Elkind / Card / Hochberg Human Performance Models for Computer-Aided Engineering

1. Auflage 2014
ISBN: 978-1-4832-7239-9
Verlag: Elsevier Science & Techn.
Format: EPUB
Kopierschutz: Adobe DRM (»Systemvoraussetzungen)

E-Book, Englisch, 344 Seiten

Human Performance Models for Computer-Aided Engineering is a collection of papers that deals with the relationship between scientific theories of human performance and practical engineering. This collection describes the emergence of a scientific engineering paradigm that uses computational theories in computational design aids. This book also considers computational human factors such as human performance models and their application in computer-based engineering designs. This text then presents applications of these models to some helicopter flight problems. This book also explains the four requirements in programming a computer-based model of the sensory performance of a pilot as 1) prediction capability; 2) measurement capability; 3) provision of compatible computer algorithms; and 4) image driven. This collection also describes cognitive structures-aspects of the human information processing system. This text then discusses resource management and time-sharing issues that is related to competition of scarce resources, which can be predictive of the quality of information processing. This book also describes other modeling scenarios such as those predicting human errors, decision making, and shape modeling. This text can prove valuable for computer programmers, engineers, physicists, and research scientists dealing with psychophysics.

Autoren/Hrsg.

Elkind, Jerome I.

Card, Stuart K.

