CYBERNETICS FORUM - univie.ac.at · CYBERNETICS FORUM A PUBLICATION OF THE AMERICAN SOCIETY FOR...

CYBERNETICS FORUM

A PUBLICATION OF THE AMERICAN SOCIETY FOR CYBERNETICS

SPRING 1975 VOLUME VII NUMBER 1

IN THIS ISSUE

Editorial: Men of Indostan and Other Places

Articles:

Toward Quantification of Conscious/Intelligent Processes in At Least Some Cybernetic Systems

Probe-Brain Research in Automata Theory: A Need for Application

Cybemetics at the University of Massachusetts at Amherst

Reverberations: Mind: The Software of the Brain

Book Review

Features:

From the Desk of the President.

ASC FORUM Notes Letters to the Editor Conference Calendar About the Authors

CODEN-ASCFC

Milton S. Katz

Frederick Kile

Richard E. Schneider

Michael A. Arbib

Roland Fisher

Christopher Longyear

1

3

6

9

15

17

2

22 . 24 . 25 . 27

Charles I. Bartfeld School of Business Admini

stration, American University Mass. & Nebraska Ave. NW Washington, D.C. 20076

N. A. Coulter, Jr. Oepartment of Surgery

Curriculum in Biomedical Engineering

University of North Carolina School of Medicine

Chapel Hilf, N.C. 27574

V. G. Drozin Foreign Correspondents Oepartment of Physics

Bucknell University Lewisburg, PA 7 7837

Charles H. Dym Oym, Frank & Company 7875 Connecticut Avenue, N. W. Washington, D.C. 20009

BOARD OF EDITORS Editor

Milton S. Katz Computer Systems Department, The Mitre Corporation

Westgate Research Park McLean, VA 22101

Associate Editors

Roland Fischer Maryland Psychiatrie Research

Center Box 3235 Baltimore, MD 21228

Gertrude Herrmann Conference Ca/endar Editor 77 3 7 University Boulevard West,

#2722 Si/ver Spring, MD 20902

Harold K. Hughes The State University College Potsdam, NY 73767

Akira lshikawa Graduate Schoo/ of Business

Administration, New York University

700 Trinity P!ace New York, NY 70006

Frederick Kile Aid Association for Lutherans App!eton, Wl 5497 7

Felix F. Kopstein 7 9 7 3 Walnut Street Philadelphia, PA 79703

J ulius Korein EEG Laboratory, Bellevue

Hospital Oept. of Neuro!ogy, New

York Un iversity Medical Center

550 Fifth Avenue New York, NY 70076

Robert M. Landau S!A Science Information

Association 3574 P!yers Mi/! Road Kensington, MD 20795

Christopher Longyear Book Review Editor Department of English

University of Washington Seattle, WA 98795

Mark N. Ozer Oepartment of Child Health

Development Children 's Hospital National Medica/ Center

The George Washington University School of Medicine and Health Seiences

3000 Connecticut Avenue N. W. Washington, D.C. 20008

Doreen Ray Steg Departmen t of Human

Be havior & Development, Drexel University

Philadelphia, PA 79704

Paul Studer Schoo/ of Library and Infor

mation Science, State University College of A rts and Science

Geneseo, NY 74454

ASIS Liaison Managing Editor

jack Lass Laurence B. Heilprin University of Maryland

4800 Berwyn Hause Road College Park, MD 20740

Roy Herrmann, President

OFFICERS - 1974

American Society for Information Science 7755 76th Street, N.W.

Washington, D.C. 20036

Gary D. Bearden, Vice President (Administrative) Bonnie W. Dunning, Vice President (Educational) Felix F. Kopstein, Vice President (Technical)

Kenneth A. Morris, Vice President (Awards) Herbert W. Robinson, Vice President (Publications) Frank S. Speck, Treasurer Carlis A. Taylor, Secretary

Carl Hammer, Chairman of the Board Roy Herrmann, President Charles H. Dym Eleanor L. lson Franktin

DIRECTORS - 1974

William E. Hanna, J r. Daniel Howland Douglas E. Knight Kumpati S. Narendra

Corporate Member Advanced Computer T echniques Corp.

237 Madison Avenue New Y ork, NY 1 0022

Mark N. Ozer Louise B. Speck Murray Turoff Heinz Von Foerster

Annual dues of the American Society for Cybernetics are $20.00 ($5.00 for students) of which $5.00 ($3.00 for students) is in payment of the regular annual subscription price to the ASC CYBERNET/CS FORUM. Special non-member subscriptions to the CYBERNETICS FORUMare $18.00 per year ($15.00 to members of ASIS) domestic, and $25.00 per year foreign, postpaid in the U.S. single copies may be purchased at $5.00 each. Copyright © 1974 American Society for Cybernetics.

The ASC CYBERNETICS FORUM is published quarterly by the American Society for Information Science for the American Society for Cvbernetics. Editorial Offices: ASC c/o Suite 911, 1025 Connecticut Avenue, Washington, DC 20036. Production offices: ASIS, Suite 210, 1155 16th Street, N.W., Washington, DC 20036.

-~---

Editorial Men of lndostan and Other Places

It was six men of Indostan To learning much inclined,

Who went to see the Elephant (Though all of them were blind),

That each by observation Might satisfy his mind.

-John Godfrey Saxe

At least 17 publications are distributed throughout the world which have in common the single-word title, "FORUM," according to Ulrich 's International Periodicals Directory (15th Edition, 1973-74). Many more periodicals contain the word Forum in their titles. The Directory does not yet include the ASC FORUM, but that is not surprising because we took our present form only last Summer.

In all, ten countries, five of the United States, and the District of Columbia host the publication of a FORUM. The array of topical content among the 17 FOR UM's is nearly as diverse as their geographic origins. Here is how they line up:

Canada-Insurance England-Human Relations and Psychosexual Studies England-Business and lndustry Germany-Children and Youth lndia-General Interest N etherlands-Architecture Pakistan-Literary and Political Reviews Po land-Literary and Political Reviews Switzerland-Literature Turkey-Literary and Political Reviews Yugoslavia-Literature

and, in the United States: Connecticu t-Local Interest District of Columbia-Law Illinois-Law Nevada-College Student's Affairs Pennsylvania-Ethnic Interests (Chicano)

The Netherlands may have the best claim of all to the title since it carries the double reference to an architectural entity, the Forum of ancient Rome, and to another dictionary meaning, "any medium for open discussion."

The law publications, on the other hand, may have an equal claim, with Forum denoting a court of law or tribunal, and connoting the judicial activities which were carried on in the Fora of ancient Roman cities. In modern terms, however, anyone may feel free to use the title, FORUM, provided that a public medium or meeting place for open discussion is intended. We prefer the modern usage because, among other reasons, it no Ionger implies meetingout of doors as did the ancient Fora.

To distinguish ourselves, at least in name, from all the other FORUM's, we commence calling ourselves the ASC CYBERNET/CS FORUM with this issue. However appropriate it might have been to keep as non-committal a title as FORUM, our range of subjects and interests impinges on, and extends far beyond, the preoccupations of all the other FORUM's listed in Ulrich's Directory. Articles in just our first three issues span major topics including: clinical medicine; experimental psychology; economics; the information industry; scientific research policy; sociology; and neuropsychology. Still other disciplines are intermingled, and more will be featured in coming issues.

Michael Arbib's lucid and enticing description (see p. 9) of cybernetics at the University of Massachusetts (don't we wish we were students again? -and how lucky are those who are!), and Heinz von Foerster and Richard Howe's feast of "Cybernetica at Illinois" in past issues, give intimations of our manifest destiny. As we continue through the seasons to serve up the orientations and activities of still more centers of cybernetics, a fuller picture of our scope and depth will emerge. The more the collection of articles we publish grows, the more evidently will our cyclopean point of view penetrate and illuminate those numerous and multifarious disciplines which line our vistas.

Those of us who were privileged to attend the ASC Tenth Anniversary Conference on Communications and Control in Social Processes at the University of Pennsylvania last November (see ASC Notes, p. 22) heard distinguished addresses and engaged in stimulating colloquia on a large variety of subjects more or less deeply touched by cybernetics. Yet the aggregation barely overlaps the programs of study and research at Massachusetts or Illinois.

Are we all, then, men of Indostan, journeying ... ?

Milton S. Katz

ASC CYBERNETICS FORUM Valurne VII, Number I-Spring 1975

From the Desk of the President

I am pleased to report that our efforts over a two-year period have been crowned with success and the American Society for Cybemetics has been accepted by the American Association for the Advancement of Science (AAAS) as an affiliated member. (See Letters, p. 26.) This membership if used judiciously, should be of no little value, because "big brother" is prepared to provide assistance in many forms and situations to smaller organizations which are affiliated with AAAS.

* * * *

Some further good news: During 1974, the ASC FORUM has developed into a full-fledged periodical and attracted new members and subscribers. The Board of Editors makes every effort to publish interesting and challenging subjects, including tutorial papers. Most of the new membership applications we have been receiving in a steady flow came from university instructors, researchers and others equally qualified.

