Michael R. Fehling

A decision support system for project portfolio selection

Abstract A decision support system (DSS) is developed to help managers select the most appropriate sequences of plans for product research and development (R&D) projects that have strict constraints on budget, time, and resources. The primary objective of the DSS is to provide an optimal combination of R&D projects. The DSS consists of several subsystems, each of which has a specific function. At the core of the DSS are a cost model, which covers time-cost tradeoff analysis, and a strategic selection algorithm, which, based on dynamic programming, provides an optimal development plan for managing R&D projects. A working board supports an interactive environment between managers and the DSS. A data checking system eliminates inconsistent data and plans in advance. This paper identifies key issues in the arrangement of R&D projects and describes various systems that have been interlinked to make the DSS a success. It also reveals that the DSS can be expanded to a decision support system shell to support similar types of problems.

Dynamic construction and refinement of utility-based categorization models

The actions taken by an automated decision-making agent can be enhanced by including mechanisms that enable the agent to categorize concepts effectively. We pose a utility-based approach to categorization based on the idea that categorization should be carried out in the service of action. The choice of concepts is critical in the effective selection of actions under resource constraints. We propose a decision-theoretic framework for categorization which involves reasoning about alternative categorization models consisting of sets of interrelated concepts at varying levels of abstraction. Categorization models that are too abstract may overlook details that are critical for selecting the most appropriate actions. Categorization models that are too detailed, however, may be too expensive to process and may contain irrelevant information. Categorization models are therefore evaluated on the basis of the expected value of their recommended action, taking into account the resource cost of their evaluation. A knowledge representation scheme, known as probabilistic conceptual networks, has been developed to support the dynamic construction of models at varying levels of abstraction. This scheme combines the formalisms of influence diagrams from decision analysis and inheritance/abstraction hierarchies from AI. We also propose an incremental approach to categorical reasoning. By applying decision-theoretic control of model refinement, a resource-constrained actor iteratively decides between continuing to improve the current level of abstraction in the model, or to act immediately. >

Unified Theories of Cognition: modeling cognitive competence☆

Abstract In his recent text, Unified Theories of Cognition, Allen Newell offers an exciting mixture of theoretical and methodological advice to cognitive scientists on how to begin developing more comprehensive accounts of human problem solving. Newells perspective is at once both exciting and frustrating. His concept of a unified theory of cognition (UTC), and his attempt to illustrate a UTC with his Soar problem solving architecture, is exciting because it suggests how scientists might use the computational methods of cognitive science and artificial intelligence to formulate and explore both broader and deeper aspects of intelligence in people and in machines. Newells perspective is equally frustrating because it dictates a behaviorist methodology for evaluating cognitive models. Newell views a UTC as a simulation of behavior. I explore the surprising similarity of Newells approach to theory to the approaches of classical behaviorists such as John Watson and Edward Chace Tolman. I suggest that Newells behaviorist methodology is incompatible with his commitment to building theories in terms of complex computational systems. I offer a modification to Newells approach in which a UTC provides an architecture in which to explore the nature of competence—the requisite body of knowledge—that underlies an intelligent agents ability to perform tasks in a particular domain. I compare this normative perspective to Newells commitment to performance modeling. I conclude that his key theoretical concepts, such as the problem space hypothesis, knowledge level systems, and intelligence as approximation to the knowledge level are fundamentally competence constructs. I raise specific concerns about the indeterminacy of evaluating a UTC like Soar against performance data. Finally, I suggest that competence modeling more thoroughly exploits the insights of cognitive scientists like Newell and reduces the gap between the aims of cognitive science and artificial intelligence.

A structured, probabilistic representation of action

When agents devise plans for execution in the real world, they face two important forms of uncertainty: they can never have complete knowledge about the state of the world, and they do not have complete control, as the effects of their actions are uncertain. While most classical planning methods avoid explicit uncertainty reasoning, we believe that uncertainty should be explicitly represented and reasoned about. We develop a probabilistic representation for states and actions, based on belief networks. We define conditional belief nets (CBNs) to capture the probabilistic dependency of the effects of an action upon the state of the world. We also use a CBN to represent the intrinsic relationships among entities in the environment, which persist from state to state. We present a simple projection algorithm to construct the belief network of the state succeeding an action, using the environment CBN model to infer indirect effects. We discuss how the qualitative aspects of belief networks and CBNs make them appropriate for the various stages of the problem solving process, from model construction to the design of planning algorithms.