Hochberg, Julian

Inhaltsverzeichnis

1;Front Cover;1
2;Human Performance Models for Computer-Aided Engineering;4
3;Copyright Page;5
4;Table of Contents;8
5;Preface to the New Edition;6
6;Foreword;14
7;Preface;16
8;Part I;20
8.1;Chapter 1. Introduction;22
8.1.1;HELICOPTER FLIGHT PROBLEMS AND APPLICATIONS OF HUMAN PERFORMANCE MODELS;28
8.1.2;DETECTABILITY AND VISIBILITY (CHAPTER 5);28
8.1.3;SURFACE AND MOTION ESTIMATION (CHAPTER 8);29
8.1.4;OBJECT RECOGNITION (CHAPTER 9);30
8.1.5;HETERO-OCULAR VISION (CHAPTER 11);31
8.1.6;WORKLOAD AND PILOT PERFORMANCE (CHAPTER 15);32
8.1.7;DECISION THEORY (CHAPTER 20);33
8.1.8;MEMORY OVERLOAD (CHAPTER 16);33
8.1.9;SKILL ACQUISITION (CHAPTER 17);34
8.1.10;HUMAN ERROR (CHAPTER 19);34
8.1.11;REFERENCES;35
8.2;Chapter 2. Preview of Models;36
8.2.1;FRAMEWORK;36
8.2.2;ASSESSMENT OF MODELS;38
8.3;Chapter 3. Use and Integration of Models;42
8.3.1;DESIGN PROCESS;43
8.3.2;TOOLBOX FRAMEWORK;45
8.3.3;SELECTING TOOLS AND MODELS;48
8.3.4;ENGINEERING ANALYSES;50
8.3.5;DISCUSSION;70
8.3.6;AFTERWORD;71
8.3.7;REFERENCES;72
9;Part II;74
9.1;Chapter 4. Introduction to Vision Models;76
9.2;Chapter 5. Models in Early Vision;80
9.2.1;OVERVIEW;80
9.2.2;INTRODUCTION;81
9.2.3;WHAT IS A MODEL?;82
9.2.4;MODEL ATTRIBUTES;82
9.2.5;SPATIAL VISION;83
9.2.6;TEMPORAL SENSITIVITY;87
9.2.7;MOTION PROCESSING;88
9.2.8;SUMMARY;90
9.2.9;REFERENCES;90
9.3;Chapter 6. Models of Static Form Perception;94
9.3.1;IMAGE GENERATION;94
9.3.2;IMAGE ANAYLSIS;95
9.3.3;POTENTIAL APPLICATIONS;103
9.3.4;REFERENCES;104
9.4;Chapter 7. Structure From Motion;108
9.4.1;OVERVIEW;108
9.4.2;INTRODUCTION;110
9.4.3;MODELS;114
9.4.4;CONCLUSION;118
9.4.5;RESEARCH NEEDS: STRUCTURE FROM MOTION;119
9.4.6;REFERENCES;122
9.5;Chapter 8. Motion-Based State Estimation and Shape Modeling;125
9.5.1;INTRODUCTION AND SUMMARY;125
9.5.2;FRAMEWORK FOR MOTION-BASED STATE ESTIMATION AND SHAPE MODELING;127
9.5.3;REVIEW OF RESEARCH IN MOTION-BASED STATE ESTIMATION AND SHAPE MODELING;131
9.5.4;MODEL APPLICATIONS AND LIMITATIONS;138
9.5.5;FUTURE RESEARCH;140
9.5.6;REFERENCES;142
9.6;Chapter 9. Real-Time Human Image Understanding in Pilot Performance Models;145
9.6.1;THEORIES OF OBJECT RECOGNITION;146
9.6.2;MODEL-BASED MATCHING: LOWE'S SCERPO AND ULLMAN'S ALIGNMENT MODELS;151
9.6.3;PERCEPTION OF MULTIOBJECT DISPLAYS;157
9.6.4;REFERENCES;161
9.7;Chapter 10. Manipulation of Visual Information;163
9.7.1;SUMMARY;163
9.7.2;INTRODUCTION;164
9.7.3;TRANSFORMATIONS ON INFORMATION PRESENTED IN A STATIC VISUAL DISPLAY;165
9.7.4;MEMORY FOR POSITIONS IN A SEQUENCE OF STATIC DISPLAYS;169
9.7.5;EXTRAPOLATION OF PERCEPTUALLY DRIVEN SPATIAL TRANSFORMATIONS;170