* * * *

A survey of association policies, particularly those which affect administration and publicity, indicated that the methods and procedures applied to implement the decisions passed by the policy making bodies depend to a large degree upon the size of an organization. Although no finite number, such as break-even point or some other measure of effectiveness, has been established which could be applied, it appears reasonably safe and logical to consider the size of the registered membership, as they are utlimately responsible for passing or rejecting the decisions of the ruling board.

In the Winter 1974 edition of the ASC FORUM one of our esteemed members of the Board of Editors has pointed a finger at the lack of devotion to the cause on the part of members (See Editorial, Cybemetics and Self-Organization, p.l.) In essence he may be correct. In the real world with which we have to cope it is difficult, if not impossible, to get "volunteers" to take all the time needed to help run the operation. In the case of a small society ("small" in terms of the number of active members) administrative tasks can only be performed by volunteers for very obvious fiscal reasons. As soon as an association develops beyond a certain size, volunteers can no longer be relied on to do the job, and a profes-

sional, salaried staff must take over some of the routine duties.

The alternative, applicable to border-line cases, is usually a temporary solution which relies on one or two "volunteers" doing most, if not all, of the work. This is scarcely a fair distribution of the duties and responsibilities which require more than just attention. This is a serious problern for ASC which must be solved in principle before the new slate of members of the Board of Directors and Officers can be elected and, when elected, can take office. The Nominating Committee, which is completing its recommendations for the coming elections, is aware of this problern and will discuss this issue with those candidates who are willing to serve.

* * * *

In the past fiscal year, there have been changes within the membership. Some old members have dropped out, mostly for economic reasons, and another group has not yet paid their dues. These are dropouts and for the most part repeaters whose names appear every year as stragglers on the last list of Final Notices. Unfortunately, it costs money to send out reminders-postage has gone up and so have all other costs. It will therefore be necessary to take all members not in good standing off the membership list as of 1 April 1975, much to our regret. In order to be reinstated the individual will be charged an extra fee of $2.00 to reimbUTse the Society for the extra expense to process these late comers.

* * * *

For reasons mentioned above, OUT membership drives have been unsystematic and poorly coordinated. This we intend to change. Under the direction of the Chairman of the Membership Committee an active membership program is to be developed. A target of 200 new members is feasible and should not only be achieved but also exceeded with the aggressive assistance of OUT officers and membership. One of the first duties of the incoming Membership Committee will be the streamlining of all recruitment materials. It is generally accepted policy, even in these troublesome tim es, to plan promotional campaigns carefully and carry the message of ASC to every potential member, individual and corporate.

2 ASC CYBERNETICS FORUM Volume VII, Number I -Spring 1975

T oward Quantification of Conscious/lntelligent

Processes in At Least Some Cybernetic Systems

Frederick Kile Aid Association for Lutherans

Appleton, WI 54911

A Word on Models

Present discussion of the utility and social significance of computer-based models seems only slightly more organized than random shouting. This paper offers a few definitions and criteria for measuring consciousness and intelligence in computer models and some other cybernetic systems. Hopefully, these suggestions will help to provide a focus for the discussion, at least by giving vociferous critics something to shout at.

Two types of computer models offer ready-made points of contact with current research:

1. Models of socioeconomic systems 2. Models of intelligence.

We must assume a large body of shared semantic and contextual knowledge, but some definitions will serve as convenient reference points for further discussion.

Consciousness

Let us characterize consciousness as that aspect of problem-solving activity which defines the problern to be solved (many relevant insights can be found in the article by William T. Powers in the Winter 1974 issue of FORUM) . This definition affords a sliding scale with which to measure consciousness: e.g. , it might be said that an amoeba which detects a particle of nutrient has defined a problern (the need to ingest the nutrient) and is therefore "conscious" at some level; it might also be said that an "IF" statement in a computerprogram is an element of consciousness since it defines a problern (albeit short-lived) for the computation process to solve; in this sense people, social groups, and computer models are all characterized by varying levels of consciousness .

This definition allows partitioning of all forms of consciousness into two types of sets, finite and infi-

nite , according to the number of problem-defining elements involved. It is convenient to introduce an intermediate dass of consciousness called semi-infinite and define as semi-infinite those conscious processes which vary by approximately one or two orders of magnitude from the consciousness potential of the human brain. The inspiration for this fuzzy definition comes from the embryonie discipline called fuzzy-set theory.

According to our definition, even this much of our (author-reader) mutual effort at communication is consciousness, but let us extend our effort to the domain of computer-based models.

Artificial I ntell igence

It is convenient to introduce another definition at this point. We will define as intelligent those aspects of problem-solving activity which follow logically the problem-defining activity called consciousness. By analogy with our classification of conscious activity we partition intelligent activity into three types of sets: finite, semi-infinite, and infinite . We can call the nutrient-ingesting activity of an amoeba intelligent. This intelligent act of nutrient ingestion follows the conscious act of nutrient detection . We can also classify as intelligent that computation or manipulation of data within a computer program which follows an IF statement. There is obviously intelligent activity by people, social systems and computer models.

We have now defined some consciousnessfintelligence qualifiers:

1. Consciousness = problem-defining 2. Intelligence = problem-solving dependent on

consciousness 3. Consciousness and intelligence partitioned into:

a. Finite b. Semi-infinite c. Infinite

ASC CYBERNETICS FORUM Volume VII, Number I-Spring 1975 3

Given these presuppositions we can further state: Sorne cornputer rnodels act consciously and intelligently.

Now let the context for discussion be further clarified by the following observations:

l. Intelligent action is defined only within the frarne of reference of consciousness and intelligent action is not systern-destabilizing, i.e., within the designated frarne of reference it seeks to solve the defined problern (or reduce the perceived error signal).

2. Intelligent use of rnodels depends on the rnodeler's consciousness (problern definition) within an appropriate frarne of reference (rneta-systern) as well as on the internal in telligence of the rnodel.

Time Frames of Validity

Addition of another qualifier will provide a rneaningful link between consciousness and intelligence. We will call this qualifier the time frarne of validity.

A conscious elernent of a systern defines a problern. A second conscious elernent in the sarne systern rnust define the proper time frarne for intelligent action in solving the problern. This time frarne is defined either by setting a type of binary switch which allows the systern to proceed only so far without responding to the particular problern definition or the time frarne is defined by setting an analog value which gradually decays (or gradually increases) until the appropriate intelligent problern-solving action ensues. lf there is not sorne sort of tirne-frarne rnechanisrn systern activity will either stop or degenerate into a sort of Brownian rnotion, i. e., the systern would becorne totally disorganized and collapse.

For exarnple, if an arnoeba detects a nutrient rnass in contact with its surface this unicellular systern rnust act intelligently by cornpletely ingesting the nutrient, and by doing so within a certain time lirnit, or it will die. This is no new wisdorn, having been perceived by Aesop who told the story of the ass which starved to death rnidway between two haystacks. Aesop's ass had no tirne-frarne rnechanisrn to cornpel systern action within the proper lirnits.

Sirnilarly, in a simple cornputer prograrn, certain IF staternents are followed by one of two (or three) staternents depending on the prograrn's perception of the outcorne of the IF. In this case, the cornputer's rnaster clock and operating systern provide for the program to function at specified tirnes and the program provides for a specific action. This arnounts to a binary switch which dictates that program action will not proceed until the relevant condition is rnet. The time frarne is essentially zero, i.e., the required action is next and nothing else will occur until the action is cornpleted.

Multiprocessing/rnultiprograrnrning systerns . provide for continuing action in sorne solution paths even

while one path is holding. The human individual and the social systern provide for a serni-infinite nurnber of actions while a particular action is holding. This leads to the generalization that the positive direction along the axis of increasing intelligence can be characterized by the increasing capacity of a systern to carry on multiple actions while multiple other actions are holding.

Now we can add another qualifier of intelligence: the capacity for a systern to change both its consciousness, i.e. , redefine the problern, and its tirne-frarne definition, i. e., reset a binary switch or analog tirnefrarne level for a holding action while the rest of the systern is operating. We now have three additional consciousness/intelligence qualifiers:

4. The nurnber of simultaneaus actions a systern is capable of

5. The nurnber capable of proceed

of holding actions a systern is while other systern functions

6. The capacity of a systern to redefine a problern while an action is holding and the capacity of a systern for relevant tirne-frarne redefinition while an action is holding.

The "Stupid" Computer

Our list of consciousness/intelligence qualifiers, together with our clarifying Observations, helps us to see why so rnany critics have adarnantly insisted that cornputers/cornputer prograrns are "stupid." Both software and hardware developrnents thus far have been very low on the intelligence scale we suggest here, i.e., present-day cornputer prograrns are essentially unable to:

1. Alter consciousness (redefine the problern) of an action while that action is holding

2. Update the time frarne of validity (reset the governing binary switch or analog value) for an action while that action is holding.