Probabilistic conceptual network: a belief representation scheme for utility-based categorization

Probabilistic conceptual network is a knowledge representation scheme designed for reasoning about concepts and categorical abstractions in utility-based categorization. The scheme combines the formalisms of abstraction and inheritance hierarchies from artificial intelligence, and probabilistic networks from decision analysis. It provides a common framework for representing conceptual knowledge, hierarchical knowledge, and uncertainty. It facilitates dynamic construction of categorization decision models at varying levels of abstraction. The scheme is applied to an automated machining problem for reasoning about the state of the machine at varying levels of abstraction in support of actions for maintaining competitiveness of the plant.

Computationally-optimal real-resource strategies

Managing the cost of deliberation before action in problems where the overall quality of the solution reflects costs incurred and resources consumed in deliberation as well as the cost and benefit of execution, and both the resource consumption in deliberation phase and the costs in deliberation and execution are uncertain and may be described by probability distribution functions, is addressed. A feasible (in terms of resource consumption) strategy that minimizes the expected total cost is termed computationally optimal. For a situation with several independent, uninterruptible methods to solve the problem, a pseudopolynomial-time algorithm that constructs a generate-and-test computationally optimal strategy is developed. This strategy construction problem is shown to be NP-complete, and Bellmans optimality principle is used to solve it efficiently. The results readily extend to the case of multiple resources.<<ETX>>

Cognitive conflict resolution: mediation analysis and strategies

Mediation provides an important and effective approach to conflict resolution. The desire to view mediation as a science rather than an art has made mediation analysis a rich research arena. Despite the fact that mediation is a widespread and useful element of interpersonal problem solving, there remains no coherent theoretical account of how its benefits may be achieved in general. In this paper, we explore fundamental problems in resolving conflict using mediation. We describe our aims in managing these problems using a cognitive mediation approach. The primary discussion focuses on a theoretical account of cognitive decision-making agents, a two-phase mediation process model, a representation of interpersonal conflict, and a framework for managing dialogue and interactions among agents and with the mediator during the mediation process. We then describe how we are assessing the applicability of this approach within an experimental mediation environment and compare it with other widely used approaches in various conflict settings.<<ETX>>

Decision methods for adaptive task-sharing in associate systems

This paper describes some results of research on associate systems: knowledge-based systems that flexibly and adaptively support their human users in carrying out complex, time-dependent problem-solving tasks under uncertainty. Based on principles derived from decision theory and decision analysis, a problem-solving approach is presented which can overcome many of the limitations of traditional expert-systems. This approach implements an explicit model of the human users problem-solving capabilities as an integral element in the overall problem solving architecture. This integrated model, represented as an influence diagram, is the basis for achieving adaptive task sharing behavior between the associate system and the human user. This associate system model has been applied toward ongoing research on a Mars Rover Managers Associate (MRMA). MRMAs role would be to manage a small fleet of robotic rovers on the Martian surface. The paper describes results for a specific scenario where MRMA examines the benefits and costs of consulting human experts on Earth to assist a Mars rover with a complex resource management decision.

IRMA: real-time, distributed process management using Schemer

Distributed production processes make complex demands. A control system with only one kind of solution process cannot readily handle these complex challenges. This sets out a multiprocessing system and a suite of tools to meet these challenges. We begin by setting out our aims in this research. We describe an application to control of emissions from geothermal power plants. Our discussion focuses on IRMA (Intelligent Real-time Monitoring and Assessment)-a toolkit that supports our view of process management as a collection of diverse problem-solving activities, and Schemer, a general-purpose computational architecture for building distributed, real-time knowledge-based systems such as IRMA. We assess our application and compare it with other knowledge-based systems for real-time process management. We discuss a suite of tools in IRMA that are evidently helpful in a process-management toolkit. We conclude that Schemers integrated implementation of multiprocessing and interprocess communication clarifies and simplifies the use of multiple tools.<<ETX>>

Formulating collaborative engineering design using machine learning method and decision theory

Abstract This paper discusses about the formulation of the collaborative design by multiple agents using the technique from machine learning in artificial intelligence and uncertainty reasoning from the decision theory. We introduce a learning technique for concept formation from prior examples (i.e., design precedents) as a method for constructing a design agents own perspectives. Then, a design coordinators activity is formulated decision-theoretically concerning with the selection of design prototypes. The formulation is illustrated using the examples of designing girder of the bridge.