9.7.6;JUDGMENTS OF OBJECT STRUCTURE FROM PARTIAL VIEWS;172
9.7.7;FUTURE RESEARCH;173
9.7.8;REFERENCES;175
9.8;Chapter 11. Combining Views;178
9.8.1;INTEGRATION OF SUCCESSIVE VIEWS;178
9.8.2;BINOCULAR COMBINATION;181
9.8.3;REFERENCES;182
9.9;Chapter 12. Afterword;185
10;Part III;188
10.1;Chapter 13. Introduction to Cognition Models;190
10.2;Chapter 14. Cognitive Architectures;192
10.2.1;SYMBOLIST ARCHITECTURES;193
10.2.2;CONNECTIONIST MODELS;195
10.2.3;REFERENCES;197
10.3;Chapter 15. Resource Management and Time-Sharing;199
10.3.1;OVERVIEW;199
10.3.2;SERIAL ALLOCATION;201
10.3.3;PARALLEL ALLOCATION;206
10.3.4;SERIAL COMPETITION;207
10.3.5;PARALLEL COMPETITION;208
10.3.6;SYNTHESIS OF THE OPTIMAL MODEL;212
10.3.7;CONCLUSION;216
10.3.8;REFERENCES;217
10.4;Chapter 16. Models of Working Memory;222
10.4.1;PHENOMENA OF WORKING MEMORY;223
10.4.2;MODELS OF WORKING MEMORY;227
10.4.3;REFERENCES;231
10.5;Chapter 17. Training Models to Estimate Training Costs for New Systems;234
10.5.1;OVERVIEW;234
10.5.2;SKILL DEVELOPMENT;235
10.5.3;MODELS FOR PREDICTING HUMAN PERFORMANCE;239
10.5.4;ENGINEERING GUIDANCE WITHOUT AN ALL-INCLUSIVE MODEL;245
10.5.5;USE OF RAPID PROTOTYPING AND QUICK EMPIRICAL EVALUATIONS;246
10.5.6;NEEDED RESEARCH;247
10.5.7;REFERENCES;248
10.6;Chapter 18. Modeling Scenarios for Action;252
10.6.1;FIXED SCENARIOS;252
10.6.2;SCENARIOS WITH SIMPLE CONTINGENCIES;259
10.6.3;MODELING MORE COMPLEX SCENARIOS;261
10.6.4;REFERENCES;264
10.7;Chapter 19. Modeling and Predicting Human Error;267
10.7.1;INTRODUCTION;267
10.7.2;ERROR MODELING;272
10.7.3;REFERENCES;288
10.8;Chapter 20. Modeling Decision Making for System Design;294
10.8.1;WHY DECISION MAKING SEEMS EASY TO MODEL—SOMETIMES;295
10.8.2;IMPLICATIONS FOR MODELING OPERATOR PERFORMANCE;297
10.8.3;MODELING WITHOUT OPTIMALITY;300
10.8.4;MAKING BEHAVIOR MORE MODEL-LIKE;303
10.8.5;TESTING THE LIMITS OF DECISION MAKING;305
10.8.6;REFERENCES;306
10.9;Chapter 21. Knowledge Elicitation and Representation;310
10.9.1;KNOWLEDGE ELICITATION;310
10.9.2;KNOWLEDGE REPRESENTATION;313
10.9.3;MENTAL MODELS AND DESIGN DECISIONS;316
10.9.4;REFERENCES;317
10.10;Chapter 22. Afterword;318
10.10.1;REFERENCE;319
11;Part IV;320
11.1;Chapter 23. Findings and Recommendations;322
11.1.1;DESIRABLE ATTRIBUTES AND TYPES OF MODELS;323
11.1.2;ADEQUACY OF MODELS FOR THE A 3I DESIGN FACILITY;324
11.1.3;VALIDATION;325
11.1.4;NEED FOR ACCESS TO HUMAN FACTORS DATA BASE;325
11.1.5;BROADER CONTEXT OF COMPUTATIONAL HUMAN FACTORS;326
11.1.6;IMPORTANCE OF THE SYSTEMS DESIGN CONTEXT FOR RESEARCH ON MODELS;326
11.1.7;FOCUSING THE A 3I PROGRAM;327
11.1.8;PROVIDING A FRAMEWORK AND A BOX OF TOOLS;328
12;Index;329

