leff@smu.UUCP (Laurence Leff) (12/20/88)
Subject: AI-Related Dissertations from SIGART No. 103 (only one file) The following is a list of dissertation titles and abstracts related to Artificial Intelligence taken taken from the Dissertation Abstracts International (DAI) database. The list is assembled by Susanne Humphrey and myself and is published in the SIGART Newsletter (that list doesn't include the abstracts). The dissertation titles and abstracts contained here are published with the permission of University Microfilms International, publishers of the DAI database. University Microfilms has granted permission for this list to be redistributed electronically and for extracts and hardcopies to be made of it, provided that this notice is included and provided that the list is not sold. Copies of the dissertations may be obtained by addressing your request to: University Microfilms International Dissertation Copies Post Office Box 1764 Ann Arbor, Michigan 48106 or by telephoning (toll-free) 1-800-521-3042 (except for Michigan, Hawaii, and Alaska). In Canada: 1-800-268-6090. From SIGART Newsletter No. 103 File 1 of 1 ------------------------------------------------------------------------ AN University Microfilms Order Number ADG87-13824. AU KRAEMER, JAMES RICHARD. IN The University of Oklahoma Ph.D. 1987, 166 pages. TI ADMINISTRATIVE CONTROL BY EXPERT SYSTEM: A FRAMEWORK FOR EXPERT SYSTEMS IN MANAGEMENT. SO DAI V48(03), SecA, pp694. DE Business Administration, Management. AB Administrative Control Expert Systems (ACES) can be developed to control an administrative process in the same way that PERT, CPM, and Gantt charts control a project. The fundamental difference between an administrative process and a project is that tasks to be performed are determined by policy and procedures, not a fixed schedule. Traditional administrative control systems are driven by data and rules, called policy, procedures, objectives, and budgets. Expert systems are, likewise, driven by their data and rule base. The ACES framework makes explicit the implicit similarity between administrative control systems and expert systems. The overall design of the ACES framework and each of its five major subsystems are described. A methodology is presented to allow the integration of an ACES into an existing transaction processing environment. Extensions to the expert system methodology are presented to provide capabilities to specifically support administrators. Security issues for expert systems are discussed, along with the integration of microcomputers into a mainframe ACES environment. A limited example of ACES rule structure, data, and processing is included in an appendix. AN University Microfilms Order Number ADG87-14769. AU OLIVERO, RAMON ALFREDO. IN University of Houston Ph.D. 1987, 272 pages. TI SELECTION OF EXPERIMENTAL DESIGNS FOR ANALYTICAL CHEMISTRY WITH THE AID OF AN EXPERT SYSTEM. SO DAI V48(04), SecB, pp1028. DE Chemistry, Analytical. AB An expert system for selecting statistical experimental designs in analytical chemistry has been designed and implemented. The resulting expert system computer program (DXPERT) uses information about the characteristics and constraints of the analytical chemical system as well as information about the user's interests and resources to assess the suitability of each of thirteen candidate experimental designs. A dedicated "inference engine" was constructed to utilize a knowledge base containing the experience of an expert in the field of statistical experimental design, the knowledge of this writer, and information from the literature. The selection of experimental designs is determined by the answers given by the analytical chemist to the questions posed by the expert system in an interactive consultation session. The questions are presented in an order determined by a criterion of maximum potential information gain. Fuzzy set logic and arithmetic are applied to the knowledge representation and to the calculation of the experimental designs' desirabilities. The program operates on an IBM-PC('TM) or compatible personal computer and is written in Pascal language. It has user-friendly features like "why" explanations, a help facility, reviewing and revision options, and menus. A number of test runs with representative problems were carried out to validate the system and to evaluate its performance. It was found that the system assigned appropriate desirabilities to experimental designs in these test cases, as determined by comparison with the solutions recommended by human experts. AN University Microfilms Order Number ADG87-13877. AU ACKLEY, DAVID HOWARD. IN Carnegie-Mellon University Ph.D. 1987, 238 pages. TI STOCHASTIC ITERATED GENETIC HILLCLIMBING. SO DAI V48(03), SecB, pp808. DE Computer Science. AB In the "black box function optimization" problem, a search strategy is required to find an extremal point of a function without knowing the structure of the function or the range of possible function values. Solving such problems efficiently requires two abilities. On the one hand, a strategy must be capable of learning while searching: It must gather global information about the space and concentrate the search in the most promising regions. On the other hand, a strategy must be capable of sustained exploration: If a search of the most promising region does not uncover a satisfactory point, the strategy must redirect its efforts into other regions of the space. This dissertation describes a connectionist learning machine that produces a search strategy called stochastic iterated genetic hillclimbing (SIGH). Viewed over a short period of time, SIGH displays a coarse-to-fine searching strategy, like simulated annealing and genetic algorithms. However, in SIGH the convergence process is reversible. The connectionist implementation makes it possible to diverge the search after it has converged, and to recover coarse-grained information about the space that was suppressed during convergence. The successful optimization of a complex function by SIGH usually involves a series of such converge/diverge cycles. SIGH can be viewed as a generalization of a genetic algorithm and a stochastic hillclimbing algorithm, in which genetic search discovers starting points for subsequent hillclimbing, and hillclimbing biases the population for subsequent genetic search. Several search stratgies--including SIGH, hillclimbers, genetic algorithms, and simulated annealing--are tested on a set of illustrative functions and on a series of graph partitioning problems. SIGH is competitive with genetic algorithms and simulated annealing in most cases, and markedly superior in a function where the uphill directions usually lead away from the global maximum. In that case, SIGH's ability to pass information from one coarse-to-fine search to the next is crucial. Combinations of genetic and hillclimbing techniques can offer dramatic performance improvements over either technique alone. AN This item is not available from University Microfilms International ADG05-60485. AU BAPA RAO, KOTCHERLAKOTA V. IN University of Southern California Ph.D. 1987. TI AN EXTENSIBLE OBJECT-ORIENTED FRAMEWORK FOR ENGINEERING DESIGN DATABASES. SO DAI V48(04), SecB, pp1095. DE Computer Science. AB This thesis describes DOM (Design Object Model), a model of objects in a database to support computer-aided design of complex artifacts such as VLSI chips and software systems. Most database models are designed with administrative domains in view and hence are ill-suited to cope with the complex structural hierarchies, multiple representations, and incremental evolution at both the object and meta-object (schema) levels that are typical of design objects. DOM aims to provide a generic framework of high-level concepts for representing these aspects of data and meta-data in the design environment. Important considerations in the design of DOM are uniformity of representation, integration of concepts, and the ability to represent design data and the more conventional kinds of data in a common framework. DOM is object-oriented in that it seeks to directly capture the properties of real-world objects; it is extensible in that a DOM database and schema can be incrementally extended to accommodate evolution in the real world. DOM has been developed in two phases. First, the conceptual requirements of design data models are formulated as a set of abstract concepts obtained by analyzing the properties of design environments. These are organized into four dimensions: (1) Static structure describing the design object considered as a static entity; (2) Evolution structure describing the evolutionary stages of design and their relationships; (3) Level denoting whether an object is an extensional object or a schema; (4) Originality denoting whether an object is a design in its own right or an instantiation of a design. In the second phase, the abstract concepts are mapped to a simple object-based data model, and thus articulated as concrete concepts. Simple and compound objects realize the static structure dimension. Generic and realization objects implement evolution structure permitting multiple evolutionary alternatives. Schema objects represent meta-designs, and copy objects represent instantiations. These concepts are extended in the static structure dimension to enable the description of design objects via abstractions such as interface, implementation, and views, the last denoting multiple representations. The application of DOM is demonstrated by modelling the domain of VLSI design objects. (Copies available exclusively from Micrographics Department, Doheny Library, USC, Los Angeles, CA 90089-0182.). AN University Microfilms Order Number ADG87-13885. AU CHRIST, JAMES P. IN Carnegie-Mellon University Ph.D. 1987, 138 pages. TI SHAPE ESTIMATION AND OBJECT RECOGNITION USING SPATIAL PROBABILITY DISTRIBUTIONS. SO DAI V48(03), SecB, pp809. DE Computer Science. AB This thesis describes an algorithm for performing object recognition and shape estimation from sparse sensor data. The algorithm is based on a spatial likelihood map which estimates the probability density for the surface of the object in space. The spatial likelihood map is calculated using an iterative, finite element approach based on a local probabilistic model for the object's surface. This algorithm is particularly useful for problems involving tactile sensor data. An object classification algorithm using the spatial likelihood map was developed and implemented using simulated tactile data. The implementation for the tactile problem was in two dimensions for the sake of clarity and computational speed, and is easily generalized to three dimensions. The spatial likelihood map is also useful for multi-sensor data fusion problems. This is illustrated with an application drawn from the study of mobile robots. AN University Microfilms Order Number ADG87-14481. AU KELLER, RICHARD MICHAEL. IN Rutgers University The State U. of New Jersey (New Brunswick) Ph.D. 1987, 352 pages. TI THE ROLE OF EXPLICIT CONTEXTUAL KNOWLEDGE IN LEARNING CONCEPTS TO IMPROVE PERFORMANCE. SO DAI V48(03), SecB, pp811. DE Computer Science. AB This dissertation addresses some of the difficulties encountered when using artificial intelligence-based, inductive concept learning methods to improve an existing system's performance. The underlying problem is that inductive methods are insensitive to changes in the system being improved by learning. This insensitivity is due to the manner in which contextual knowledge is represented in an inductive system. Contextual knowledge consists of knowledge about the context in which concept learning takes place, including knowledge about the desired form and content of concept descriptions to be learned (target concept knowledge), and knowledge about the system to be improved by learning and the type of improvement desired (performance system knowledge). A considerable amount of contextual knowledge is "compiled" by an inductive system's designers into its data structures and procedures. Unfortunately, in this compiled form, it is difficult for the learning system to modify its contextual knowledge to accommodate changes in the learning context over time. This research investigates the advantages of making contextual knowledge explicit in a concept learning system by representing that knowledge directly, in terms of express declarative structures. The thesis of this research is that aside from facilitating adaptation to change, explicit contextual knowledge can support two additional capabilities not supported in most existing inductive systems. First, using explicit contextual knowledge, a system can learn approximate concept descriptions when necessary or desirable in order to improve performance. Second, with explicit contextual knowledge, a learning system can generate its own concept learning tasks. To investigate the thesis, this study introduces an alternative concept learning framework--the concept operationalization framework--that requires various types of contextual knowledge as explicit inputs. To test this new framework, an existing inductive concept learning system (the LEX system Mitchell et al. 81 ) was rewritten as a concept operationalization system (the MetaLEX system). This document describes the design of MetaLEX and reports the results of several experiments performed to test the system. Results confirm the utility of explicit contextual knowledge, and suggest possible improvements in the representations and methods used by the system. AN University Microfilms Order Number ADG87-14072. AU LANKA, SITARAMASWAMY VENKATA. IN University of Pennsylvania Ph.D. 1987, 148 pages. TI AN AID TO DATABASE DESIGN: AN INDUCTIVE INFERENCE APPROACH. SO DAI V48(03), SecB, pp811. DE Computer Science. AB The conventional approach to the design of databases has the drawback that to specify a database schema, it requires the user to have knowledge about both the domain and the data model. That is, the onus of encoding the domain information in terms of concepts foreign to the domain falls on the user. The goal of this research is to free the user of such burdens. We propose a system that designs a database based on its functional requirements. The user need only provide information on how the database is expected to be used, and the system infers a schema from this. Furthermore, the information is expressed in a language which is independent of the underlying data model. The above problem has been cast as an inductive inference problem. The input is in the form of Natural Language (English) queries and a conceptual database schema is inferred from this. The crux of the inference mechanism is that the hypotheses are synthesized compositionally and this is described in terms of Knuth's attribute grammars. In certain situations the inference mechanism has the potential to synthesize false hypothesis. We have advanced a method to detect these potentially false hypotheses, and refine them to obtain acceptable hypotheses. A prototype of such a system has been implemented on the symbolics Lisp machine. AN University Microfilms Order Number ADG87-12346. AU MICHON, GERARD PHILIPPE. IN University of California, Los Angeles Ph.D. 1983, 125 pages. TI RECURSIVE RANDOM GAMES: A PROBABILISTIC MODEL FOR PERFECT INFORMATION GAMES. SO DAI V48(03), SecB, pp813. DE Computer Science. AB A simple probabilistic model for game trees is described which exhibits features likely to be found in realistic games. The model allows any node to have n offsprings (including n = 0) with probability f(,n) and assigns each terminal node a WIN status with probability p and a LOSS status with probability q = 1 - p. Our model may include infinite game trees and/or games that never end when played perfectly. The statistical properties of games and the computational complexities of various game solving approaches are quantified and compared. A simple analysis of game pathology and quiescence is also given. The model provides a theoretical justification for the observed good behavior of game-playing programs whose search horizon is not rigid. Pathological features that were recently found to be inherent in some former game models are put in a new perspective. AN University Microfilms Order Number ADG87-13872. AU MUELLER, ERIK THOMAS. IN University of California, Los Angeles Ph.D. 1987, 763 pages. TI DAYDREAMING AND COMPUTATION: A COMPUTER MODEL OF EVERYDAY CREATIVITY, LEARNING, AND EMOTIONS IN THE HUMAN STREAM OF THOUGHT. SO DAI V48(03), SecB, pp813. DE Computer Science. AB This dissertation presents a computational theory of daydreaming: the spontaneous human activity--carried out in a stream of thought--of recalling past experiences, imagining alternative versions of past experiences, and imagining possible future experiences. Although on the face of it, daydreaming may seem like a useless distraction from a task being performed, we argue that daydreaming serves several important functions for both humans and computers: (1) learning from imagined experiences, (2) creative problem solving, and (3) a useful interaction with emotions. The theory is implemented within a computer program called DAYDREAMER which models the daydreaming of a human in the domain of interpersonal relations and common everyday occurrences. As input, DAYDREAMER takes English descriptions of external world events. As output, the program produces English descriptions of (1) actions it performs in the external world and (2) its internal "stream of thought" or "daydreams": sequences of events in imaginary past and future worlds. Five major research issues are considered: (1) the generation and incremental modification of realistic and fanciful solutions or daydreams, (2) focusing attention in the presence of multiple active problems, (3) the recognition and exploitation of accidental relationships among problems, (4) the use of previous solutions or daydreams in generating new solutions or daydreams, and (5) the interaction between emotions and daydreaming. DAYDREAMER consists of a collection of processing mechanisms and strategies which address each of the above issues: (1) a planner, a collection of personal goals, daydreaming goals, and planning and inference rules for the domain, and a mutation mechanism; (2) a control mechanism based on emotions as motivation; (3) a serendipity mechanism; (4) an analogical planner which stores, retrieves, and applies solutions or daydreams in a long-term episodic memory; and (5) mechanisms for initiating and modifying emotions during daydreaming and for influencing daydreaming in response to emotions. DAYDREAMER is able to generate a number of daydreams and demonstrate how daydreaming enables learning, creative problem solving, and a useful interaction with emotions. AN University Microfilms Order Number ADG87-14099. AU NADATHUR, GOPALAN. IN University of Pennsylvania Ph.D. 1987, 169 pages. TI A HIGHER-ORDER LOGIC AS THE BASIS FOR LOGIC PROGRAMMING. SO DAI V48(03), SecB, pp813. DE Computer Science. AB The objective of this thesis is to provide a formal basis for higher-order features in the paradigm of logic programming. Towards this end, a non-extensional form of higher-order logic that is based on Church's simple theory of types is used to provide a generalisation to the definite clauses of first-order logic. Specifically, a class of formulas that are called higher-order definite sentences is described. These formulas extend definite clauses by replacing first-order terms by the terms of a typed (lamda)-calculus and by providing for quantification over predicate and function variables. It is shown that these formulas, together with the notion of a proof in the higher-order logic, provide an abstract description of computation that is akin to the one in the first-order case. While the construction of a proof in a higher-order logic is often complicated by the task of finding appropriate substitutions for predicate variables, it is shown that the necessary substitutions for predicate variables can be tightly constrained in the context of higher-order definite sentences. This observation enables the description of a complete theorem-proving procedure for these formulas. The procedure constructs proofs essentially by interweaving higher-order unification with backchaining on implication, and constitutes a generalisation, to the higher-order context, of the well-known SLD-resolution procedure for definite clauses. The results of these investigations are used to describe a logic programming language called (lamda)Prolog. This language contains all the features of a language such as Prolog, and, in addition, possesses certain higher-order features. The nature of these additional features is illustrated, and it is shown how the use of the terms of a (typed) (lamda)-calculus as data structures provides a source of richness to the logic programming paradigm. AN University Microfilms Order Number ADG87-13897. AU OFLAZER, KEMAL. IN Carnegie-Mellon University Ph.D. 1987, 210 pages. TI PARTITIONING IN PARALLEL PROCESSING OF PRODUCTION SYSTEMS. SO DAI V48(03), SecB, pp814. DE Computer Science. AB This thesis presents research on certain issues related to parallel processing of production systems. It first presents a parallel production system interpreter that has been implemented on a four-processor multiprocessor. This parallel interpreter is based on Forgy's OPS5 interpreter and exploits production-level parallelism in production systems. Runs on the multiprocessor system indicate that it is possible to obtain speed-up of around 1.7 in the match computation for certain production systems when productions are split into three sets that are processed in parallel. However for production systems that are already relatively fast on uniprocessors, the communication overhead imposed by the implementation environment essentially offsets any gains when productions are split for parallel match. The next issue addressed is that of partitioning a set of rules to processors in a parallel interpreter with production-level parallelism, and the extent of additional improvement in performance. The partitioning problem is formulated and an algorithm for approximate solutions is presented. Simulation results from a number of OPS5 production systems indicate that partitionings using information about run time behaviour of the production systems can improve the match performance by a factor of 1.10 to 1.25, compared to partitionings obtained using various simpler schemes. The thesis next presents a parallel processing scheme for OPS5 production systems that allows some redundancy in the match computation. This redundancy enables the processing of a production to be divided into units of medium granularity each of which can be processed in parallel. Subsequently, a parallel processor architecture for implementing the parallel processing algorithm is presented. This architecture is based on an array of simple processors which can be clustered into groups of potentially different sizes, each group processing an affected production during a cycle of execution. Simulation results for a number of production systems indicate that the proposed algorithm performs better than other proposed massively parallel architectures like DADO, or NON-VON that use much larger number of processors. However, for certain systems, the performance is in the same range or sometimes worse than that can be obtained while a parallel interpreter based on Forgy's RETE algorithm such as an interpreter using production-level parallelism implemented on a small number of powerful processors, or an interpreter based on Gupta's parallel version of Forgy's RETE algorithm, implemented on a shared memory multiprocessor with 32 - 64 processors. AN University Microfilms Order Number ADG87-14990. AU POTTER, WALTER DONNELL. IN University of South Carolina Ph.D. 1987, 247 pages. TI A KNOWLEDGE-BASED APPROACH TO ENTERPRISE MODELING: THE FOUNDATION. SO DAI V48(04), SecB, pp1097. DE Computer Science. AB This dissertation describes the Knowledge/Data Model. The description includes the modeling foundation and primitives, the representational paradigm, a formal schema specification language, and a prototype implementation based upon the model. The Knowledge/Data Model captures both knowledge semantics, as specified in Knowledge Based Systems, and data semantics, as represented by Semantic Data Models. The Knowledge/Data Model can be thought of as an instance of a newly defined class of data models, called Hyper-Semantic Data Models, that facilitate the incorporation of knowledge in the form of heuristics, uncertainty, constraints and other Artificial Intelligence Concepts, together with object-oriented concepts found in Semantic Data Models. The unified knowledge/data modeling features provide a novel mechanism for combining Artificial Intelligence and Database Management techniques to establish the foundation of a Knowledge/Data Model for an Expert Database System. These features are provided via the constructs of the specification language, called the Knowledge/Data Language. The Knowledge/Data Language is the formal specification language for the Knowledge/Data Model. It is characterized as a context free language and is represented by a collection of grammar rules that specify the syntax of the language. The constructs of the language allow the features of the Knowledge/Data Model to be utilized in a modeling situation. In addition to being context-free, the Knowledge/Data Language is self-descriptive (sometimes referred to as self-referential). Throughout the dissertation, modeling examples, including the prototype application description, are presented using the language. AN University Microfilms Order Number ADG87-14121. AU SCHOCKEN, SHIMON. IN University of Pennsylvania Ph.D. 1987, 308 pages. TI ON THE UNDERLYING RATIONALITY OF NON-DETERMINISTIC RULE-BASED INFERENCE SYSTEMS: A DECISION SCIENCES PERSPECTIVE. SO DAI V48(04), SecB, pp1099. DE Computer Science. AB This research investigates the underlying rationality of several leading mechanisms in artificial intelligence designed to elicit, represent, and synthesize experts' belief: Bayesian inference, the certainty factors (CF) calculus, and an ad-hoc Bayesian inference mechanism. The research methodology includes a review of the philosophical foundations of these "belief languages," a mathematical analysis of their proximity to a classical Bayesian belief updating model, and an empirical comparison of their performance in a controlled experiment involving human subjects and their corresponding computer-based expert systems. The major analytic finding is that the certainty factors language is a special case of the Bayesian language. This implies that the certainty factors language is consistent with its Bayesian interpretation if and only if it is restricted to a very small subset of realistic inference problems. However, the widely-used CF language might perform better than its Bayesian counterpart due to the greater semantic appeal of the former. With this in mind, the thesis compares the descriptive and external validity of the three languages in a controlled experiment. The major empirical results are (a) within the limited context of this experiment, neither the certainty factors nor the Bayesian language dominates the other in terms of descriptive validity, defined as the proximity of the system's judgment to actual experts' judgment; and (b) the correlation between the computer-based Bayesian judgment and the pooled expert judgments is significantly greater than the corresponding CF correlation. To sum up, this research shows that the classical Bayesian approach to rule-based inference appears to dominate the certainty factors language, both on analytic and empirical grounds. At the same time, the proven success of CF-based systems (e.g. MYCIN) and its wide popularity suggest that the CF approach to inference is indeed appealing to many designers and users of expert systems. It is suggested that future research attempt to formulate a synthetic approach to knowledge engineering, i.e. one that combines the attractive descriptive features of the CF language with the normative rigor of a Bayesian design. It is hoped that this will strike a balance between preserving the intuitive element of human reasoning, and, at the same time, enforcing a certain degree of normative rationality. AN University Microfilms Order Number ADG87-15371. AU UMRIGAR, ZERKSIS DHUNJISHA. IN Syracuse University Ph.D. 1986, 259 pages. TI AUTOMATION OF HARDWARE-CORRECTNESS PROOFS. SO DAI V48(04), SecB, pp1099. DE Computer Science. AB The ubiquity of the digital computer and its use in critical applications makes verification of its correctness an extremely important issue. Unfortunately present verification methodologies, which rely almost exclusively on simulation, have difficulty handling the complexity of modern hardware designs. In this dissertation we explore an alternate verification methodology in which the functional correctness of a design is proved using formal proof techniques. To prove the correctness of a design, a formal hardware verification system is given two formal descriptions of the design which correspond to a functional specification and an implementation. It must then establish an implication or equivalence between these two descriptions. This can be done using exhaustive simulation, but this is slow and cannot be used to verify parameterized circuits. A more general method is to use algebraic simulation to derive verification conditions and then use a theorem prover to establish the validity of these verification conditions. An interactive general purpose theorem prover which is a partial decision procedure for first-order logic is used as a shell for more efficient but specialized algorithms. A specialized algorithm, called the bounds algorithm is used to establish the validity of formulas involving universally quantified linear inequalities over the integer domain. This algorithm is goal-directed and is easily extended to handle some properties of interpreted functions. Theoretical properties of these theorem proving procedures are established. The usefulness of the formal verification system is limited by its theorem proving component. It has successfully been used to verify the functional correctness of simple arithmetic circuits, including an array multiplier. AN University Microfilms Order Number ADG87-13707. AU ZERNIK, URI. IN University of California, Los Angeles Ph.D. 1987, 346 pages. TI STRATEGIES IN LANGUAGE ACQUISITION: LEARNING PHRASES IN CONTEXT. SO DAI V48(03), SecB, pp815. DE Computer Science. AB How is language acquired by people, and how can we make computers simulate language acquisition? Although current linguistic models have investigated extensively parsing and generation, so far, there has been no model of learning new lexical phrases from examples in context. We have identified four issues in language acquisition. (a) How can a phrase be extracted from a single example? (b) How can phrases be refined as further examples are provided? (c) How can the context be incorporated as part of a new phrase? (d) How can acquired phrases be used in parsing and in generation? In solving this problems, we have established three theoretical points. (a) We have shown how a dynamic lexicon is structured as a phrasal hierarchy. (b) We have constructed strategies for learning phrases. (c) We have constructed a parsing mechanism which can operate even in the presence of lexical gaps. The program RINA has incorporated these elements in modeling a second-language speaker who augments her lexical knowledge by being exposed to examples in context. AN University Microfilms Order Number ADG87-13708. AU CRAIG, ELAINE M. IN University of California, Los Angeles Ph.D. 1987, 283 pages. TI EXPERT AND NOVICE PROBLEM SOLVING IN A COMPLEX COMPUTER GAME. SO DAI V48(03), SecA, pp600. DE Education, Psychology. AB This study examined the problem solving processes involved in playing a complex computer game and explored the utility of computer games for research on problem solving and for instruction in problem solving skills. The study characterized and compared the problem solving behaviors of "expert" and "novice" game players. It compared expert/novice contrasts in computer game players with expert/novice problem solving differences in other domains such as physics, computer programming, and errand planning. The study also looked at the changes in problem solving behaviors that occurred when novices moved toward expert play and considered the potential for incorporating computer game activities in problem solving instructional programs. The Opportunistic Planning Model (OPM) (Hayes-Roth & Hayes-Roth, 1978, 1979) provided the theoretical basis and the methodological framework for the study which looked at the problem solving behaviors of 18 university undergraduates playing an "off-the-shelf" computer game. Measures of subjects' problem solving behaviors included audio recordings of what they said while playing the game ("think aloud" protocols), detailed observations of their game play, and interviews before and after game play. Data were analyzed using t-tests and chi square tests. The study found the following problem solving behaviors to be associated with success at a computer game: making high level decisions, exploiting world knowledge, showing sensitivity to constraints, clustering tasks, using a system to organize information, considering alternatives, and assessing the state of one's knowledge. The study found very few increases in problem solving behaviors as subjects became more experienced with the game. It also found that computer game play involved subjects in many of the same activities that are incorporated in problem solving instructional programs. AN University Microfilms Order Number ADG87-13882. AU BUSHNELL, MICHAEL LEE. IN Carnegie-Mellon University Ph.D. 1987, 250 pages. TI ULYSSES -- AN EXPERT-SYSTEM BASED VLSI DESIGN ENVIRONMENT. SO DAI V48(03), SecB, pp833. DE Engineering, Electronics and Electrical. AB Ulysses is a VLSI computer-aided design (CAD) environment which effectively addresses the problems associated with CAD tool integration. Specifically, Ulysses allows the integration of CAD tools into a design automation (DA) system, the codification of a design methodology, and the representation of a design space. Ulysses keeps track of the progress of a design and allows exploration of the design space. The environment employs artificial intelligence techniques, functions as an interactive expert system, and interprets descriptions of design tasks encoded in the scripts language. An integrated circuit silicon compilation task is presented as an example of the ability of Ulysses to automatically execute CAD tools to solve a problem where inferencing is required to obtain a viable VLSI layout. The inferencing mechanism, in the form of a controlled production system, allows Ulysses to recover when routing channel congestion or over-constrained leaf-cell boundary conditions make it impossible for CAD tools to complete layouts. Also, Ulysses allows the designer to intervene while design activities are being carried out. Consistency maintenance rules encoded in the scripts language enforce geometric floor plan consistency when CAD tools fail and when the designer makes adjustments to a VLSI chip layout. Consistency maintenance is discussed extensively using floor planning, leaf-cell synthesis, and channel routing tasks as examples. Ulysses has been implemented as a computer program and a chip layout that was semi-automatically generated by Ulysses is presented to illustrate the performance of the program. AN University Microfilms Order Number ADG87-13900. AU SAUK, BRIAN EDWARD. IN Carnegie-Mellon University Ph.D. 1987, 98 pages. TI LEILA: AN EXPERT SYSTEM FOR ESTIMATING CHEMICAL REACTION RATES. SO DAI V48(03), SecB, pp840. DE Engineering, Electronics and Electrical. AB This work describes an expert system, named Leila, capable of estimating chemical reaction rates. These estimates are based on fundamental data and a hierarchy of reaction rate theories. The theories are encoded in the form of production rules, and the expert system methodology chosen for Leila is that of a production rule system. Unlike most production systems, the rules in Leila are segmented into nodes. Nodes represent knowledge about a specific area of the reaction rate domain. During a rate determination, attention is focused on only one node at a time, thus minimizing the number of rules that need to be considered at each step. In addition, since nodes represent a specific area of expertise, extensions and modifications are simplified, since they only deal with a small portion of the knowledge base. Leila also provides a model for the solution of reaction rate problems. The steps of this model are defined by rules, thereby enabling modifications to the model without extensive recoding. During a rate determination, Leila checks for balanced reactions, classifies reactions, performs rate determinations based on hierarchies of theories, estimates unknown data, performs any unit conversions, and shows the solution path taken by the determination, if requested. The rate theories present in Leila deal primarily with low-pressure gas phase reactions, and in particular, recombination and ionization reactions. A summary of the reactions that Leila can handle is given. For some reactions, many theories apply, while for others, only one theory can be used. A number of comparisons to experimental data is also presented. In many cases, the theoretical estimates are in good agreement with experiment, while for others agreement is poor. Reasons for disagreement are given. AN University Microfilms Order Number ADG87-13817. AU CHEN, JEN-GWO. IN The University of Oklahoma Ph.D. 1987, 143 pages. TI PROTOTYPE EXPERT SYSTEM FOR PHYSICAL WORK STRESS ANALYSIS (dBASE III). SO DAI V48(03), SecB, pp845. DE Engineering, Industrial. AB This research involves the development of an interactive knowledge-based Ergonomics Analysis SYstem (EASY) for physical work stress analysis. EASY was written in dBASE III and BASIC for IBM-PC compatible microcomputers. The system consists of three major components: the Physical Work Stress Index (PWSI) used by the supervisor or ergonomist for further investigation of problem situations, the Ergonomics Information Analysis System (EIAS) for evaluation of tasks by the worker, and the Dynamic Lifting Analysis System (DLAS) for manual material handling tasks. The Physical Work Stress Index is an observational method of physical work stress analysis which possesses the ease of application of traditional work study techniques but provides better accounting of human and task variables. The technique involves activity sampling of various physical components of the work including body location, base of support, orientation, hand position, acceleration and thermal load. The PWSI is derived from observational data and is classified into six different levels: very low, low, moderate, high, very high and extremely high. The EIAS includes four sections: case identification, problem description, job description and operator-operation interaction. The last two sections record quantitative data as opposed to the qualitative data collected in the first two sections. The quantitative data consists of a 5-point scale which describes the seriousness of each aspect of the problem. The EIAS provides general guidelines to tell the user how to avoid unnecessary problems and improve performance. The DLAS includes three components: lifting capacity analysis, biomechanical analysis and NIOSH guidelines analysis. Extensive use of menus for database entry/editing and analysis provides an efficient and friendly interface design. The system was evaluated by comparing the results of EASY and individuals with an introductory knowledge of ergonomics with experts' conclusions for nine test jobs involving a variety of physical work stressors. The evaluation indicated that 83% of EASY's diagnoses were accepted by the experts with some variation between individual experts and between EASY and the other diagnosticians. AN University Microfilms Order Number ADG87-14897. AU WU, SZU-YUNG DAVID. IN The Pennsylvania State University Ph.D. 1987, 235 pages. TI AN EXPERT SYSTEM APPROACH FOR THE CONTROL AND SCHEDULING OF FLEXIBLE MANUFACTURING CELLS. SO DAI V48(04), SecB, pp1125. DE Engineering, Industrial. AB An expert system is a computer program that uses knowledge and inference procedures to solve problems. Today, most expert systems contain a substantial amount of domain expertise (i.e., knowledge) organized for efficient problem solving. However, most of the existing design philosophies for expert systems do not lend themselves to real-time control environments. Expert systems are currently being touted as a means of resolving factory scheduling problems. Unfortunately, the expert systems developed to date are neither generic nor responsive enough to be used for on-line system control. In this research, an architecture is created which takes advantage of both expert system technology and discrete event simulation. The simulation is used as a prediction mechanism to evaluate several possible control alternatives provided by the expert system. A performance measure is obtained from the simulation for each of the suggested alternatives. A control effector is then employed to affect the physical control of the cell based on the performance measure. This performance measure is worth a great deal of domain-specific knowledge that would otherwise have to be included in the expert knowledge base. The integration of the expert control system, the simulation, and the control effectors, form a system called MPECS. MPECS is used to control Flexible Manufacturing Cells (FMC). Specific software and algorithms are developed to define and implement the system. The control architecture is examined using the information from an existing FMC to demonstrate its feasibility. AN University Microfilms Order Number ADG87-13891. AU GURSOZ, ESAT LEVENT. IN Carnegie-Mellon University Ph.D. 1987, 193 pages. TI EXPERT TASK PLANNING IN ROBOTIC CUTTING OPERATIONS. SO DAI V48(03), SecB, pp854. DE Engineering, Mechanical. AB In this thesis, an expert system is developed for a class of automated cutting operations. These operations include plasma cutting, oxy-fuel cutting, laser cutting and water-jet cutting. The common features in these processes, which define this class of cutting operations, are the following: first, the work material is cut by the sweeping action of a line segment emanating from the process tool; second, the cutting effect terminates at an imprecise point along that cutting segment; and third, the cutting task at hand can be fully described by the surface-boundary representation of the workpiece and the surface to be cut. The surface-boundary representation is a fairly standard form of modeling in CAD systems. Hence the description of the cutting task can easily be supplied by a CAD database if it exists or can be interactively defined within a CAD system. The specific concern in this thesis is robotic applications in such tasks. Given such a CAD description of the cutting task, we have developed an expert system to generate the robot program which shall execute the desired cut. This overall transformation from the task description to the robot program can be naturally divided into two phases. In the first phase, the cutting task is formulated in a manipulator independent fashion to the level where relative movements of the cutting segment are prescribed. In the second phase, a robot program which articulates the prescribed cutting segment motions are generated. The focus of this study deals with the first phase in which the cutting task is planned. The fundamental problem in such a planning task is that neither a strictly geometrical analysis, nor a purely heuristic approach is a sufficient basis when considered alone. Commonly, geometric modeling is used in simulating manufacturing operations. Knowledge-based robot task planning, on the other hand, has usually been implemented for the cases where complicated spatial reasoning is not required. In this thesis, we have developed a knowledge-based system which blends heuristics with spatial reasoning within the framework of a solid modeling system. Although an implementation of robotic flame cutting of structural beams is used to provide the fundamental knowledge and the context, this system is constructed in a general fashion to cover all of the addressed cutting operations. Furthermore, it is possible to extend the developed planning concepts to other manufacturing applications where spatial reasoning is crucial. AN University Microfilms Order Number ADG87-10627. AU STANDLER, NANCY ANN. IN The University of Rochester Ph.D. 1986, 508 pages. TI AUTOMATED IDENTIFICATION OF A COMBINED POPULATION OF NEURONS AND ASTROCYTES: APPLICATION OF A PROGRAMMING APPROACH SUITABLE FOR HIGH RESOLUTION HISTOLOGICAL SPECIMENS. SO DAI V48(03), SecB, pp710. DE Health Sciences, Pathology. AB A combined population of neurons and astrocytes in semithin (1 micrometer) sections of mouse cortex is automatically identified with greater than 95% accuracy. The computer algorithms use a new programming approach that shows promise of being applicable to the identification of a wide variety of structures in complex, high resolution images of histological sections. The approach stresses the use of histologically meaningful distinctions between similar sites in cells and in other structures in the histological section. The use of logical trees to identify the cells enables the algorithms to tolerate large variations in appearance from cell to cell while retaining the ability to make subtle distinctions between particular cells and non-cell structures with very similar appearance. Difficulties with segmenting the cells from the background are avoided by using branch point tests in the logical trees that do not require segmented images. AN University Microfilms Order Number ADG87-16349. AU CHANG, HSI ALEX. IN The University of Arizona Ph.D. 1987, 407 pages. TI AN ARCHITECTURE FOR ELECTRONIC MESSAGING IN ORGANIZATIONS: A DISTRIBUTED PROBLEM-SOLVING PERSPECTIVE. SO DAI V48(04), SecA, pp768. DE Information Science. AB This dissertation provides a foundation for electronic information management in organizations. It focuses on the relationships among communication, control, and information flows of the organization. The main thesis addresses the question of how electronic mail messages may be managed according to their contents, ensuring at the same time, the preservation of organizational and social relationships. A taxonomy for the management of unstructured electronic information relevance based on the treatment of information is derived from current research. Among the three paradigms, the information processing, the information distribution, and the information sharing paradigms, the inadequacy of the first two is recognized, and the treatment of information in its active mode is proposed. This taxonomy can be used to quickly differentiate one research from another and evaluate its adequacy. Three concepts, four cornerstones, and an architecture constitute our framework of information relevance management. The cornerstones are knowledge of the organization, knowledge of the individual, information construction, and information interpretation. Through knowledge of the organization and the individual, the machine production systems are able to distribute and manage information according to the logic of human production systems. The other two cornerstones together improve the unity of interpretation among the organizational members. The physical architecture can adapt a number of applications, each of which, may not only have different knowledge presentations and inference mothods, but also may co-exist in the system simultaneously. An integrated knowledge-based electronic messaging system, the AI-MAIL system, is built, tested, and evaluated through a case study to demonstrate the feasibility of the architecture and its applicability to the real-world environment. The three operating levels, interorganizational, intraorganizational, and individual, are illustrated through a study of the U.S. Army. From three large scale field studies, the existing AUTODIN I system, a backbone of the Army's communications, is analyzed and evaluated to illustrate the applicability and benefits of the three operating levels. This dissertation contributes to the field of Management Information Systems by offering a methodology, a taxonomy, a new paradigm, a framework, and a system for information management and a method of adaptive organizational design. In addition, it points toward future research directions. Among them are research to deal with ethical issues, organizational research, knowledge engineering, multi-processor configuration, and internal protocols for applications. AN University Microfilms Order Number ADG87-16352. AU FJELDSTAD, OYSTEIN DEVIK. IN The University of Arizona Ph.D. 1987, 394 pages. TI ON THE REAPPORTIONMENT OF COGNITIVE RESPONSIBILITIES IN INFORMATION SYSTEMS. SO DAI V48(04), SecA, pp768. DE Information Science. AB As the number of information system users increases, we are witnessing a related increase in the complexity and the diversity of their applications. The increasing functional complexity amplifies the degree of functional and technical understanding required of the user to make productive use of the application tools. Emerging technologies, increased and varied user interests and radical changes in the nature of applications give rise to the opportunity and necessity to re-examine the proper apportionment of cognitive responsibilities in human-system interaction. We present a framework for the examination of the allocation of cognitive responsibilities in information systems. These cognitive tasks involve skills associated with the models and tools that are provided by information systems and the domain knowledge and problem knowledge that are associated with the user. The term cognitor is introduced to refer to a cognitive capacity for assuming such responsibilities. These capacities are resident in the human user and they are now feasible in information system architectures. Illustrations are given of how this framework can be used in understanding and assessing the apportionment of responsibilities. Implications of shifting and redistributing cognitive task from the system-user environment to the system environment are discussed. Metrics are provided to assess the degree of change under alternative architectures. An architecture for the design of alternative responsibility allocations, named Reapportionment of Cognitive Activities, (RCA), is presented. The architecture describes knowledge and responsibilities associated with facilitating dynamic allocation of cognitive responsibilities. Knowledge bases are used to support and describe alternative apportionments. RCA illustrates how knowledge representations, search techniques and dialogue management can be combined to accommodate multiple cooperating cognitors, each assuming unique roles, in an effort to share the responsibilities associated with the use of an information system. A design process for responsibility allocation is outlined. Examples of alternative responsibility allocation feasible within this architecture are provided. Cases implementing the architecture are described. We advocate treating the allocation of cognitive responsibilities as a design variable and illustrate through the architecture and the cases the elements necessary in reapportioning these responsibilities in information systems dialogues. AN University Microfilms Order Number ADG87-12660. AU CRITTENDEN, CHARLOTTE CLEMENTS. IN University of Georgia Ph.D. 1987, 181 pages. TI A STUDY OF SIX PRONOUN USAGES: FOR PRACTICAL PURPOSES. SO DAI V48(03), SecA, pp640. DE Language, Linguistics. AB This study covers six areas of pronoun usage: broad reference of which, that, and this; impersonal you; agreement with indefinites; agreement with collective nouns; whose as genitive of which; and the and which or and who constructions. The method used was to edit approximately one million words of selected material published in 1983. Three types of primary sources have contributed to this survey: twenty nonfiction best sellers, articles from ten periodicals chosen from a variety of readership levels, and newspaper editorials from five representative geographical areas. The editing identified usages that are different from those advanced by a number of traditional grammar books and handbooks used on the college level. Included in this study is historical information from the OED and from scholars like Otto Jespersen, Albert Marckwardt, Fred G. Walcott, and J. Lesslie Hall. Additionally, a number of twentieth-century usage studies were surveyed, including, among others, those of Paul Roberts, C. C. Fries, Robert Pooley, Margaret Nicholson, and Bergen and Cornelia Evans. Several studies written by journalists, e.g., Roy Copperud and Wilson Follett, contribute added perspective. Further descriptive information comes from two dictionaries often used on the college level: Webster's Third and the American Heritage. After listing the approximate number of identified examples of each usage being investigated in all three types of primary sources and citing typical quotations, this study makes observations about the actual use of each pronoun construction in relation to its history, reports from usage studies, dictionary notes, and handbook information. The study finally draws general conclusions and discusses implications appropriate for an effective approach in using and teaching these six areas of pronoun reference. AN This item is not available from University Microfilms International ADG05-60501. AU HALL, CHRISTOPHER JOHN. IN University of Southern California Ph.D. 1987. TI LANGUAGE STRUCTURE AND EXPLANATION: A CASE FROM MORPHOLOGY. SO DAI V48(04), SecA, pp914. DE Language, Linguistics. AB This investigation examines the contribution of psycholinguistic and diachronic factors to the development across languages of a preference for suffixing over prefixing. It argues for an approach to explanation in linguistics that stresses: (a) the need for an investigation of potential underlying psychological or functional principles, involving the cooperation of the various subdisciplines of linguistics; and (b) the need for an explicit description of the mechanism(s) of "linkage" between structure and explanation, i.e., an account of how languages developed the properties in question. The investigation draws on principles of lexical processing, diachronic change, universals/typology, theoretical morphology, and semantics in order to provide a fuller and more motivated explanation than has previously been offered. It critically evaluates the major prior effort to explain the suffixing preference provided by Cutler, Hawkins & Gilligan (1985). The discussion presented here suggests that, although their fundamental insights were correct and provide the basis for the present work, there are three areas of inadequacy: (a) the processing explanation is inaccurate in some details; (b) it is incomplete in that no explanation of the mechanism of linkage is provided; and (c) the Head Ordering Principle, formulated to "explain" the basic pattern of the crosslinguistic data, is based on an erroneous assumption, and is, in any case, more a statement of a generalisation than an explanation of the facts. The explanation offered in the present work refines the processing explanation and introduces factors from language change into the explanatory hypothesis. It is argued that the suffixing preference is, in actuality, a prefixing dispreference that ultimately derives from the conflict between two driving forces of language change, namely, the opposing pressures of economy and clarity. The former leads to semantic redundancy and phonological reduction within words, and this interacts with the latter which leads to maintenance of stem initial strength and a resistance to prefixing, for reasons of efficient processing. Two original experiments on word recognition in English are also reported. Experiment I examines the processing of prefixed words at various stages of reduction; Experiment II focuses on the hypothesised locus of the dispreference for prefixing. The results yield initial support for the account proposed. (Copies available exclusively from Micrographics Department, Doheny Library, USC, Los Angeles, CA 90089-0182.). AN University Microfilms Order Number ADG87-12184. AU OSHIRO, MARIAN MIDORI. IN The University of Michigan Ph.D. 1987, 264 pages. TI A TAGMEMIC ANALYSIS OF CONVERSATION. SO DAI V48(03), SecA, pp641. DE Language, Linguistics. AB The tagmemic method of linguistic analysis as developed by Kenneth L. Pike is applied to the analysis of informal multi-party verbal interaction ('conversation'). The three part-whole hierarchies of units of tagmemic analysis--grammatical, referential, and phonological--are each discussed with reference to prior analysts' choices of units. Methodological problems of analyzing conversation are discussed and the hierarchies reevaluated and modified in response to them. Methodological questions include (1) identification of nuclei and margins, and boundary definitions of units, (2) differences between written and oral texts, and implications of the presence of hearer/respondent(s) in spontaneous verbal interactions, and (3) the nature of cohesion and the degree and kind of convergence of the three hierarchies at their upper levels. A central question is how to treat speakership in the analysis. The conclusions reached are that alternation of speakers should not be used as a feature of grammatical units; that speakership is reflected in the purpose (an element of cohesion) of the 'move', which is a unit of the referential analysis; and that the individual speaker's voice is a feature of the unit labeled the 'turn' in the phonological hierarchy of units. Although the word 'turn' is used in this dissertation as a technical term limited to a single hierarchy, the tri-hierarchical approach of tagmemic analysis is found to contribute toward an understanding of what is commonly referred to as a turn (an interactional component). The analysis of speech into three distinctive systems clarifies the problem of defining a turn by identifying multiple points in an interaction--hierarchical unit boundaries--at which a change of speakers may take place. All three hierarchies as constructed for conversational analysis include the Episode and History as their highest-level units. The other units of the revised grammatical hierarchy are the Morpheme, Morpheme Cluster, Word, Grammatical Phrase, Grammatical Clause, Grammatical Sentence, and Grammatical Paragraph. For the referential hierarchy, the other units are the Concept, Concept Complex, Monolog, Exchange, Interlogue, and Speech Event. For the phonological hierarchy, they are the Phoneme, Syllable, Word, Phonological Phrase, Phonological Clause, Phonological Sentence, Turn, Phonological Paragraph (projected), and Conversation. AN University Microfilms Order Number ADG87-13789. AU RAVIN, YAEL. IN City University of New York Ph.D. 1987, 319 pages. TI A DECOMPOSITIONAL APPROACH TO PREDICATES DENOTING EVENTS. SO DAI V48(03), SecA, pp641. DE Language, Linguistics. AB The semantic representation of predicates has received renewed attention in recent linguistic research, following the 1981 publication of Chomsky's Lectures on Government and Binding. One of the major features of Chomsky's new theory is the reinstitution of thematic roles, such as Agent and Patient, to express semantic relations between predicates and their arguments. These roles are posited as primitives and play a prominent part in the derivation of syntactic structures. The first part of this dissertation argues against theories such as Chomsky's, which rely on thematic roles. It is shown that their underlying Restrictive approach prevents them from accounting for the syntax and semantics of propositions denoting events. The second part of the dissertation argues in favor of a Nonrestrictive, non-thematic approach to semantics. J. Katz's Decompositional Theory is the Nonrestrictive model adopted here. The meaning of several predicates and propositions denoting events is analyzed and represented in terms of Katz's Theory. The Decompositional analysis is contrasted with the different thematic analyses to reveal a formal system for semantic representation which is complete and consistent and a set of principles which determine semantic properties and relations. AN University Microfilms Order Number ADG87-15458. AU BRINGSJORD, SELMER C. IN Brown University Ph.D. 1987, 226 pages. TI THE FAILURE OF COMPUTATIONALISM. SO DAI V48(04), SecA, pp937. DE Philosophy. AB This dissertation is composed of a number of arguments against the thesis that persons are automata. AN University Microfilms Order Number ADG87-14418. AU CLING, ANDREW DEAN. IN Vanderbilt University Ph.D. 1987, 216 pages. TI DISAPPEARANCE AND KNOWLEDGE: AN EXAMINATION OF THE EPISTEMOLOGICAL IMPLICATIONS OF ELIMINATIVE MATERIALISM. SO DAI V48(03), SecA, pp667. DE Philosophy. AB The purpose of this dissertation is to consider Paul Churchland's arguments for eliminative materialism and for the abolition of traditional epistemology. It is shown that these arguments are faulty and that there is more to be said for our commonsense conception of mentality than the eliminative materialist supposes. The essay begins by explaining the eliminative materialists' claim that our commonsense conception of mentality is an outmoded theory which will, or at least should, be replaced by a theory to be drawn from completed brain science. Drawing on contemporary work in metaphysics and the philosophy of science, it is shown that supervenience is an important intertheoretical relation which is not equivalent to reduction or elimination. Supervenience allows us to reconcile the claim that everything is physical with the claim that not all properties are expressible in the language of physics. Using this result, I argue that three of Churchland's four arguments for eliminative materialism rest on the dubious metaphysical assumption that all theories will either reduce to or be eliminated by completed physical science. It is shown that this failure is deeply ironic given Churchland's claim that disputes in the philosophy of mind are largely empirical in character. It is also shown, however, that eliminative materialists can easily respond to charges that their view is somehow self-referentially incoherent. It is shown that Churchland's fourth argument for eliminative materialism, and for the claim that traditional epistemology should be abolished, depends upon his first three arguments and is, therefore, flawed. It is also shown that the argument is a failure in its own right. The essay concludes by showing that there are some important respects in which our commonsense conception of mentality and traditional epistemology are superior to purely materialistic accounts. This superiority stems, in large part, from the availability of intentional states such as beliefs. AN University Microfilms Order Number ADG87-16367. AU DADZIE, S. S. IN Temple University Ph.D. 1987, 164 pages. TI THE GRICE PROBLEM: A CRITICAL ANALYSIS OF THE CAUSAL THEORY OF PERCEPTION. SO DAI V48(04), SecA, pp937. DE Philosophy. AB The essay examines H. P. Grice's attempt to formulate the necessary and sufficient conditions of perceiving in purely causal terms. It involves appraisal of P. F. Strawson's criticism of the thesis as inherently circular; George Pitcher's defence of it against Strawson's challenge; Alvin I. Goldman's Historical Reliabilism, a causal-cum-belief theory of knowledge which had started off as a strictly Gricean analysis; and, finally, Donald Davidson's theory of the explanation of action which construes reasons as causes and, hence, explanation by reasons as only a species of ordinary causal explanation. According to our finding, Grice's thesis is indeed vulnerable to Strawson's objection; Pitcher fails to deflect the force of Strawson's attack, his own composite account of perception (couched in causal, behavioral and direct realist terms) fails to improve the prospects of Grice's doctrine; and its merits notwithstanding, Strawson's critique lacks the wherewithal to make it a decisive argument against the causal program. Our argument thence: the necessary and sufficient conditions of perception cannot be provided in causal terms; an adequate account has to be non-causal or, at least, include (or reflect) factors which are demonstrably refractory to causal analysis (for example, the concept's integrally cognitive force, plus its intensional properties). The study does not pretend to offer a comprehensive theory, however, specifying the necessary and sufficient conditions of perception in non-causal terms; it merely sketches the kind of lines necessary for doing this if this were viable. The results are fruitful, nonetheless: for, along with its central task of settling a heretofore unresolved dispute in perception theory proper (that between the Strawsons and the Pitchers), the study affords a sense of the interconnections among seemingly disparate issues, illuminating some age-old puzzles in philosophical debate; notable among these being, of course, the two-fold flaw disclosed in the causalist's program (Grice's as well as Goldman's and Davidson's), namely, its weak grasp of the intensional complexity of the concepts in question and, thence, its taking the general concession that causal factors are relevant, to somehow lead to the conclusion--without sufficient argument--that a causal theory of those concepts is adequate. AN University Microfilms Order Number ADG87-16113. AU HARBORT, ROBERT A., JR. IN Emory University Ph.D. 1987, 213 pages. TI APPLICATION OF HERMENEUTICS TO MODELS OF MEDICAL INFORMATION. SO DAI V48(04), SecA, pp939. DE Philosophy. AB A hermeneutic is an interpretation of something that integrates understanding and application. Derived from the name of the god Hermes, and referring to his bringing the gift of language to humanity, it most often refers to interpretation and application of biblical texts. From the late nineteenth century it has been used by philosophers and literary critics to apply to a wider field of interpretation. It is different from exegesis or explanation in the scientific sense, which is divorced from practicality. Hans-Georg Gadamer has been instrumental in linking the idea of interpretation as the integration of explanation and application with Aristotle's idea of "practical philosophy" as found in the Ethics. He used analogies with everyday activities to illustrate ideas about interpretation of text; I turned the process around to ask whether there would be any advantage to modeling certain nonliterary activities as interpretive processes. In particular, I was interested in modeling various processes associated with medicine. The hermeneutic model does not necessarily generate more precision (in the scientific sense) in descriptions of medical activities, but it does allow the model to include self-awareness. This has not been available to models of medical activity with any degree of objective content, yet treatises on the philosophy of medicine list it as an important characteristic. Medicine is an example of a hermeneutic activity at several levels. Medical education, the individual practice of medicine by one physician with one patient, the health care delivery system, and medical ethics are all examples of medicine as hermeneutics. Previous work in modeling of information and information processing in medicine has been based primarily on scientific or existential epistemologies. I will examine hermeneutics as a context in which models of medical information and information processing are to be judged for effectiveness. The purpose of the dissertation is to establish the validity of the hermeneutic model and to use it to evaluate several models of information and information processing in medicine. AN University Microfilms Order Number ADG87-12439. AU SAUNDERS, RUTH ANN. IN The University of Wisconsin - Madison Ph.D. 1987, 278 pages. TI KNOWLEDGE WITHOUT BELIEF: FODOR VERSUS PIAGET ON COGNITIVE EXPLANATIONS OF COGNITIVE CHANGE. SO DAI V48(03), SecA, pp669. DE Philosophy. AB Jerry Fodor has recently argued for a version of nativism based on the claim that it is impossible to give a cognitive account of how new cognitive powers are acquired. Piaget has insisted that without such an account, it is impossible to understand what cognition is. My main concern in this work has been to expose and clarify the deeper philosophical disagreements that underlie the surface dispute. This work brings to light basic disagreements over the nature of knowledge, over what the fundamental units of cognitive psychology are, and over what cognitive psychology ought to explain. For each side of the dispute, I devote two chapters to articulating a set of basic assumptions, defending their prima facie plausibility, and showing how they lead to either Fodor's or Piaget's claim. Fodor's nativism is presented as a true claim about the logical character of certain sorts of representational theories of cognition. Piaget's theory is interpreted as an account of increasing knowledge of objects rather than as an account of internal mental organization. So interpreted, Piaget's theory avoids Fodor's charge of incoherence, avoids some common objections to the notions of stage and equilibration, and presents a radically new understanding of knowledge and cognition. To explicate Fodor's claim, I show how it arises from one line of thought within standard views about the nature of epistemology and cognitive psychology. In the process, I identify assumptions that are crucial for understanding the conflict between Fodor and Piaget. The contrasting assumptions I develop to make sense of Piaget's claim are that: (1) knowledge of objects is direct (rather than mediated by knowledge of facts about the objects); (2) the fundamental units of cognitive psychology are interactions and interaction patterns (i.e., relationships between knowers and known objects, rather than internal causal states with narrow content); and, (3) cognitive explanations show how present interaction patterns and the nature of the known object generate new cognitive powers (rather than showing how processes of belief formation and manipulation issue in behavior and new beliefs). AN University Microfilms Order Number ADG87-12973. AU SAYRE, PATRICIA ANN WHITE. IN University of Notre Dame Ph.D. 1987, 260 pages. TI MICHAEL DUMMETT ON THE THEORY OF MEANING. SO DAI V48(03), SecA, pp669. DE Philosophy. AB The dissertation examines Dummett's recommendations regarding the construction of a theory of meaning. It begins by taking up the question of why a theory of meaning is wanted. It is argued that the sense in which Dummett is concerned with meaning is broad enough to give no offense to those with Quinean prejudices against "meanings". It is also argued that the sense in which Dummett is concerned to construct a theory is narrow enough to place a number of constraints on the construction of a theory of meaning. Many of these constraints may appear arbitrary at first, but can be given a rationale by leaning hard on Dummett's suggestion that an adequate theory of meaning must have a "genuinely scientific character". This rationale can be extended to provide a basis for Dummett's objections to Davidson's truth-conditional theory of meaning, namely, his objections on the grounds that the theory is modest, holistic, and faces difficulties in dealing with undecidable sentences. Unfortunately, the rationale also provides a basis for objections to Dummett's verificationist and falsificationist alternatives to Davidson's theory. Dummett's alternatives are explicitly designed to be neither modest nor holistic, but they do face difficulties when it comes to undecidable sentences. It is argued that although these difficulties are not in principle insuperable, they do suggest that Dummett's constraints on the construction of a theory of meaning may make such a theory impossible to construct. AN University Microfilms Order Number ADG87-16392. AU SEREMBUS, JOHN HERMAN. IN Temple University Ph.D. 1987, 214 pages. TI ABSOLUTE COMPARATIVE PROBABILISTIC SEMANTICS. SO DAI V48(04), SecA, pp941. DE Philosophy. AB The thesis of the dissertation is that relations between statements of a formal language, which are suitably constrained to mirror the non-quantitative probability relation 'is not more probable than', can serve as a semantics for that language and that this absolute, comparative, probabilistic semantics is a generalization of absolute, quantitative, probabilistic semantics, that is, the semantics for a formal language that employs one-place functions that obey the laws of the probability calculus. Chapter one provides an historical sketch of the area to which the dissertation is a contribution. It traces the development of what came to be known as probabilistic semantics from the work of Sir Karl Popper through Robert Stalnaker, William Harper, Hartry Field, Kent Bendall, and Hugues Leblanc. It also provides a brief history of probability as a non-quantitative (comparative) concept by discussing the work of Bruno De Finetti, Bernard Koopman, and Charles Morgan. It concludes by explaining the thesis of the dissertation in light of the just-sketched tradition and spells out the program for the rest of the work. Chapter two presents the syntax of a propositional language PL and provides an absolute comparative probabilistic semantics for it. It then shows that the language is sound and complete with respect to that semantics. The last section gives an account of generalization and argues that this semantics is a generalization of the absolute comparative probabilistic semantics for PL. This amounts to claiming that for every probability function there is a corresponding probability relation and for every member of a proper subset of probability relations, namely, that set which contains only comparable relations, there is at least one probability function corresponding to it. Chapter three offers the same kind of results obtained in chapter two for a first order language FL. The final chapter offers a summation of the results and highlights some of the features of absolute comparative probabilistic semantics such as the intensionality of the logical operators and the existence of what are termed 'assumption sets'. It also suggests possible avenues of application and research involving the new semantics. AN University Microfilms Order Number ADG87-15497. AU GOLDEN, RICHARD MARK. IN Brown University Ph.D. 1987, 122 pages. TI MODELLING CAUSAL SCHEMATA IN HUMAN MEMORY: A CONNECTIONIST APPROACH. SO DAI V48(04), SecB, pp1174. DE Psychology, Experimental. AB Causal schemata represent knowledge of routine event sequences, and are constructed from causal relationships. A computational theory of causal schemata is proposed for controlling behavior and recalling actions from memory. Within this theory, learning is viewed as a procedure that involves estimating the probability distribution of causal relationships in the world. The memory recall process is a complementary procedure that uses the probability distribution function estimated by the learning process to select the most probable action to be executed or recalled within a particular situation. A neurophysiological implementation of this computational theory involving Anderson, Silverstein, Ritz, and Jones's (1977, Psychological Review, 84, 413-451) Brain-State-in-a-Box neural model and a procedure for representing causal schemata as sets of neural activation patterns is proposed. An important feature of the resulting system is that actions are indirectly linked together through commonalities in the internal structure of situations associated with those actions. The model successfully accounts for the gap size effect in causal schemata (G. B. Bower, J. B. Black, & T. J. Turner, 1979, Cognitive Psychology, 11, 177-220), effects of causal relatedness (J. M. Keenan, S. D. Baillet, & P. Brown, 1984, Journal of Verbal Learning and Verbal Behavior, 23, 115-126), certain types of confusion errors in human memory for stories (Bower et al., 1979), and characteristics of human memory for obstacles and irrelevancies in stories (Bower et al., 1979; A. C. Graesser, S. B. Woll, D. J. Kowalski, & D. A. Smith, 1980, Journal of Experimental Psychology: Human Learning and Memory, 6, 503-515). AN University Microfilms Order Number ADG87-13615. AU WILLIAMS, PATRICK SWINNY. IN Texas Tech University Ph.D. 1987, 257 pages. TI QUALITY IN LINGUISTIC METAPHORS. SO DAI V48(03), SecB, pp903. DE Psychology, Experimental. AB This study examined a neglected question about linguistic metaphor: What are the psycholinguistic characteristics accounting for some metaphors being better than others? Metaphor quality was hypothesized to be a function of three components--comprehensibility, aptness, and novelty. Nine other variables were hypothesized to influence quality in metaphors. Subjects rated a set of constructed metaphors on these 12 variables. Correlational analyses revealed that metaphor quality was primarily a function of comprehensibility and novelty; higher quality metaphors were highly comprehensible and familiar. Metaphor quality, so defined, was found to be influenced primarily by denotative and connotative similarity between a metaphor's subject and predicate. A second experiment examined hypotheses derived from Ortony's compactness, inexpressibility, and vividness theses. It was predicted that high quality metaphors would differ from low quality metaphors by (1) being more difficult to paraphrase, (2) having topics which undergo greater connotative meaning change, and (3) being easier to recall. Differences between high and low quality metaphors on connotative meaning change were significant. In sum, the hypothesis that metaphor quality is a function of comprehensibility, aptness, and novelty was partially supported. Regarding the effects of quality level, it was concluded that metaphors which are low in quality due to extremely high or low levels of comprehensibility, aptness, and novelty differ from each other in their effects on ease of paraphrase and recall. Truly high quality metaphors, however, are hypothesized to have moderately high levels of comprehensibility, aptness, and novelty. Such metaphors differ from metaphors of moderately low quality primarily on dimensions of connotative meaning. AN University Microfilms Order Number ADG87-15237. AU SCHOEN, LAWRENCE MICHAEL. IN Kansas State University Ph.D. 1987, 81 pages. TI SEMANTIC REPRESENTATION: TYPICALITY, FLEXIBILITY, AND VARIABLE FEATURES. SO DAI V48(04), SecB, pp1180. DE Psychology, Personality. AB Previous research has demonstrated that apparently simple lexical items (e.g., piano) can be instantiated in very different ways as a function of context. Schoen (1986) gathered salience ratings of words' properties across various sentential contexts. Using a variable feature system, he described the instantiated meaning of a word as the collection of the mean salience ratings of its properties for a given context. The present study continues and expands upon this research by (1) examining shared properties of multiple exemplars (both high and low levels of typicality) from within the same category (e.g., robin and chicken), across a variety of contexts; (2) correlating salience weights of semantic properties with typicality ratings of category exemplars, (3) gathering salience ratings of properties pertaining to superordinate categories themselves (e.g., bird) and comparing these ratings to ratings obtained for individual category exemplars, and (4) exploring the merits of two different methodological procedures for gathering property salience ratings. The results of these experiments were discussed in terms of current models of semantic representation, as well as a new perspective, the variable feature approach. End of File