We can add a further set of criteria before pausing to suggest a tentative quantification. lf any cornbination of characteristics would allow us to judge that a finitely-conscious/finitely-in telligent systern can be reclassified as belonging to the category of the serniinfinite, these characteristics would include:

l. The systern is capable of redefining a problern and redefining the time frarne for action on that problern while the particular action is holding;

2. The systern is capable of processing multiple concurrent holding actions while multiple systern functions are proceeding; and

3. The systern is capable of redefining the rnode of problern-solving action (for exarnple, reassigning variables or coefficients in the relevant solution equation), and of redefining the problern tirne-frarne for action while that action is holding.

4 ASC CYBERNETICS FORUM Valurne VII, Nurnber i-Spring 1975

By now the exasperated critic might say, "Why stop here? Why not define as still more intelligent, possibly even as infinitely intelligent (ultra-intelligent), the system which could alter its problern definition, its time-frame definition , and its problem-solving action while the problern solving action is taking place?"

There are both practical and metaphysical reasons for halting this line of argument, which could easily become a sort of chain-letter of reasoning.

1. A system which is unable to fix a problern definition and move toward resolution of that problern in a clearly-defined frame of reference could be viewed as merely random and of no further interest

2. Gödel proved that no set of logical operations, and thus by extension no system, is totally self-contained (without prior assumptions or outside referent), and thus at least one metastaterneut or meta-system must be presupposed

3. Even metaphysical speculations such as the Indian tradition that the world is on the back of an elephant, that elephant on another, and so on, provide no answer to where the bottarn elephant stands or even whether there is a bottarn elephant .

favor the first of these reasons because it is most consistent with the contextual frame of reference for our discussion of consciousness/intelligence.

ldeas for Quantification

Any function which quantifies consciousness/intelligence in a cybernetic system is necessarily arbitrary . I would like to suggest that if C is the consciousness/ intelligence level of some cybernetic system and:

P = log(number of problem-definition elements of C)

Q = log(number of problem-solving elements of C)

R = log(number of problem-redefining elements of C)

S = log(number of time-frame redefining elements of C)

T = a constant, we can obtain an approximate measure of C by an equation of the following form:

C = T(R+S)(P·Q).

Comments on Socioeconomic Simulations

Simulations are run in real time or in compressed time. Real-time simulations are run in parallel with events in the problern set of interest. Other models are run in compressed time, ice., faster than the events themselves. We are especially interested in models of the latter type which use multiple feedback loops in an attempt to forecast at least a qualitative response to a

system perturbation and , hopefully, even a quantitative response.

A qualitative forecast from this type of model would indicate the direction of a particular change in response to an input of known direction. For example, a model of the qualitative type might be designed to indicate whether the direction of change in human nutrition levels for the poorest ten percent of humanity would be positive or negative as a result of higher energy prices charged by the OPEC oil cartel. This qualitative model might, for example, indicate that the direction of change in nutrition would be downward for a while and then upward because of the increased flow of oil reserves from cartel states as aid to poor nations or the model might show that the direction of change in human nutrition for the poor ten percent of humanity would be downward initially and more sharply downward after a period of time due to lower economic activity in major food producing areas.

Linear reasoning now leads us to conclude that intelligent use of a forecasting model requires the modeler to:

1. Place the model in a clearly defined contextual frame of reference

2. Be conscious of the multiple, non-coincident time-frames of validity of assumptions, both contextual and within the model

3. Refine data and equations according to an "intelligent" process

4. Advise non-technical users regarding both conscious/intelligent aspects of the model and non-conscious/non-intelligent ("stupid") characteristics of the model.

Comments on Artificial I ntelligence

The researcher in artificial intelligence is obliged to: 1. Specify the problem-definingfproblem-solving

context of the artificially-intelligent system; 2. Specify the level of problem-definition and

problem-solving action sought; 3. Define and solve the problern of time-frame

definition , both for model context and within the model;

4. Advise non-technical analysts and users regarding both intelligent and non-intelligent aspects of the artificially-intelligent system;

5. Work toward development of a multiply-defining/multiply-acting system with multiple holding and with redefinition , re-timeframing capability .

If this paper has contributed anything to intelligent debate, it has done so by indicating the size of the task and by indicating at least one sine qua non of intelligence in computer-based systems: time-frame definition. If we have achieved anything beyond this, it is to indicate how quantification can be used to measure progress in development of computer-based, conscious/ intelligent cybernetic systems.

ASC CYBERNETICS FORUM Valurne VII , Number 1- Spring 1975 5

Probe Brain Research in Automata Theory:

A Need for Application

Richard E. Schneider The MITRE Corporadon Westgate Research Park

McLean, Virginia 22101

Basic research has been slipping steadily from its once exalted, if poorly understood, public station. The cry from the public, in the Congress, and even among scientists, grows lauder and more persistent for research directed to goals which will benefit the material welfare of society. Mr. Schneider raises the issue for one area of cybernetics- Automata Theory. The same considerations apply in any research field.

Implicit in this probe are two questions: why basic research? ; and, how can we best realize the benefits of targeted research? To answer the first, we need articulate proponents who will do for basic research what Paul deKruif and others did for medical research. For the second question, we need to originate, examine, enlarge on and perfect action proposals such as Mr. Schneider makes here.

-Editor

The Need

Some automata theorists apparently feel that Automata Theory is purely mathematical and that problems in this field should be studied simply because they exist. This is one reason that Automata Theory as a whole, and neural network studies (i.e., brain theory) specifically, have few practical applications to show for a field in which so much research has been conducted and so many papers written.

Increasing pressure is building to translate more theoretical work, in all fields, into applications which will improve the quality of life . A recent report by the National Academy of Engineering stated that Federal Agencies spent $17 billion on research and development during fiscal 1973. Further, $1 billion of this was earmarked for the transfer and dissernination of technological information, when in fact only $43 million

(less than 5%) was actually spent to transform developed technology into products, processes or services other than those originally intended. The Academy deplored the trend of Federal R&D, which is away from improving mankind's lot in life.

The Problem

Preservation of life and associated good health head the list of subjects comprising the concept of the quality of life. Some of the most dreaded threats to this preservation of life are brain diseases such as polio, Huntington 's disease, epilepsy, Parkinson's disease, multiple sclerosis, cerebral palsy, etc. Another threat is a simple blood clot in the brain which can cause a strake, resulting in paralysis or death (strokes were the third leading cause of death in the U.S. in 1973 according to the National Safety Council). Symptoms of these conditions have been known for years but progress toward prevention and correction are bounded by our lack of knowledge of brain function. Exploratory surgery is seldom performed in the brain because of this lack of knowledge and our apprehension about the delicacy of this highly sophisticated, finely tuned mechanism. Very slight alterations in the brain can engender remarkably large external changes in such complex functions as personality, speech and sight, mental capacity and motor coordination.

Short-term fixes to problems may be found even if the problems are in the brain. For instance, drugs and surgery have been used to relieve symptoms, often only tt'mporarily or partially. But long-term solutions to brain problems are inextricably linked to our knowledge (or, rather, our lack of knowledge) of how the brain operates. Thus, we must first recognize that the chief obstacle to solutions of brain problems is this

6 ASC CYBERNETICS FORUM Valurne VII, Number 1-Spring 1975

lack of knowledge. Certainly, any contribution that Automata Theory can make to increase knowledge of brain operation would be welcomed as a significant contribution toward improving the quality of life.

Automata Theory

The primary motivations for founding a theory of automata three decades ago were to understand the operation of the human brain (and the whole Central Nervous System, for that matter), to describe various functions of brain operation by logical formulations (the formalism in Automata Theory), and to use these formulations to engineer better information systems (such as computers). Early mentors in this field felt that the human brain contained the secrets of highly efficient information processing. Understanding these secrets would Iead to highly intelligent machines which could greatly improve the quality of life. Although both suppositions were (and are) true, the work required to disclose just a small fraction of the brain's operation was grossly underestimated. Early disappointments led to a decline in brain theory activity by automata theorists.

But activity did not die out. lnstead, through persistence and the passage of time, a few things were learned about the brain and were used in the design of information systems. What also evolved during this time, and which probably was a saving grace, was the ability of automata theorists to weave brain research data into a pattern of brain theory. As more data were revealed, automata theorists used their formalisms and system engineering approaches to improve their theories of brain operation.

Much of this theory gained credibility and acceptance in the medical field. Today, much of the theory of brain operation developed by automata theorists is found in the most authoritative medical text and reference books, and is used directly in medical specialties involving the treatment of brain diseases and disorders, in brain surgery and in brain research. This theory has been accepted so completely that even though it originated in Automata Theory, it often can not be distinguished in context from other medical information.

Automata theorists have much to offer in the area of brain theory . Through their training in systems engineering and formalisms, they can provide a different perspective, that of a cohesive systems approach, on a theory of a very complex engineering system. In this case the system happens to be the brain. Thus they are able to generate new theories of brain operation from the wealth of new brain research data being supplied by biological and medical researchers.