Leseproben

1 Introduction
Publisher Summary
This chapter presents an introduction to models of human performance and their use in computer-based engineering. It focuses on a particular human factors design problem, which is the problem of design of cockpit systems for advanced helicopters, and on a particular aspect of human performance, its vision, and related cognitive functions. A model is a representation or description of all or a part of an object or process. There are many different types of models, and they are developed for a variety of reasons. Models of human performance have long been used in the human factors design of complex systems to answer questions about the ability of the human to function satisfactorily in the system and about the ability of the system to achieve the objectives for which it is being designed. Modern computer technology is changing this situation and is making it possible to develop models of much greater complexity and comprehensiveness that can represent human performance with greater depth and breadth than was possible with earlier modeling technologies. This report discusses a topic important to the field of computational human factors: models of human performance and their use in computer-based engineering facilities for the design of complex systems. It focuses on a particular human factors design problem—the design of cockpit systems for advanced helicopters—and on a particular aspect of human performance—vision and related cognitive functions. By focusing in this way, the authors were able to address the selected topics in some depth and develop findings and recommendations that they believe have application to many other aspects of human performance and to other design domains. The report is addressed to human factors professionals and others interested in human performance models, human factors design methodology, and design tools. It describes some of the key vision-related problems of helicopter flight and cockpit design as a way of introducing the reader to the design domain on which the report is focused. It discusses issues in the integration of models into a computer-based human factors design facility and the use of such a facility in the design process, and it reviews existing models of vision and cognition with special attention to their use in a computer-based design facility. It concludes with a set of findings about the adequacy of existing models for a computational human factors facility and a related set of recommendations for research that is needed to provide a stronger foundation of models upon which to base such a facility. A model is a representation or description of all or part of an object or process. There are many different types of models and they are developed for a variety of reasons. In a design context, models can be considered to be a “thing” of which we ask questions about some aspect of a design. Models of human performance have long been used in the human factors design of complex systems to answer questions about the ability of the human to function satisfactorily in the system and the ability of the system to achieve the objectives for which it is being designed. Early models used for human factors design were of necessity verbal, statistical, or mathematical descriptions or theories of some limited aspect of human performance. Sometimes they were narrow, sometimes broad, sometimes shallow, sometimes deep, but rarely broad and deep. It was not possible to cope with the complexity of comprehensive models of human performance. Modern computer technology is changing this situation and is making it possible to develop models of much greater complexity and comprehensiveness that can represent human performance with greater depth and breadth than was possible with earlier modeling technologies. The impact of computer technology on modeling is sufficiently profound to warrant a distinctive name, computational models, for models that use this technology. Computational models are simply those models that can be, or have been, implemented on a computer. Important properties of computational models are (1) they can be constructed of component models assembled from different sources; (2) they can be integrated into a computer-based facility that provides users with the ability to manipulate these models and apply them to design in a very flexible manner; (3) the models can be made to run using either real or simulated data from the physical environment and thereby provide a simulation of the part of pilot performance being modeled; and (4) since designs can also be represented in a computational form, they can coreside with the computational models and tested against them in a common facility. Computer technology has also fostered new forms of computational models of human performance in the fields of artificial intelligence and cognitive science. Computational models take many forms, their taxonomy has many dimensions, and it can be structured in many ways depending upon the perspective. For a computer-aided engineering facility, it is of interest to know whether one has a simulation model or an analytic model. Simulation models typically use a computer program to imitate human performance. They are of particular interest for the design of aircraft systems because they allow connection to the physical environment in which the human is to function, have explicit inputs derived from the environment and produce outputs that approximate those produced by the human. These can be fed back to the environment so that the closed-loop behavior of the pilot-vehicle system can be simulated and examined. Analytic models represent human performance mathematically, typically in terms of algebraic or differential equations. Both the form of the equations and their parameters are of interest to the psychologist and the designer. Analytic models often provide concise descriptions and even “laws” governing human behavior that are of enormous value in the design process. Some models attempt to represent specific human processes, usually by simulation, and as a result are known as process models. Others attempt to predict only human output without claiming to be good representations of the human processes involved, and are known as performance models. Models of the processes used by the human to accomplish the task under study are more powerful than those that just describe the observed external behavior (outputs) because they are more likely to be applicable to a wider range of tasks and conditions. Most models in the literature are descriptive in the sense that they were developed to describe observed human behavior, performance, or processes. A few, however, are prescriptive in the sense that they prescribe how the human should perform if he were to behave in a rational way that takes into account the information available, the constraints that exist, the risks, rewards, and objectives. Some rational models are based on strong theories of optimality, such as those that have been developed in the fields of control, decisions, and signal detection, and are known as ideal observer or ideal operator models. We will often refer to prescriptive models as rational action or normative models. Until fairly recently, most human performance models were numerical or quantitative and lent themselves to classical, numerically-based computation. As a result of progress in artificial intelligence and cognitive science, a substantial body of non-numerical, qualitative, but calculable models, has been developed. These models are necessary for representing cognitive behavior. Although they are qualitative, they are computational and, as such, are amenable for inclusion in a computer-based engineering facility. Many of the reviews of vision and cognition in this report address qualitative models. Models can represent behavior at different levels and with different amounts of detail. There are mission-level models that attempt to encompass the whole mission or major mission segments by representing human behavior at a high level of abstraction. Such models are concerned, typically, with issues such as the workload on the human operator. Models can address entire human subsystems, such as vision or motor control, or be focused on a part of a complex task or of a human subsystem. There is the goal of building models that tie together detailed models of several human subsystems to obtain a “complete” representation of human behavior in a complex system. However, most comprehensive models contain little detail about specific aspects of human performance, reflecting the harsh reality of the trade-off of breadth against depth. Most existing models of human performance were developed with a simple task in mind, but there have been numerous efforts to build more comprehensive models that attempt to represent more complex behavior, often by assembling and integrating simple task models within a uniform framework. As a result of decades of research, a large collection of models now exists for many aspects of human perceptual, motor, cognitive, and biomechanical performance. The extent to which these simple task models can be usefully integrated to represent more comprehensive behavior depends upon the nature of the gaps in the coverage of the models and on the completeness of the linkages among them. Both of these problems are addressed in the reviews of models in Parts II and III of this report. Much of the progress in modeling that has occurred in recent years has been due to the remarkable increase that has occurred in...