptions, a consultant program must be able to explain what it is doing and why it is doing it. Explanation ....

Swartout, W. (1983). XPLAIN: A System for Creating and Explaining Expert Consulting Programs. Artificial Intelligence, 21, 285--325.

....includes facilities to determine what information to include. Given that this information does not directly mirror the user s question, the system also needs to determine how to phrase the information in language. Similarly, expert system explanation is another application where it has been shown [127] that a simple translation of the underlying inference trace (as is often done using templates [112] does not produce a satisfactory explanation of the system s reasoning. Finally, in machine translation, where the content of the generated text is determined by parsing the source language, ....

W.R. Swartout. XPLAIN: a system for creating and explaining expert consulting systems. Artificial Intelligence, 3(2):285--325, 1983.

....important reasons for interest in causality in the context of intelligent systems. The first is that models that include causal information are natural and in general easier to construct and modify [14, 21] Such models are also easier for the system to explain and for their users to comprehend [3, 25]. The theoretical link between 282 Marek J. Druzdzel structural equations models and directed probabilistic graphs shows how prior theoretical knowledge about a domain, captured in structural equations, can aid construction of BBNs. If we happen to know the mechanism tying a group of variables, ....

William R. Swartout. XPLAIN: a system for creating and explaining expert consulting programs. Artificial Intelligence, 21(3):285--325, September 1983.

....of natural explanations in the domain. But as researchers in expert systems have found, this places an undue burden on knowledge engineers, who are already trying to mediate between the sometimes conflicting demands of writing expert programs that are correct, efficient, and maintainable (Swartout 1983). Second, traversal mechanisms are only suitable for text types whose structure is dictated primarily by domain structure, and therefore they can only be used for certain genres. Even then, they impose stringent constraints on how domain knowledge must be represented. Third, this approach does not ....

Swartout, W. R. (1983, September). XPLAIN: A system for creating and explaining expert consulting systems. Artificial Intelligence 21 (3), 285--325.

....potential errors in her plan, justifications must be included in support of important points. The importance of explanation for enhancing the acceptance of expert systems is well known [86] Techniques for generating explanations based on the knowledge and reasoning process have been described in [57, 60, 61, 85]. It is also understood that explanations should be tailored to the user s current goals [9, 88] The level of explanation currently available by directly accessing TraumAID s knowledgebase is limited to the information needed by the system for its planning and reasoning. Since TraumAID s ....

William Swartout. XPLAIN: A system for creating and explaining expert consulting programs. Research Report, Information Sciences Institute, Marina del Rey, California, 1983.

....model that suggested how to titrate ideal doses given the patient s clinical state, evidence of possible digitoxicity, and urgency of care. This program also served as the vehicle for several programs on explanation and justification. Perhaps the most interesting was Swartout s PhD thesis [Swar83] in which he used an automatic programmer to generate the digitalis advisor from underlying medical knowledge and treatment principles, which then formed the basis for a sophisticated explanation capability that guaranteed consistency with the program s actual operation. Other explanation work ....

. Swartout, W. R. Xplain: A system for creating and explaining expert consulting programs. Artificial Intelligence, 21:285-325, 1983.

....over the past ten years although there has been much research on different approaches to the problem. These can be divided into approaches which concentrate on the development of new forms of representation of the target domain and the system s reasoning (Clancey 1983, Hasling et al. 1984, Swartout 1983) and those which focus on producing a more human like exchange between the system and the user (Weiner 1980, Goguen et al. 1983) although inevitably there is a great deal of overlap between the approaches. Two examples of the latter approach which were influential in the development of my own ....

Swartout, W.R. XPLAIN: a system for creating and explaining expert consulting programs. Artificial Intelligence, 21, 285-325, 1983.

....that the relation between the events is incidental, has been demonstrated in a multitude of situations. Quite understandably, experiments with early decision support systems, such as Mycin, have indicated that rules alone are not sufficient for generating understandable explanations. 5] Swartout [27] proposes equipping programs with a domain model and representation of problem solving control strategies for the purpose of explanation. The need for a domain model rises from the inability of traditional rule based systems to explicitly examine their control strategy and their behavior and their ....

