|
p.3 This chapter... A leading thread is the interpretation of evolution as progressive organization of information in object systems for regulation of internal behavior and control over external events.
p.14 Looking back over this swift survey of evolution from inorganic seeds to human intelligence, we ask whether evolution may be profitably viewed as the development of organized information in a system context. Intelligence is organized information in action and with a purpose; it is the directed application of information through associated means to control selected events and achieve desired ends for an object system in a specified environment... Experimental inquiry is a fallible and uncertain type of intelligence by its hypothetical nature, but one that has led, more successfully than any other method in the long run, to genuine cumulative advance in understanding and control over the areas to which it has been applied.
p.22 C. S. Peirce (1839-1914), often credited as the originator of American pragmatism, saw experimental method as the prototype of human inquiry and problem solving. The validity of experimental inquiry, he claimed, was determined by the consequences of human action based on experimental findings. Peirce saw in experimental method the most effective means available for guiding human judgment and behavior.
p.36 The family and the individual have generally felt the least impact from computers. [JLJ - probably true in 1967]
p.37 If we do not destroy one another in a nuclear holocaust, or asphyxiate ourselves in an unbridled population explosion, or enslave and exploit each other in new forms of autocratic and dictatorial tyranny, or murder one another for the sake of anachronistic ideological systems [JLJ - threats still present 50 years later...]
p.38-39 the experimental method since its inception has been almost exclusively contained within the scientific and technological domain by social institutions and by the vast inertia of authoritarian tradition at all levels of society. The advent of computer-serviced societies marks a crucial crossroad for the extension of experimental intelligence to human affairs.
p.43 a realtime computing system is one capable of following and controlling selected events in its related environment at the time they occur. [JLJ - perhaps theory involving complex strategic games played by a computer should be reconsidered as a branch of realtime computing]
p.43 the broad evolutionary trend in the development of computer-serviced societies is from the initial organization of information to its ultimate use for control. And control eventually attains its most efficient and effective form as some type of realtime control. Thus a study of realtime systems invariably is more likely to anticipate the structure, function, and some of the leading technical and human problems of computer-serviced societies.
p.43 The mounting rate of scientific and technological change... One of the prime responses to this challenge is the widespread development of computer-based realtime systems, which provide early warning in detecting problems coupled with a capability for dealing with them at the time they occur rather than after the event.
p.48 System reliability is buttressed by contingency modes for various types of emergencies. [JLJ - the reference was to a computerized banking system, but true in any type of general application]
p.53 The experimental laboratory setting leads to a strong emphasis on system performance evaluation... the diagnostic routines that direct the course of learning may be used to accumulate considerable performance data... This observation may be generalized to man-machine digital systems in that the central computer may sample measures of system effectiveness in the course of normal operations. Performance evaluation may be spun off as a byproduct of live operations.
p.62-63 The online operation is linked to realtime regulation of events in the embedding environment... In each case an appropriate and timely system response is required... The type and nature of system degradation under increasing input traffic and under the malfunctions and errors peculiar to each system is of central concern in assessing realtime system performance. Much of this book is concerned with this general problem.
p.65 We shall see, time and time again throughout this book, how experimental endeavor is not merely academic nor limited to strictly scientific circles, but lies at the heart of technological and social change.
p.85 Perhaps the greatest obstacle in the path of successful classification is the refusal to face up to the complexity of the task. Armchair deduction is not enough. Exploratory empirical induction is one of the first steps
p.116 The frequent succession of geographical reconfigurations in SAGE reflects the continuing problem of evaluating system capability and adapting existing and projected capability to changing requirements, a perpetual problem for any geographically dispersed computer-based system.
p.116 no one could have conceivably foreseen the critical events and changes that have transpired... constantly updated "best" estimates... and evolutionary adaptation to changing system needs, in the spirit of calculated risks, are as much as may be expected.
p.122 The computer performs numerous vital functions and integrates massive information services in realtime, and, in Adam Smith's words, enables "one man to do the work of many."
p.137 the common use of "realtime" to describe a system that falters in some serious manner if the computer doesn't keep up with the pace of events in the object environment.