Shortco mi ngs

Useful discoveries are few compared to the amount of effort put forth by automata theorists working on

brain theory. The reasons are that the scope of many small group efforts is narrow and there is little Co

ordination, either among groups, or between the groups and the medical field. For the most part, brain theorizing has been flowing in one direction: from the researchers who collect data to the theorists who organize it. Other related pro blems make discovery difficult. The overwhelming tide of data and proposed theories coming from biological and medical researchers far surpass the integrative capabilities of automata theorists. Also, difficulty still exists in devising methods to observe neurological operation in the human brain in vivo. As a result, we are too often misled to hypothesize that:

1. what is observed in vitro is also true in vivo; or 2. what is observed in animals is also true in hu-

mans; or 3. functional operation can be derived solely from

structural (anatomical) data.

Goals

One way to increase the effectiveness of brain theorists as a group is to organize goals which contribute directly to a better understanding of brain operations as they relate to the more serious clinical problems. Such an organization of goals would have the familiar pyramidal shape with the top goal being a theory of brain operation. The next Ievel down would be the long-term goals of the theories of operation for the major brain components such as the cerebral cortex, two cerebral hemispheres, thalamus and hypothalamus, midbrain, cerebellum, pons Varolii, and the medulla oblongata. It is recognized that there is such a tremendous amount of interrelationship between even the subcortical areas that understanding these interfaces alone could represent long-term goals. At the next Ievel are several short-term goals which contribute directly to each long-term goal. Many of the short-term goals will be clear, concise and workable with modest resources.

Each short-term goal would require that an associated record be kept, containing a bibliography of research reports pertinent to that goal, a Iist of peop!e who have been working on that goal and a collection of the most recent theories based on the research. Each such record should be inexpensive and readily accessible, perhaps through some widely used network such as ARPA.

One key to the success of this operational plan is to keep it continuously updated and readily available. To insure the proper calibration, interpretation, and use of data and theories, the users (neurosurgeons, brain researchers and experimenters, therapists) and the developers (automata theorists) must maintain lines of communication and attend conferences together. Also, contribution to-or involvement in-these goals should be on a voluntary basis. That is, people should not be assigned tasks in a formal se!lse. This need not pre-

ASC CYBERNETICS FORUM Volume VII , Number I-Spring 1975 7

clude an agency or group from focusing on any efforts which might appear to be particularly promising.

An Example

For example, one long-term goal is a more detailed theory of the operation of the cerebral cortex. Some of the short-term goals in support of this would be regional analysis and stratification; neuron (pyramidal, stellate, and fusiform) topology and orientation (anatomy), chemistry, dimensions, electrical properties, populations, and interconnection patterns. As these pieces are collected and studied, new theories will be proposed, leading to more emphasis on research of in vivo operation in the human brain in an effort to prove these new theories.

A better understanding of the operation of the cerebral cortex can Iead to a better understanding of strokes. For instance, it may be possible to localize the cause, determine the nature of the problern (open artery, tumor, clot, etc.), and specify more specific corrective action and therapy. There are only some 60,000 !arge pyramidal neurons (Betz Cells) in Layer V of the cerebral cortex and they directly control, via the pyramidal tract, many body muscles. Knowledge of the location and extent of darnage to these cells would be useful in determining a patient's chances for recovery. That is, if one area of cells die due to a clot in the arteriole feeding them, the muscles they control will never work again. On the other hand, if destruction occurred to some of the processes feeding the cells, some degree of muscle action may return.

That some of this work has already been done is not an issue, what is at issue is that results are scattered. Collection and organization of these results, permitting cross fertilization and analysis, would be beneficial to all. Duplication of effort would be reduced through the organization of goals. One could very quickly survey the field on one's own, based on the list of references, challenge established theories and propose new ones, and point toward areas in need of further research.

Expansion of Brain Theory Field

Automata Theory as a whole, and brain theory as one of its shining examples, does not seem to exhibit a

deep interest in moving people from the learning or student phase into the independent study phase. A student can elect to take a couple of courses in Automata Theory at a university, but after this point, when his interest and motivation are high, there is no followon next step to take. [See article by Arbib p.9, Ed.] As a result, lacking further opportunity or guidance, motivation dies and interest subsides, to be replaced by other interests. Advanced courses leading toward independent study in brain theory are not widely available. Industry does not support any aspects of Automata Theory. Libraries of books and papers on the subject of the engineering approach to brain theory are less than thin, and brain research positions are offered only to people with doctorates in specialized areas.

There is a huge gap between that level a student attains after a couple of courses and qualification to conduct independent study (and write nontrivial papers and give lucid, factual lectures). In order to bridge this gap, courses on brain theory from an engineering viewpoint are needed. Once some of the vocabulary, techniques and anatomy are picked up, a graduate student could move into books and technical journals on his own. Many of the goals discussed earlier are not too sophisticated for graduate students to work on in their effort to upgrade their base of knowledge. As a student becomes more familiar with the material, he can contribute to theoretical development. Thus there could be sufficient incentive for people at the learning Ievel of Automata Theory to get involved in and contribute to brain theory.

Conclusion

Automata theorists have much to offer the medical field in the evolutionary process of developing a theory of brain operation. It is suggested that the efforts of the automata theorists in this area be coordinated in order for them, and the medical field, to derive maximum benefit from their efforts. To orient efforts by automata theorists in the right direction, as well as passing results of these efforts on to the medical field, communication between the two groups must be improved. In order for engineered brain theory to grow, we must also provide the means and encouragement to Automata Theory students to bridge the gap to independent study.

8 ASC CYBERNETICS FORUM Volume VII, Nurober !-Spring 1975

Cybernetics at the University of Massachusetts at Amherst

Michael A. Arbib Computer and Information Science

University of Massachusetts Amherst, MA 01002

What is Cybernetics?

Norbert Wiener defined Cybernetics as "the study of communication and control in the animal and the machine." Since then, the nurober of definitions of cybernetics has grown to a bewildering profusion. One of the most limited definitions is "the study of feedback mechanisms." One of the most all-embracing definitions is "any study of complex systems, preferably with the use of the computer." It is the latter definition which has found favor in the Soviet Union, and which probably corresponds most closely to the activities of the American Society for Cybernetics.

At the University of Massachusetts at Amherst, we use the nam-e to describe a field far broader than the study of feedback mechanisms, and yet more restricted than the general study of complex systems. For us, the study of cybernetics is primarily the study of attempts to build artificial intelligence (AI) and of information processing in the brain. Our concern also extends to mathematical models of control and communication and, at the other end of the spectrum, to the extrapolation of our principles to discuss societal problems. Nonetheless, the twofold focus is essential in allowing the program to have intellectual cohesion and rigor.

COINS of the Realm

The Department of Computer and Information Science (COINS) of the University of Massachusetts at Amherst has 13 faculty members, of whom five owe their primary allegiance to cybernetics. Three members of the cybernetics Group-Michael Arbib, William Kilmer, and Nico Spinelli-emphasize brain research; while two members - Ed Riseman and Dan Fishmanemphasize studies of artificial intelligence. However, their studies overlap greatly, both in teaching and research. The work in artificial intelligence is aided by Allen Hanson of nearby Hampshire College, and Roger Ehrich of the University's Department of Electrical and Computer Engineering. The brain research is not only aided by a nurober of colleagues in zoology and psychology conducting neurophysiological and neurobehavioral

research, but also by the newly created Center for Systems Neuroscience. This Center, created in January of 1974, as an adjunct to the COINS Department, is directed by an Executive Committee-Michael Arbib (Director), William Kilmer and Nico Spinelll- and is staffed by Fred Lenherr as Center Associate, and Art Karshmer as Computer Director . The mission of the Center is twofold: with its GT-44 graphics system, which will have a tie-line to the university computer, to develop new techniques for brain simulation; and to welcome four Fellows a year-experienced experimentalists in the main part-to work with us on closing the gap between theory and experiment in the neurosciences.

Graduate Training in Computer and Information Science

Of the 70 graduate students currently in COINS, a dozen are now working towards a Ph.D. in the cybernetics area, with a roughly equal division of emphasis between brain theory and artificial intelligence. All students who enter the graduate program in COINS are required to take three core courses: "Fundamentals of Computation," which serves to give them a basic knowledge of algebraic techniques and mathematical concepts of automata theory and the theory of computations; "Fundamentals of Computing Systems," which builds upon an undergraduate background in programming and computer architecture to provide core concepts of computer software and hardware, with special emphasis on operating systems; and "Fundamentals of Cybernetics," which synthesizes the basic concepts of artificial intelligence and brain theory, using as text Arbib's The Metaphorical Brain (Wiley-Interscience, 1972).

Having completed these core requirements, a student proceeding to the M.S. must take five other graduate courses-several of them being quite advanced-and complete a six-credit project in one semester. This project can range from an exercise in complex programming to a preliminary literatme survey prior to Ph.D. research, to a mini-thesis.

ASC CYBERNETICS FORUM Volume VII, Nurober 1-Spring 1975 9

Students who have completed the M.S. program may either leave to enter a computer-related profession, or may continue working towards the Ph.D. To obtain the Ph.D., the student must pass a qualifying exam, complete six courses beyond the M.S., and must write a Ph.D. thesis which constitutes a substantial contribution to computer and information science. The thrust of concentration-after the overview provided entering graduate students by the three core coursesmay be in any one of three areas: computing systems, theory of computation, or cybernetics itself.

Courses in Cybernetics

Some idea of the diversity of courses available in the department may be seen from Fig. 1, which provides the course structure in cybernetics for 1974-5. In addition to these ten cybernetics courses, the department offers a further 26 courses-at all levels from freshman to Ph.D. research-in computer systems and theory of computation. "Computers and Society" is a freshman course designed to help students confront the question, "What are some of the major problems facing the world today?" They then move on to an assessment of the proper role of technology - with a special emphasis on cybernetics and computers-in such issues as global resource allocation, local and regional planning, education, privacy and government. This course is designed for the general student body, rather than for the COINS major, although it is hoped that this course will attract students to computer science who might not otherwise have realized its interest.

"Statistical Foundations of Computer Simulation" is a service course designed to help students understand how techniques from differential equations and statistics can be used in simulating real world systems on computers.

The remairring eight courses of Fig. 1 provide a typical graduate program of cybernetics courses. However, several of the advanced seminars vary from year to year, and many students will take only six or seven

of the ten courses during their graduate career. After the overview of brain theory (BT) and artificial intelligence given in "Fundamentals of Cybernetics," the courses divide into two sets: those which emphasize brain theory, and those which emphasize artificial intelligence. However, as will be seen from the more detailed description of these courses, there is considerable overlap.

"Computational Cybernetics" is a course rotated between Arbib, Kilmer and Spinelli, and varies drastically depending upon the professor! This year it covers techniques for representing knowledge both within brains and in computers; a detailed analysis of the neural control of movement by the cerebellum and spinal cord; parallel preprocessing and scene analysis of outdoor scenes; an introduction to algebraic methods for representing computation and control; cybernetic perspectives on computational linguistics; some ideas on the increasing role of artificial intelligence in question-answering systems; an attempt at a theory of distributed computation with applications to both artificial intelligence and neurology; and a look at theoretical embryology with the perspective of distributed computation.

"Neurocybernetics of Cerebral Cortex" surveys many of the leading models of cerebral cortex and many different experimental studies of cerebral cortex, and seeks to discern the common threads running through them, as well as to point up topics for future research.

In "Natural vs. Artificial Intelligence" both brain theory and artificial intelligence are brought into perspective with the study of questions of learning in both AI systems and neural nets; the development of mental abilities; and strategies of animal problem-solving, with their possible implications for problem-solving by computers.

The courses in artificial intelligence start with the survey course "Artificial Intelligence" which builds upon the basic material presented in "Fundamentals of Cybernetics" to give the student an idea of the current

Fundamentals of Cybernetics

/ ' Computational Cybernetics Neurocybernetics of Cerebra! Cortex Natural vs. Artificial lntelligence

Artificial lntelligence Seminar on Theorem Proving lmplementing a Question-Answering System Seminar on Scene Analysis

Computers and Society

Statistical Foundations of Computer Simulation

Fig. 1. Cybernetics Courses in COINS University of Massachusetts at Amherst 1974-75

10 ASC CYBERNETICS FORUM Volume VII, Number 1-Spring 1975

state of knowledge in such areas as robotics, heuristic programming, Al languages, natural language processing, and scene analysis. The remairring courses, which vary from year to year, then explore individual topics in further detail, bringing the student up to the level of Ph.D. research in the department.

"Seminar on Theorem Proving" not only surveys such techniques as resolution for proving theorems by machine, but also looks at the general role of semantics in deductive systems, explores some cognitive psychology studies which offer comparisons with machine performance, and compares the back-tracking techniques embedded in Al language with some of the techniques available in deductive systems.

The seminar on "lmplementing a Question-Answering System" is then a sequel, in which students can use this knowledge to build a system with the best features surveyed in the previous semester.

Finally, "Seminar on Scene Analysis" presents methods for parallel preprocessing to extract crucial features from a visual scene (as well as looking at some of the techniques whereby vision routines can make sense of these features) to reconstitute a description of the objects represented in the scene.

Research in Cybernetics

With this overview of the structure of the cybernetics program, and with this sampling of the available courses, we can turn to a description of the research program, expanding upon the summary given in Fig. 2.

Our general concern with the structures of control, communication, and computation in a system communicating with a complex environment embraces both the analysis of how these were achieved within the brain (BT) and the synthesis of mechanisms for achicving these within the computer (AI). Our approach is very much in the spirit of Warren McCulloch, with whom both Arbib and Kilmer worked at MIT in the early '60s.

In interacting with a complex world, a robot or an organism must have receptors whereby it can extract important information from its environment, it must have effectors whereby it can interact with the environment, and it must have internal structures which enable it to combine current information with the record of prior experience to determine current action. In seeking to analyze the way in which a system can extract information about its world, we have emphasized the analysis of visual systems of animals and robots.

Our analysis of internal structures has focussed on the construction of "internal models" or "semantic data bases" on the one hand, and on the study. of adaptation in neural nets on the other-with the latter using both theoretical and experimental techniques. Finally, our study of effectors has focussed on the role of the cerebellum in modulating the control of move-

ment by the spinal cord. With this overview, we now turn to a more detailed account of our work in vision, learning, computation and movement .

The brain group's work on vision has many connections with the work of the Al group. lt has two main sources. One is the theoretical perspective created by Arbib with his "slide-box" metaphor, in which visual input is used to cue both the retrieval of appropriate "slides" from long-term memory, and their relative positioning, to provide an adequate "description" of the input which gives the system access to programs which can be compiled in a context-dependent manner to control action. The other main source of our work in vision is the series of neurophysiological experiments conducted by Spinelli on the developing visual system of cats and other animals.

In a recent study of eye movements and visual perception, Didday (now at the University of California at Santa Cruz) and Arbib built on Didday's earlier model of the frog tectum-the region of frog midbrain which controls the snapping of the frog at flies-to suggest how a slide-box visual cortex could use midbrain systems to direct eye movements to gain the necessary information for building up an internal model. This study was of particular interest both for its demonstration of parallelism in these mechanisms, and for its use of a hybrid Al/BT model, with cortex being given a gross functional model, while the midbrain was given a detailed neurophysiological model.

A study by Parvati Dev (now at the Neuroseiences Research Program of MIT) showed how to design neural networks which could segment a visual scene into different regions on the basis of coherence of particular features. One particularly interesting application of this work was to stereopsis, and Peter Burt (a Ph.D. student) is now extending the model to encompass moving scenes. This model is both shedding new light on neural mechanisms of motion perception, and also is setting up a number of ideas which may eventually augment the VISIONS system (described below) which at the moment only operates upon static scenes.

Spinelli's demonstration that restricted visual experience in the developing kitten yields brain cells which are "imprinted" with the experienced patterns has many interesting implications. Lenherr has carried out further experiments on the effects of binocularity upon such development, as well as making related studies of the auditory system. He has developed an adaptive neural net model of von der Malsburg of Göttingen to account for the experimental data he has gathered. Fanya Montalvo has also built upon the von der Malsburg model, this time to explain some of the psychophysical data upon visual adaptation in humans, as well as other aspects of human visual perception.

Riseman and Hanson are developing their VISIONS system for the analysis of outdoor scenes. Their aim is to extract gross descriptions-such as the type and location of major objects, or gross terrain features-


BRAIN RESEARCH GROUP ART I FICI AL INTELLIGENCE GROUP

Faculty: Arbib, Kilmer, Spinelli CSN Staff: Karshmer, Lenherr

[Didday on tectum) [Arbib's 'Siide-Box')

Faculty: Fishman, Riseman Associated Faculty: Ehrich, Hanson

1 ---------+ Riseman, Fishman and Hanson on Semantic Networks for Scene Analysis

Didday and Arbib / f on Eye Movements and / I Visual Perception / Riseman and Hanson

1 / Burt on Motion / Devon*

Segmentation ----+ and Stereopsis

Montalvo

on visual ----adaptation

Spinelli and Lenherr on OCCAM, [von der Malsburg) and Visual Development

t I ~

Kilmer [Mclardy and Stanley on Ranck) on Learning Habituation in Hippocampus

/ Killer [Kilmer & / on Animal Learning McCulloch r on RETIC) Kilmer on Ethology

l Arbib and Montalvo on competition and consensus

Arbib on Distributed -----+ Cooperative Computation

r Arbib and Manes on Machines in a Category

Arbib, Karshmer and Lenherr on CORETEX

[Arbib, Franklin & [Greene on Nilsson on Cerebellum) Strategies of

\ /.Movementl

Boylls on Cerebellum*

! Maulucci & Ariel on Mechanisms of Locomotion

on Parallel Preprocessing

Riseman, Ehrfeh and Fisher on Pattern Recognition using Context

1 [ Riseman on Feature Selectionl

Soloway on Automatie lnduction

Fishman on Rote-Learning in Goal - Directed Systems

I Fishman on Deductive Semantics

Fishman, Riseman, Soloway and Arbib on Al Languages

[ ) indicates background werk done elsewhere; * indicates completed Ph.D. research project .

Figure 2. Cybernetics Research Coins Department UNIVERSITY OF MASSACHUSETTS AT AMHERST

12 ASC CYBERNETICS FORUM Vo1ume VII, Number 1-Spring 1975

-

with details to be filled in depending on system application. Their project consists of two parts. The first involves the development of "preprocessing cones"these are layered parallel computational structures (though currently simulated in the serial computations of our PDP-15) which extract low-level visual features such as edges, regions, color, texture and shape. The next stage of the research will involve the use of these visual features by vision routines, one for each possible object such as "sky ," "tree," or "road" in the scene being analyzed, to get confidence levels for presence of objects in various regions. The vision routines will process cone output using a sem;ntic data base of general knowledge; such context settings as "winter," "outdoors," "rural"; and information from the partial model constructed so far; as well as context frames which provide information about expected models or stereotypes.

The building of the model of the scene under analysis will be controlled by an executive, which will invoke subprocesses, switch messages between subprocesses, integrate responses, examine implications, and resolve conflicts. Of particular interest will be the way in which the executive handles the semantic data base, and here a collaborative effort is under way to use semantic networks for scene analysis via techniques developed by Fishman in his work on deductive semantics.

Briefly, Fishman has shown how semantic cues can be used in developing extremely efficient search strategies for answering questions about a relatively large qata base-some 400 assertions concerning kinship relations. The next step will be to adapt such techniques to handle a 1 000-assertion da ta base concerning the properties of, and the relationships between, typical objects in outdoor scenes. In addition to developing these techniques with Riseman and Hanson, Fishman will be continuing to look for general methods for speeding up deduction, and will be building a powerful new system for deductive semantics.

Riseman is also working with Ed Fisher (a Ph.D. student) on pattern recognition techniques which make use of context. Very high levels of error detection and correction of words is achieved by a contextual postprocessing system which follows a standard pattern classifier. The current application of these techniques is to recognizing addresses and zip codes despite the type of degradation typical in the usual material handled by the Post Office.

An important historical source of our work in brain theory is the collaboration in the early '60s of Kilmer with the late Warren McCulloch of MIT on a model of the reticular formation-the command and control system in the brainstem of mammals. In current work, descended from this study of the reticular formation, Kilmer has been modeHing the hippocampus as a learning network. A spin-off of this work is the study by Jim Stanley (a Ph.D. student) of habituation and wave

propagation in hippocampus-like neural networks. In another development, Kilmer is working with such neurophysiologists as Jim Ranck of the University of Michigan on trying to make sense of their data on the correlates of activity of single cells in the hippocampus during different types of animal behavior. This leads into a general study of ethology-animal behavior-with emphasis on setting up AI learning schemes which can account for a number of Observations on animallearning.

Learning has received surprisingly little emphasis in recent work in artificial intelligence. In addition to Kilmer's work, a project by Elliot Soloway (a Ph.D. student) seeks to remedy this by exploring ways in which the rules underlying some complex performance may be automatically inferred. As his model world, Soloway is using baseball, trying to devise general strategies which would enable a system to pass from the description at the low level of running, swinging the bat, and throwing the ball, to a higher-level description in terms of runs and innings, etc.

Turning from the specific studies of vision and learning to more general questions, we may note that Arbib and Montalvo have been looking for a general theory of competition and consensus in neural networks which generalizes Didday's study of the tectum, Kilmer and McCulloch's study of the reticular formation, and Dev's study of segmentation. Arbib is also looking at some general models of distributive cooperative computation, in the hope that they will both provide a more natural way for looking at the structure of certain AI problems, and also shed light on brain lesion studies.

The Center for Systems Neuroscience group is currently developing a new language, CORETEX-this is not an acronym, but combines the words "core" and "cortex" to suggest the use of computers to aid in modeHing the brain-which will have great inherent parallelism, and make it very easy to set up simulations of neural nets on the GT-44 system, and the larger backup system that it will utilize for number crunching. The development of CORETEX goes hand in hand with the concern of the whole cybernetics group to analyze the key features of various Al languages currently available, to see whether it will be possible to reconfigure them into a new language in which the expression of parallelism becomes far more natural, without sacrificing any of the powerful backup features currently available.

The one area of "classic" cybernetics, in the sense of Wiener, in which the department is deficient is the general area of information and control theory. Courses in these areas are offered in the Department of Electrical and Computer Engineering, but little of this work has any effect upon the programs of our cybernetics students, unfortunately. There is also some related work in our Theory of Computation group, but this-again-has little impact upon the cybernetics program. Arbib and


his colleague Ernest Manes-of the University's Mathematics Department-are applying the abstract algebraic concepts of the field known as category theory to build a general theory of computation and control which, for example, unifies the study of observability , reachability, and minimal realization for sequential machines, linear systems and tree automata. Other studies in this area look at the general concept of nondeterminism-the theory of systems in a fuzzy world-as well as such topics of programming theory as the manipulation of syntactic and semantic trees.

The final research area of our brain research group concerns the neural control of movement. This study grew out of a preliminary view of the role of the cerebellum in the control of movement conducted by Arbib, Gene Franktin of Stanford University and Nils Nilsson of the Stanford Research Institute in 1968, and work dorre at the University of Chicago by Peter Greene on the strategy of movement. Curt Boylls (now at the Neurophysiology group of the Good Samaritan Hospital in Portland, Oregon) developed a model in which the cerebellum provided the off-line setting of parameters to be used by algorithms resident in the spinal cord which control locomotion. In building upon this model, Arbib, Ruth Maulucci (a Ph.D. student) and Gideon Arie! (a Research Associate) are taking cinematographic records of cats walking on treadmills under varying conditions to provide better insight into that which the spinal algorithms are to control.

Suggestions for Further Study

The reader wishing to learn more about graduate study in the department may write to Dr. E.R. Riseman, Graduate Admissions Director, COINS/GRC, University of Massachusetts, Amherst, MA 01002, U.S.A. For a more detailed perspective on much of the research outlined above, the reader should consult: J. Szentagothai and M.A. Arbib: Conceptual Models of Neural Organization , published as Valurne 12, Number 3 (October 1974) of the Neuroseiences Research Program Bulletin - for more on the work of Arbib, Boylls, Dev and Didday. M.A. Arbib, W.L. Kilmer and D.N. Spinelli: "Neural Models and Memory" -for more on the brain group's studies of learning. D.H. Fishman, A. Hanson and E. Riseman: "A Model-Building System for Scene Analysis" - for more on deductive systems and the VISIONS system. M.A. Arbib and E.G. Manes: "A Category Theory Approach to Systems in a Fuzzy World"-for more on the unified theory of computation and control.

The last three references are available as COINS technical reports, and may be obtained by writing to any of the authors. Each paper contains a bibliography with many references to other papers published by the COINS Cybernetics group of the University of Massachusetts at Amherst. The Center for Systems Neuroscience will shortly begirr publication of the Brain Theory News/etter . Enquiries should be addressed to Fred Lenherr, the Executive Editor.

14 ASC CYBERNETICS FORUM Valurne VII, Number I-Spring 1975

I

I t

I I

I

Reverberations Mind: The Software of the Brain

(Variations on N.A. Coulter's theme)

Roland Fischer Maryland Psychiatrie Research Center

Baltimore, Maryland 21228

When a solid object or a perturbation lies in the path of a sonic probe, reverberations are often detectable. Characteristically, the returning signal carries something of the nature of the probe emitted and something of the conformation and composition of the object probed. A printed probe is analogaus to a sonic probe, as witness Dr. Roland Fischer's reverberations to Dr. N.A. Coulter's Probe in the Winter 1974 issue of FORUM.

-Editor

I ntroduction

We should be grateful to Dr. Coulter for having written such a mindful article (of faith) that truly matters. I have thoroughly enjoyed reading it and offer here in response to his most stimulating and indeed provocative lines of thought a few reverberating reflections.

Reverberation One

For Dr. Coulter, mind is simply an organization (of the instruction code) of the brain. Aristotle made a similar proposition when making a clear distinction between the body and the actual organization of the body: our soul. My point is that the mind (Aristotle's soul) may be regarded as the software of not just the brain but ,of the whole self-organizing system, the living organism. Heart and blood, liver, genitals and even the big toe, are no less important in their contributions to mind-behavior than, for instance, neurons or glia. Non-linear, superadditive, "sophisticated," i. e., goal directed · (mind) behavior cannot be localized within

part(s) of the system since it is inherent in, and a property of, the hierarchically structured complexity of the whole system.

Foxhunting, for instance, to use Woodger's* example, is a novel goal-directed behavior of the equimo (a combination of horse "[equus] and man [homo] and it is quite evident that one cannot localize foxhunting within any particular part of this "sufficiently complex system."

But what is a sufficiently complex system? In cantrast with a "simple" amoeba with a rigid program of responses we conceive of a sufficiently complex system as one wherein the system's interaction with the environment becomes part of the program-as in metazoa. It is, then, in these self-programming systems that behavior manifests itself as "detached," non-localizable, and as a symbolic interpretation of the system's interaction with its environment.