....of the direction of causal relation allows for concluding that manipulation of the weather will have impact on the barometer reading and manipulation of the barometer will have no impact on the weather. Obviously, knowledge of causal interactions will, as noted by Clancey [5] and Swartout [27], enhance the explanations greatly. Graphical structure of a Bayesian belief network model is in itself an important component of an explanation and has been used as a part of the user interface in several practical decision support systems shells. 6 FLEXIBILITY IN THE LEVEL OF REASONING Users of ....

William R. Swartout. XPLAIN: a system for creating and explaining expert consulting programs. Artificial Intelligence, 21(3):285--325, September 1983.

.... that a major source of difficulties in understanding, modifying and augmenting first generation knowledge based systems stemmed from the use of lowlevel knowledge representations that failed to distinguish different kinds of knowledge (see [Chandrasekaran and Mittal, 1982; Clancey, 1983b; Swartout, 1983] In a first generation system, domain facts, problem solving know ledge, and terminological definitions were all expressed in rules. A single rule might mix together clauses concerned with the user interface, the system s problem solving stra tegy and internal record keeping. Because none of ....

....focused on their failure to provide good explanations, these architectural flaws create problems for acquisition as well. A number of second generation expert system frameworks have emerged in recent years (see [Chandrasekaran, 1986; Clancey, 1983a; Hasling et al. 1984; Neches et al., 1985; Swartout, 1983; Swartout et al. 1991; Wielinga and Breuker, 1986] A common theme among these frameworks is that they encourage a more abstract representation of domain knowledge and problem solving knowledge that makes distinctions between different kinds of knowledge explicit. By moving toward ....

W.R. Swartout, Xplain : A system for creating and explaining expert consulting systems. Artificial Intelligence, 21(3):285-325, September 1983. Also available as ISI/RS-83-4.

.... source of difficulties in understanding, modifying and augmenting first generation knowledge based systems stemmed from the use of low level knowledge representations that failed to distinguish different kinds of knowledge (see [Chandrasekaran and Mittal, 1982; Clancey, 1983b; Swartout, 1981; Swartout, 1983] In a first generation system, domain facts, problem solving knowledge, and terminological definitions were all expressed in rules. A single rule might mix together clauses concerned with the user interface, the system s problem solving strategy and internal record keeping. Because none of ....

....can be viewed as inverse processes, and these architectural flaws create problems for both. A number of second generation expert system frameworks that support better explanations have emerged in recent years (see [Chandrasekaran, 1986; Clancey, 1983a; Hasling et al. 1984; Neches et al., 1985; Swartout, 1983; Swartout et al. 1991; Wielinga and Breuker, 1986] A common theme among these frameworks is that they encourage a more abstract representation of domain knowledge and problem solving knowledge that makes distinctions between different kinds of knowledge explicit. By moving toward architectures ....

[Article contains additional citation context not shown here]

W.R. Swartout, Xplain : A system for creating and explaining expert consulting systems. Artificial Intelligence, 21(3):285-325, September 1983. Also available as ISI/RS-83-4.

No context found.

Swartout, W. 1983. XPLAIN: A system for Creating and Explaining Expert Consulting Programs. Artificial Intelligence. 21:285-325.

No context found.

William R. Swartout. XPLAIN: a system for creating and explaining expert consulting programs. Artificial Intelligence, 21(3):285--325, September 1983.

No context found.

Swartout W.R., XPLAIN: a system for creating and explaining expert consulting programs, Artificial Intelligence, 1983, 21, pp. 285-325.

No context found.

William R. Swartout. XPLAIN:A system for creating and explaining expert consulting programs. 44 Artificial Intelligence, 21(3):285--325, September 1983. Also appears as USC Information Sciences Institute Tech Report RS-83-4.

No context found.

Swar83. Swartout, W.R. XPLAIN: a System for Creating and Explaining Expert Consulting Programs. Artificial Intelligence , 21 (1983), 285-325.

No context found.

W.R. Swartout, XPLAIN: a System for Creating and Explaining Expert Consulting Programs. Artificial Intelligence , 21/3 (1983), 285-325.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC