If user interfaces are to reap the benefits of natural language interaction, they must be endowed with the properties that make human natural language interaction so effective. Human-human explanation is an inherently incremental and interactive process. New information must be highlighted and related to what has already been presented. In this paper, we describe the explanation component of a medical information-giving system. We describe the architectural features that enable this component to generate subsequent explanations that take into account the context created by its prior utterances.

One application of models of reasoning behavior is to allow a reasoner to introspectively detect and repair failures of its own reasoning process. We address the issues of the transferability of such models versus the specificity of the knowledge in them, the kinds of knowledge needed for self-modeling and how that knowledge is structured, and the evaluation of introspective reasoning systems. We present the ROBBIE system which implements a model of its planning processes to improve the planner in response to reasoning failures. We show how ROBBIE's hierarchical model balances model generality with access to implementation-specific details, and discuss the qualitative and quantitative measures we have used for evaluating its introspective component. Introduction Many motivations underlie current interest in introspective reasoning and learning. From a functional perspective, introspective reasoning has the potential benefit of allowing the reasoner to refine its own reasoning methods...

Case-based planning is the reuse of past successful plans in order to solve new planning problems. This paper presents a survey of case-based planning, in terms of its historical roots, underlying foundations, methods and techniques currently used, limitations, and future trends. Several authors have given overviews on case-based reasoning and specific topics such as case retrieval, case adaptation, and learning. This overview differs in focus. Its aim is to emphasize the case-based approach to planning, its methodological issues, and its relation to classical planning and the other kinds of case-based reasoning. It also provides some reference models. Keywords: Case-based planning, Case-based reasoning, Planning, Plan Retention, Plan Retrieval, Plan Reuse, Plan Revision. Statement of Exclusive Submission: This paper has not been submitted elsewhere in identical or similar form, nor will it be during the first three months after its submission to Artificial Intelligence Review. 1 Contents 1

. A significant problem for case-based reasoning (CBR) systems is determining the features to use in judging case similarity for retrieval. We describe research that addresses the feature selection problem by using introspective reasoning to learn new features for indexing. Our method augments the CBR system with an introspective reasoning component which monitors system performance to detect poor retrievals, identifies features which would lead retrieval of more adaptable cases, and refines the indexing criteria to include the needed features to avoid future failures. We explore the benefit of introspective reasoning by performing empirical tests on the implemented system. These tests examine the effect of introspective index refinement, and the effects of problem order on case and index learning, and show that introspective learning of new index features improves performance across the different problem orders. 1 Introduction Selecting the best set of features to use in indexing a c...

Complex applications in many areas, including scientific computations and business-related web services, are created from collections of components to form computational workflows. In many cases end users have requirements and preferences that depend on how the workflow unfolds, and that cannot be specified beforehand. Workflow editors therefore need to be augmented with intelligent assistance in order to help users in several key aspects of the task, namely: 1) keeping track of detailed constraints across selected components and their connections; 2) accommodating flexibly different strategies to construct workflows; e.g., from general knowledge of necessary tasks, from desired results, or from available data; and 3) taking partial or incomplete descriptions of workflows and understanding the steps needed for their completion. We have developed a system called CAT (Composition Analysis Tool) that analyzes workflows and generates error messages and suggestions in order to help users compose complete and consistent workflows. Our approach combines knowledge bases, which have rich representations of components and constraints, together with planning techniques that can track the relations and constraints among individual components. We have formalized our approach based on AI planning principles, allowing us to formulate claims about the underlying algorithms as well as the resulting workflows.

In plan reuse, refitting is the process of modifying an existing plan to make it applicable to a new problem situation. An efficient refitting strategy needs to be conservative, i.e., it should minimally modify the existing plan to fit it to the new problem situation. In this paper we present techniques for conservative refitting control by utilizing the annotated dependency structures of an existing plan. The dependency structures arc used to select the refitting choices that minimize disturbance to the applicable parts of the existing plan. This localizes the refitting process and minimizes the cost of refitting. We describe how these techniques are incorporated into PRIAR, a framework for flexible plan reuse. 1.

Reasoning from prior cases or abstractions requires that a system identify relevant similarities between the current situation and objects represented in memory. Often, relevance depends upon abstract, thematic, costly-to-infer properties of the situation. Because of the cost of inference, a case retrieval system needs to learn which descriptions are worth inferring, and how costly that inference will be. This paper outlines the properties that make an abstract thematic feature valuable to a case-based reasoner, and recasts the problem of case retrieval into a framework under which a system can explicitly and dynamically reason about the cost of acquiring features relative to their information value. 1 Retrieval, description, and learning For a case-based reasoner to make effective use of recalled prior experiences, it must be able to judge which of its cases are applicable to the current situation. This problem is not new nor is it unique to case-based reasoning: any system that re...

A routine is a habitually repeated performance of some actions. Agents use routines to guide their everyday activities and to enrich their abstract concepts about acts. This dissertation addresses the question of how an agent who is engaged in ordinary, routine activities changes its behavior over time, how the agent's internal representations about the world is affected by its interactions, and what is a good agent architecture for learning routine interactions with the world. In it, I develop a theory that proposes several key processes: (1) automaticity, (2) habituation and skill refinement, (3) abstraction-bychunking, and (4) discovery of new knowledge chunks. The process of automaticity caches the agent's knowledge about actions into a flat stimulus-response data structure that eliminates knowledge of action consequences. The stimulus-response data structure produces a response to environmental stimuli in constant time. The process of habituation and skill refinement uses environm...

A fundamental difficulty faced by groups of agents that work together is how to efficiently coordinate their efforts. This coordination problem is both ubiquitous and challenging, especially in environments where autonomous agents are motivated by personal goals. Previous AI

Guiding the generation of abductive explanations is a difficult problem. Applying casebased reasoning to abductive explanation generation---generating new explanations by retrieving and adapting explanations for prior episodes---offers the benefit of re-using successful explanatory reasoning but raises new issues concerning how to perform similarity assessment to judge the relevance of prior explanations to new situations. Similarity assessment affects two points in the case-based explanation process: deciding which explanations to retrieve and evaluating the retrieved candidates. We address the problem of identifying similar explanations to retrieve by basing that similarity assessment on a categorization of anomaly types. We show that the problem of evaluating retrieved candidate explanations is often impeded by incomplete information about the situation to be explained, and address that problem with a novel similarity assessment method which we call constructive similarity assessme...