p.161 One case in which automation was carried too far was automatic initiation of tracks. It was impossible to pack... the almost endless variety of... patterns in... data into a satisfactory automatic track initiation package. No algorithm could work as well as human surveillance teams familiar with... patterns and with local... traffic as it varies with all levels of readiness. Over-initiation swamped the system with low-priority tracks, under-initiation required excessive human effort to compensate for program deficiencies. [JLJ - IMHO, this is the same problem with computer chess over the years, the inability to distinguish between game continuations worthy of pursuing in depth and those that can be handled as a calculated risk without deeper or more thorough exploration. A human develops a diagnostic test to distinguish between the two, and the machine executes this diagnostic test. So if this is true, how exactly is the machine playing the game? Looks to me like it is just doing what it is told to do.]
p.171 Are there in existence the ideas and the knowledge that permit experimental method to be effectively used in social interests and affairs? John Dewey, The Quest for Certainty, 1929.
p.190-191 System testing is, in many respects, the climax of system development. Bona fide system testing is possible only when the system can be assembled and operated as an organized entity. The user is confident of an acceptable level of performance in his system and is willing to initiate operations only when it passes muster in a credible system test.
p.197 Field tests often become feasibility demonstrations of gross effectiveness levels of system behaviors rather than controlled experimental measurements under standardized conditions. Nevertheless, key decisions must be made on best estimates of system capability, and if evidence is lacking, subjective evaluations are manufactured to fill the gaps and attain closure.
p.198 Cost-effectiveness for testing should not be narrowly conceived in terms of immediate payoff. Testing should also be evolutionary in its aim, deliberately cultivated to plant better seeds and improve the stock of future generations of systems.
p.198 Computerization does not change fundamental system goals nor basic performance measures unless the new system is redefined or the environment is drastically modified, resulting in a related but different type of system.
p.200 The total testing process... occupies a much larger share of total design and operational effort than is commonly supposed. Usually only a small portion of the testing iceberg is made visible... Perhaps the larger share of the design, production, and installation effort is spent in informal and semiformal tests of proposed designs.
p.206-207 Can system development be conducted as a scientific enterprise? ...system development may be conceived as an evolving set of hypotheses relating system design to system performance, subject to experimental verification throughout the life cycle of the object system. This analogy states the basic condition for system development as an applied scientific enterprise; namely, the systematic application of experimental method to evolving system design and changing system performance.
p.210 The proposed philosophy for man-machine digital systems and ultimately for computer-serviced societies rests on two broad cornerstones, one for methodology and the other for human values.
p.210 The first, pertaining to method, draws heavily from two sources: experimentalism as proposed by Dewey, and planned evolution... The fusion of experimentalism and evolution into a single doctrine is called evolutionary experimentalism and is defined as follows:
Evolutionary experimentalism refers to the extension of experimental method to social systems so that the course of social action is experimentally regulated with respect to relatively long-range evolutionary goals determined by changing social requirements.
p.211 The second cornerstone of the proposed philosophy for computer-serviced societies is called "humanistic automation" and is defined as follows:
Humanistic automation is the elevation of human intelligence and the enhancement of social well-being through the use of computers for experimentally regulated evolution of social systems in response to changing human needs.
p.217 “...according to Pragmaticism, the conclusion of a reasoning power must refer to the Future. For its meaning refers to conduct, and since it is a reasoned conclusion, must refer to deliberate conduct, which is controllable conduct. But the only controllable conduct is Future conduct.” Charles S. Peirce, The Doctrine of Necessity Examined, 1892. [JLJ - but what if we are reconstructing motives for certain historical figures? We might reason what Hitler's motives were for marrying Eva Braun, shortly before their dual suicide, but our pragmatic reasoning here applies to a reconstruction of past events. So Peirce is not entirely correct here. He might claim that such a reconstruction involves writing a book, and a new interpretation of history involves future sales of that book.]
p.218-219 Heraclitus (544-484 B.C.) claimed that only change exists and is real; for example, no one can step twice into the same river. He taught that all things carry their opposites, such as life and death, and that the only reality is the state of becoming... Time, like a gas or a liquid, assumes the shape of the philosophical container into which it is poured.
p.223 real world events are unique and situationally contingent whereas science, as it has developed since the Renaissance, is primarily concerned with the pursuit of general laws that are temporally invariant. Contemporary science is temperamentally and constitutionally incapable of dealing with unique events and idiosyncratic behaviors, the stuff of which the world is made. We have built science around abstract logical classifications conjured up for the most part by the ancient Greeks... Their science was made for leisurely contemplation of universals, for after-dinner conversations among well-bred aristocrats
p.224-225 what is real time science? Real time science (a) refers to general methodology and broad empirical findings, borrowing freely from the pure sciences and from interscience; (b) it deals with temporally and situationally contingent events amenable to experimental regulation and control in a system setting; and (c) it results in an extension of human mastery over such events.
p.225 A basic tenet of the proposed approach to scientific system development is each man-made digital system be conducted as a scientific enterprise, experimentally investigating its own behavior for self-corrective adaptation to changing system conditions. This implies that each system should investigate its own real time behavior to reflect better understanding and more effective control over system events.
p.227 Realtime refers to the mode of information processing that occurs in any system that continually senses and responds to selected changes in the object environment in a manner and in time to permit the system to (a) regulate and control some aspects of ongoing events in the system and its environment while they occur, (b) within the bounds of minimal or acceptable levels of system performance, (c) as determined by experimental tests.
p.229 Realtime information processing is always provisional and hypothetical - that is, it may work and it may not work in varying degrees and at different levels in current and in future system situations. It is forever being put to test in the arena of real world events and is forever proving itself in new system experience.
p.235 Effective allocation of realtime functions presupposes a theory of real time behavior for the object system, empirical norms of real time events, initial design hypotheses for realtime allocations based on the theory and the available data, and a continued program of realtime optimization testing to verify and adjust design aspirations to empirical realities.
p.257 System capacity refers to quantitative estimates and extrapolations of system effectiveness throughout the range of expected evolutionary changes, situational contingencies, and rated load levels, expressed as testable hypotheses derived from empirical norms of system performance.
p.258 Estimates of system capacity are only as good as the existing empirical performance norms.
p.259 Evolving hypotheses relating system design to system performance serve a dual purpose at any given point in system evolution.
- The hypotheses are estimates of system performance for standardized conditions that have been directly tested or are planned to be tested.
- The hypotheses are also extrapolated estimates of system effectiveness through the range of system conditions that are not normally expected to be directly tested, but that may conceivably occur in ongoing system experience.
The evolutionary theme is prominent in the above interpretation of system capacity and in evolutionary experimentalism. In both concepts system performance data must be experimentally updated to reflect significant changes in system conditions.
p.259 In system design, as in human maturation, "the child is father to the man."
p.304 Geisler relates simulation specifically to experimental problem solving, particularly to complex problems that do not permit analytic formulation, but that may be investigated heuristically by experimentation.
p.305 war gaming as described by Paxson:
"A war game is a model of military reality set up by a judicious process of selection and aggregation, yielding the results of the interactions of opponents with conflicting objectives as these results are developed under more or less definite rules enforced by a control or umpire group" (1963, p. 1).
[JLJ - no, a war game is a contest between two sides, requiring strategic play usually in the form of sequential (or rarely simultaneous) turn taking, with objects and rules for interaction created to mimic certain aspects of actual warfare, and perhaps used as a recreation or training exercise.]
p.307 Experimentation is here interpreted in the broad sense - as any type of representational exercise of an object system, including training and operations as well as testing.
p.308 Simulation is the representational feature of experimental inquiry, and experimental testing, with its associated modes of simulation, is the essential instrumentality for regulating the course of system training and operations.
p.365 The ultimate purpose of information power is not merely to enable us to wallow around in a growing pool of information but to open up new avenues of socially useful action that will increase our mastery over our environment.
p.365 Capabilities brought about by innovations in electromechanical sensors and effectors will extend the power of man-computer teams for realtime control of selected environments.
p.365 As used in this context, artificial intelligence refers to a computer-based system capable of automatically learning from, and adapting to, an object system environment relatively independently of human influence or control. Artificial intelligence is thus a machine version of a self-organizing system partially independent of human control.
p.446 Churchman (1962) describes the "elusive experiment" as a problem-solving process characterized by iterative formulation and testing, with termination of the process occurring after a sequence of experiments reaches a satisfactory solution or a point of diminishing returns. The elusive experiment is a characterization of real time experimental behavior - a groping, trial-and-error endeavor that may lead to significant discovery and verification. Churchman comments that:
"No one seems to have tried to formalize the elusive experiment, perhaps because present-day inquiring systems in science are largely unimaginative with respect to data collection, which is regarded as a rather dull task, better delegated to less creative minds" (p. 29).
p.447-448 At this point a representative cycle of experimental inquiry in man-machine digital systems may be briefly sketched, which ties recording, reduction, and analysis into themes considered in prior chapters... As in so many other computer-supported behaviors, the cycle begins and terminates with human decisions. Human examination of system requirements leads to problem definition. Formulated problems guide development of experimental design incorporating operational hypotheses to be tested. The testing process, featuring some form of system exercising or system simulation, generates the necessary man-machine digital events to satisfy the experimental design. Data collection, including manual and automatic recording techniques, reaps the harvest from critical test events. Data reduction refines and distills this harvest; analysis consummates the process through human interpretation and evaluation of findings. Unsolved and newly discovered problems lead to a new cycle.
p.480 A powerful property of the generalized regenerative mutation illustrated in Fig. 10.5 is that it represents a paradigm of a controlled digital experiment.
p.497 Mutations may arise from any source or combination of sources including the system environment, equipment alterations, human learning and adaptation, or modifications of the computer program... This leads to deliberate cultivation and experimental selection of favorable regenerative mutations, as opposed to slower and more wasteful processes of comparatively unregulated "natural" selection, to achieve an accelerated tempo of directed system evolution.
Experimentally controlled regenerative mutations provide a fundamental and powerful technique for successful implementation of evolutionary experimentalism in man-machine digital systems.
p.507-508 In time the mounting problems of the computer-serviced society will be settled less by centralized, authoritative fiat and more directly at the source - at the grass roots of society by self-corrective, in-field evolution from open experiments that will admit provisional results subject to continual improvement.
p.512 Traditional experimental techniques are inadequate to cope with the real time requirements of man-machine digital systems... new and more powerful approaches are needed.
p.512 Why evolution? Why not use some other concept as a theoretical superstructure for man-machine digital systems? These questions go back to some fundamentals.
As indicated repeatedly, perhaps the most significant characteristic of man-machine digital systems is their ceaseless and sometimes radical rate and pattern of change. These changes, spurred externally by environmental pressures, internally by system innovation, and generally by advances in sciences and technology, result in cumulative and progressive system development. Evolutionary advance is going on anyway, at least over the long run, whether we like it or not. We have a choice to make. Shall we let evolution progress blindly or shall we regulate it? Shall we be at its mercy or shall we take firm hold of it and direct it to serve our needs? Viewed this way the problem becomes one of evolutionary control... The type of evolution espoused for man-machine digital systems is not that of inevitable social progress nor that of biological evolution. It is provisional, experimental, and adaptive, not blind or passive. It is a directed scientific and technological evolution, not the totality of human evolution.
p.513-514 Our critic might now argue that long-range planning combined with regulative testing may be adequate concepts for the control and development of man-machine digital systems. We may, therefore, profitably jettison the excess evolutionary baggage altogether. If management planning and control were sufficient to regulate system development, I would agree with this critic. But this view is only part of the total existential picture since a vast amount of system development stems from factors that management can assess and influence only partially or indirectly... system development occurs simultaneously at a vast number of interacting system levels in addition to the planning and testing levels, and that all of these must be properly taken into account for an effective system gestalt. Evolution, in contrast, characterizes the living, working system at all its levels throughout all its history more forcefully and comprehensively than planning and testing taken alone.
p.514 Evolution lends itself well to processes of real time change... Evolution simultaneously encompasses the continuity and unity of the object system, its embedding environment, the historical course, and the forward thrust of system events. It is probabilistic and contingent with regard to future events, based on differential selection of system mutations in response to open-ended system environments. [JLJ - Change is inevitable, except from a vending machine.]
p.515 Performance effectiveness is paramount in evolutionary thinking - performance in its fine aspects as well as in overall effectiveness and for sheer survival value.
p.519 Much of American pragmatism has been a protest against fixed, immutable, and absolute entities; this philosophy has provided a rallying point for the redirection of human intelligence to the central problems of uncertainty, change, and increasing human control.
p.521 It was no accident that Darwin developed his theory of evolution from worldwide field experience, as a naturalist amassing a vast array of empirical data and reasoning inductively from what he observed. He did not come upon evolution as a theoretical scientist working deductively from abstract models in the laboratory.
p.523 A hierarchy of validity exists in which more realistic simulation supersedes and displaces less realistic simulation... Since hypotheses are defined in large part by the operations used to test them, they necessarily change as their test operations are reconfigured in more realistic system exercises. There is thus a hierarchical validity structure of hypotheses - those associated with abstract and symbolic system exercises are at the bottom, whereas those verified by live system operations are at the pinnacle.
p.523-524 A cornerstone for a science of real time is the notion that natural events are potential experiments... There are a vast number of potential experiments in the flow of real time events... To create real time science... it is necessary to build a realtime information system right into the events that are the objects of the inquiring system.
p.525 The object of real time analysis is not temporalization for its own sake, but temporalization for knowledge leading to informed prediction and effective control.
p.526 Exploration and innovation are the hallmarks of a future-directed science of real time. Experimental inquiry in real time science correspondingly requires and follows a course of innovative evolution... It means that basic experimental design in realtime information systems should commit significant system resources to exploratory capability for the detection and discovery of new leads... Exploratory technique should become... a standard procedure in experimental design. Innovation is perhaps the most important product of real time science for a changing and uncertain world.
p.527 Heuristic problem-solving by man-machine or machine-only techniques, in which satisfactory solution spaces are mapped, as opposed to unique and optimal solution points, represents another broad front for real time mathematical development.
p.533 In the absence of empirical data, the initial specifications at the earliest planning stages are guesses, representing a compromise between what the system ought to do to meet its goals and what the planners think it is most likely to do in the initial configuration.
p.534 The process of evolving hypotheses continues through the life cycle of the object system, responding dynamically to internal system changes, environmental changes and fresh empirical test data.
p.540 In-field evolution emphasizes the open-ended and self-transforming characteristics of man-machine digital systems through evolution working from the situational, the concrete and the particular in system behavior... In-field evolution is a reflection of the real time process at the concrete level of ongoing system events. Existential feedback, stressing self-cultivation of system experience, is a basic mechanism for accelerating system adaptation and improvement... In-field mutations are prime proving grounds for evolutionary advance.
p.547-548 Some of the key reasons for adopting an evolutionary approach were:
- Evolution characterizes the living, working system at all its levels throughout all its history more forcefully and comprehensively than any other general concept.
- Evolution is a useful concept vehicle for describing self-corrective, self-transforming, open-ended organizations such as man-machine digital systems.
- Evolutionary concepts are well-suited to encompass a vast variety of differential and changing organizations and applications of systems...
- Evolution is a plastic theoretical concept, yielding readily to new knowledge and capable of assimilating a broad array of findings from many diverse fields.
p.549 The proposed theory of man-machine digital systems was presented and elaborated through 14 basic tenets or principles. The list of 14 propositions is as follows:
- Evolutionary experimentalism.
- Real time science.
- Humanistic automation.
- The social domain.
- The computer as a built-in system laboratory.
- Man-machine digital systems as applied science.
- Evolving hypotheses.
- The analysis system.
- The applied scientific community.
- Systematic eclecticism.
- Existential feedback.
- In-field evolution.
- Evolutionary tempo.
- Realtime control.
p.569 Creative work may be conveniently separated into the phases of problem formulation and problem solving.
p.570 wishing does not make it so.
|