Reverberation Two

Historically speaking, we may observe an increased sophistication in the mind's ability to devise models of its own function, and indeed, the evolutionary trend in brain modeling favors constructs such as the hydraulic engine, the telephone switchboard, the computer and, of late , the hologram (Pribram, 1971; Fischer and Rockey, 1967t), which require less and less energy and more and more "information space."

*Woodger, J.H. In a lecture as quoted by J. Rothstein, Columbus, Ohio (personal communication, 1968).

trrimbram, K. Languages of the Brain, Englewood Cliffs, NJ: Prentice-Hall, 49-50, 206-207 (1971). Fischer, R. and M. Roc key, A Heuristic Model of Creativity, Experientia, 23:150-152 (1967).

ASC CYBERNETICS FORUM Volume VII, Number I - Spring 1975 15

•

The small amount of energy necessary for the brain (and its models) to process information (i.e., to mind its own business) recalls the symbolic nature of the genetic code, through which living matter evades the constraint of energy. Goodwin# (1972) notes in this context that the primary structure of a protein is not determined by a condition of minimum free energy, but by the sequence of bases in its corresponding gene or cistron. The process of translation from messenger RNA to polypeptide chains carries only one-dimensional information in the sequence of amino acids. It is the interaction of this one-dimensional text with its "physiological environment" -a hypothesis testing situation- which determines three-dimensional texture. Note, however, that the function of protein has no explicit relation to its three-dimensional structure due to the hierarchical mode of information processing involved.

Evidently , there are systems which process information-which have rule-like behavior-but with which we cannot communicate in any prescriptive way. Such systems obey laws in the sense that the rules governing their behavior are not explicitly represented in their constraint. This kind of information processing is called by Conrad (1973)** nonalgorithmic and hierarchical. (Note the close similarity with the hierarchical information processing customary within the Catholic Church. Y ou often do not pray directly to God. You may pray to your guardian angel who then may convey the message to a higher ranking saint who, in turn, may "translate" the prayer to the holy Virgin who agairr may mediate between you and Christ).

Such systems allow for gradual change in function associated with gradual change in structure. Gradual-

#Goodwin, B.C. , "Biology and meaning" (in) Towards a Theo· retical Biology, Four Essays, Washington, C.H. (ed.), Chicago : Aldine/Atherton, Inc. (1972).

**Conrad, M. "Is the Brain an Effective Computer?" Interna· tional Journal of Neuroscience, 5:167· 170 (1973).

ness is, then, the property implicit in evolutionary hierarchical systems.

The systems with which we can communicate according to a definite rule have quite different evolutionary behavior than the law-governed system which otherwise simulate them. This is the result of the tradeoff between the programmability of the system and the adiabatic modifiability of its function-concludes Conrad- and I have to quote here my good friend Harry Blum (National Institutes of Health, Bethesda), who recently called my attention to an intriguing phenomenon: the impossibility of differentiating between a hyperbola and an ellipse if both are placed on top of each other- not on a plane, to which they plainly belong, but-on a sphere. They will completely cover up one another to truly illustrate that one cannat imply process from shape (i.e., function from structure) even though a process properly defines that shape.

Do we not illustrate-by looking with our brain at brain function-another cover-up?

Reverberation Three

Dr. Coulter states that the mind is the software of the brain and makes up this statement with his mind which is the software of his brain. Apparently, the rational study of brain behavior-i.e., mind-is hampered by the scientific discovery (Toulmin, 1971 tt) "that strictly causal brain-mechanisms underlie all rational thought-processes-including the scientific discovery that strictly causal brain mechanisms underlie all rational thought processes."

ttToulmin, S. and R.S. Peters, "A debate." (in) Borger, R. and F. Cioffi (eds.), Explanataion in the Behavioral Sciences, Cambridge : Cambridge University Press ( 1971 ).

16 ASC CYBERNETICS FORUM Volume VII, Number I-Spring 1975

Book Review Christopher Longyear Department of English

University of Washington Seattle, WA 98195

Management of Information Handling Systems, Paul W. Howerton (ed.), Rochelle Park, NJ : Spartan Books (Hayden Book Co., Inc.), 1974. 224 pp., $12.95.

Cognition: A Multiple View, Paul L. Garvin (ed.), New York & Washington: Spartan Books, 1970. xv + 428 pp., $22.95.

The Howerton book gives the uninitiated reader a look at the seamy world of information specialists in various unprincipled domains ( e.g., the library, the specialized technical information cen ter, the school, law, government and business). Lack of principle is an accusation that is sometimes leveled by the authors at agencies, at procedures practiced by groups or individuals, or at disciplines as currently constituted. Such "unprincipled" elements are usually more or less overt, as when we are reminded that individuals may exploit information systems in business for base personal motives of advancement (e.g., in "The Social Nature of Information Management" part of Chapter 9). Sometimes, however, the lack of princip1e seems to be innate to the subject matter, as when we are to1d in the first chapter: "The lack of a body of proven axioms stemming from generally accepted theory is a ... cause for disagreement among documentalists." Sometimes the book itself seems a shade unprincipled, for of the 11 contributing authors, every one of them is from the Washington, DC area, and (with the exception of a co-author and one other) all are affiliated with the American University. Such b1atant inbreeding is bound to give us a twinge of apprehension.

We can discount the book's claim to be a reference book (a term used in the dust jacket blurb), but we should certainly consider its claims to be, in the editor's words, "a textbook, or [a] source of additional readings through which [ the reader's] knowledge of information systems is improved," (the editor's own characterization on p. 204 ). We shall therefore try to evaluate this collection on its own terms.

A few misprints, including a gross shifting about of an entire line from page 38 to 39, do little to enhance the rhetorical effectiveness of the book. Some of the language is distressingly reminiscent of educationalese or governmentalese. We would more willingly overlook such failings if the fundamental organization of the book, the individual chapters, or the individual ideas were more sinewy.

As a textbook, it is presumably aimed at managers or would-be managers, apparently at a very preliminary

level. As such, one might indeed expect some superficiality in surveys, justification for which must be sought on pedagogical or presentational grounds. Still, the first chapter, which bravely admits the lack of theoretical underpinnings quoted above, makes at least this reviewer uneasy when it goes on to say a little later in the same paragraph: "In one sense, the organization giving thought to system planning is relieved of a considerable burden by the very absence of formal theory." Such an introduction to the management of information handling systems may honestly reflect the current dismal state of the art, but it surely bodes ill for its future development.

One is a bit puzzled by the world-weary pessimism reflecting realistic appraisals of the actualities on the one hand, sharply contrasting with quite visionary hopes for a brighter future that will come to pass when undreamed-of difficulties have disappeared, on the other. One is also distressed by some assumptions that seem to underlie many of the chapters. Chapter One, for example, seems aimed exclusively at corporate research institutions. Such an emphasis reduces science to sheer development. What David C. Weeks calls "documentalistics" seems to me to demand a re-examination of what information is. What it is not is a collection of "knowledge" stored like goods in a warehause, requiring only the application of efficient inventory labels to be retrieved economically.

We should note at the outset that there are some comparatively praiseworthy elements. Milos G. Pacak's "Computational Linguistics and Information Handling" does provide a modest survey of some of the work done in this field, even limited as it is to work reported in 1971 or earlier. It suffers, as does the rest of the book, from failure to attend to the fundamental question of what is information; instead, it lists a number of particular systems, none of which could be adequately presented in enough detail to let the reader get either a feeling for the operation of the system in practice or a sense of the underlying model of information that theoretically is implemented. Another praiseworthy element is the book's occasionally listing sources that may be useful for the intended reader. Thus an appendix listing some 75 "U.S. Government Technical Information Systems Relevant to Environmental Quality" closes Martha C. Sager and Charles M. Cargille's chapter called "Information Handling in Environmental Sciences."

But the shortcomings of the book far outweigh such peripheral advantages. There is, to be sure, an attempt


made to enlist management science to the aid of managers; again, this is only superficially treated and seems, with few exceptions, overwhelmingly directed either to management in general, or to cook-book check-lists of particulars that could be appropriate for a single type of system.

I am personally appalled at the notion of a textbook that minimizes the need for a solid theoretical foundation. Surely students, especially those who hope to be managers of information-handling systems, should be discouraged from adopting the technician's approach to tinkering with systems on purely an adhoc basis; they should be encouraged to demand an intellectually satisfying and a scientifically believable general theoretical framework. lts absence, far from absolving the student from the need to find one, should instead force him to put its development at the top of his list of priorities rather than at the bottom. Applied disciplines, whether in education, in technology or in any other realm, can only justify their existence on the basis of something to apply; justification on the basis of mere need smacks of charlatanism.

The consequences of lacking a theory are evident. What some authors seem to see as "information" is indistinguishable in nature from any other bulk commodity; an extreme example is perhaps Sager and Cargille's visionary "mental model of a fully adequate [environmental science] system" which "must include the totality of human technical knowledge." One wonders at the modest restriction to technical knowledge, for one man's art may become another's science. If we are to wait on such systems, especially if "heuristic computing" is to do all sorts of things like develop "some capacity of correlating data to synthesize new information ... the capacity to discriminate the better data from the inferior, the capacity to teach itself from printed materials and to compose grammatically correct manuscripts [sie] synthesized from information selected from the vast store of a memory bank," we had better not hold our breaths while waiting for environmental science to clear the air.

Readers of this joumal will hardly be surprised to find a few comments urging system design research; they may be surprised at how few they are. Some cybernetic principles are acknowledged, as when the editor's introduction to Terence G. Jackson, Jr.'s "The Information System Executive" points out: "We are now convinced that the person who controls the flow of information through an organization controls the organization," though the term "cybernetic(s)" does not appear in the book's index.

The book is not utterly devoid of redeeming social virtue, however. There may be a real service performed for novice managers when they are warned not to believe the blandishments of computer salesmen. For the tyro, it is probably also useful to hear of distinctions drawn between data systems and information systems, or among the activities of document-retrieval, indexing and abstracting. The book may perform yet an even greater service to the field of information systems: it may goad some of us into action. Instead of dismissing the book as irrelevant or merely selfcontradictory, let me recommend it as an example of what happens when technicians are unconstrained by

theoretical foundations. If it serves as a prod to some of us who are interested in the nature of human information, the book is worthwhile.

To launch what I hope may develop into a fruitful discussion of human information from a cybernetic point of view, I suggest that there are principles on which we may build. One basis might be called the biological basis of human information (cf. Maturana's article, discussed below). Another we might label the logical basis of human information (cf. von Foerster's and Meredith's papers, discussed below). Still another rnight be called the representational basis. * Let me draw one lesson applied to information systems here; the small informational community that is incorporated formally into some computer question-answering systems cannot long survive if it lirnits itself to preconceived categories. "Change or die" is an inexorable law of all viable systems, and a data base used by humans about real information, unlike a few privileged human beings, cannot be pardoned from the consequences of unforeseen changes.

Reports of conferences tend to be of very uneven quality. On the one hand, an editor has the responsibility of reproducing accurately the stuff of his conference- good, bad or indifferent-and on the other, he has the obligation to his readers to select and arrange material so that the book becomes more than a mere hard-bound version of the preprints. For example, in the 1378 pages of the three volumes reporting the 1969 International Congress of Cybernetics at London, there may lurk a few contributions of less than outstanding value. t Sometimes, however, the conference itself seems to be selective enough to produce a good crop of papers; when these are polished after the meetings to reflect the positive gains of the interaction meetings are supposed to produce, then such proceedings can be of extraordinary virtue. In the 428 pages of Cognition: A Multiple View edited by Paul Garvin, there are collected some extremely good papers, most of which should be of considerable interest to cyberneticians.

There are few editorial lapses. An example of a minor babble is a reference to comments of developmental psycholinguistics "as discussed in this symposium by Belugi-Klima" (p. 190); no such discussion appears in the book, though its exclusion is acknowledged some 150 pages later (p. 345). I also hope that my difficulty with the sentence, "Hence, for CLS, the f' concept is a test tree and the 11 concepts are either building strategy or selection strategy, building strategy, depending upon whether the system is used in the receptive or the selective mode" (p. 388) is due to a typographical error rather than to my grammatical inadequacy.

Compared with the Howerton book, Cognition is very different. It is the result (we are told in the "Acknowledgement") of "the proceedings of t!Je symposium on 'Cognitive Studies and Artificial Intelligence Research,' held March 2-8, 1969, at The Univer-

*Christopher Longyear, Nature's Grammar: A Representational Theory of Language. (Working title; to appear).

tJ. Rose, ed., Progress of Cybernetics, 3 vol. London: Gordon and Breach (1970).

18 ASC CYBERNETICS FORUM Volume VII, Number 1- Spring 197 5

sity of Chicago Center for Continuing Education, under the sponsorship of the Wenner-Gren Foundation for Anthropological Research." The meaty substance is presented under four headings we shall discuss briefly below.

In a brief "Introduction" (pp. ix-xiv), Paul Garvin gives us his view of the "basic issues raised," which tend to be either rather general ("The view was presented that in the structure of cognition, relationships are more important than component elements. There were varying degrees of disagreement on this point," or frankly personal ("Another [ question raised during the symposium] was the obvious question of realism in simulation. This latter question, in my opinion, is the fundamental one, and it deserves fuller elaboration"). The greatest virtue of the book resides not so much in its organization, or in the introduction, but in the unusually high level of excellence displayed in the individual papers.

Though cyberneticians should not wonder at the feasibility (or even the utility) of a cybernetic look at biological, psychological and computational disciplines, individual practitioners may boggle a bit at domains more usually associated with anthropology, literary criticism and linguistics. I shall argue that even these~ perhaps esoteric~regions can be stimulating when seen aright. In the brief description of the contents of Cognition which follows, I shall in fact be inviting scholars of cybernetics to consider, however briefly, just such domains.

First, to cover ground that is probably more familiar, and to persuade the sceptical specialist that there may be something worth looking at in the book, let me here present a selected list of authors, whose names should be already familiar from more obviously cybernetic work . Among these we might list Humberto Maturana (yes, the co-author with Lettvin, McCulloch and Pitts of the seminal article, "What the Frog's Eye Tells the Frog's Brain"); Heinz von Foerster, former President of the ASC; Patrick Meredith, author of Instruments of Communication; Saul Amarel, one of the contributors to the first ASC conference; Michael Arbib, author of Brains, Machines and Mathematics; and Gordon Pask, whom Warren McCulloch characterized as the man "generally regarded as the genius of selforganizing systems of the variety called cybernetical." Other contributors, who may be less well known to ASC members, are nevertheless important scholars who have received international recognition in their respective fields. With such a galaxy of luminaries, we expect a dazzling performance. We are not disappointed.

Part I, "Cognition and the Organism," has three chapters, one each by Maturana , von Foerster and Meredith. In a careful presentation of the informational and ecological niches of organisms, Maturana leads us from the simplest organisms responding to its environment to humans with their extraordinary ability to create informational environments of virtually infinite scope.

Heinz von Foerster presents a two-part paper, "Thoughts and Notes on Cognition." The first part amounts to a plea for terminological and especially conceptual sanity: do not confuse the properties of a model or a simulation with the properties of an

organism being simulated. The second part is an attempt to axiomatize the informational world of a simple organism; this part has also appeared as "Toward an Axiomatization of Cognition" in an interim report from the Coordinated Science Laboratories, University of Illinois.

Patrick Meredith's "Developmental Models of Cognition" presents some of the notions developed more fully in his Pergarnon book, Instruments of Communication: An Essay an Scientific Writing, which relentlessly pursues a search for the parameters involved in an act of communicating even the most elementary fact.

The secend part of Garvin's book is called "Cognition, Culture and Language." It seems to concentrate more heavily on anthropological-linguistic matters such as kinship terms and folk-tale narrative structures, while the third part is more clearly linguistically oriented. Part II contains four chapters, by D' Andrade, Wallace, Werner and Colby, names that may be unfamiliar to cyberneticians. Roy G. D'Andrade presents an illuminating discussion of the technique called componential analysis; his and other papers in Part li inevitably touch on linguistic issues, but these are faced more directly in Part III. Anthony F. C. Wallace argues for the greater psychological reality of a logic of relations (for example, "My father's sister's son" defines a relative product in English), compared with the more traditional logic of classes.

Oswald Werner sees the need for a truly global representation of a world of discourse from which explicit and implicit su bsystems such as definitions, taxonomies and other kinds of linkages can be retrieved. In his "Epilog" (p . 173), he mentions lately discovering Raphael's question-answering device, SIR,* which refers to a 1964 work. It is a pity that Werner apparently did not know either of SIR or of other question-answering systems such as DEACON** or REL *** which might satisfy a good many of his requirements for formal representations. Benjamin N. Colby's paper, "The Description of Narrative Structures," takes a rather traditional structuralist-formalist approach to describing the structures of folk-tale narratives; he insists upon keeping track of the chronological or sequential positions as well as of the form and the function of a narrative element. We verge on literary criticism here.

Part III, "Cognition and Meaning," turns out to be four chapters by representatives of various schools of linguistics. Stratificational grammar and generative semantics are represented by major figures, Lamb and McCa wley. Structuralist-formalist linguistics is repre-

*Bertram Raphael, "SIR : A Computer Program for Semantic Information Retrieval," AFIPS Conference Proceedings, vol. 26, part 1; 1964 Fall Joint Computer Conference. Spartan Books (Washington, D.C., 1966).

**James A. Craig, Susan C. Berezner, Homer C. Camey, and Christopher R . Longyear, "DEACON: Direct English Access and Control," AFIPS Conference Proceedings, vol. 29; 1966 Fall Joint Computer Conference. Spartan Books (Washington, D.C., 1966), pp. 365-380.

* **Bozena Henisz Dostert, REL-An Informational Systemfora Dynamic Environment. REL Report No. 3. California Institute of Technology (Pasadena, Dec. 1971).

ASC CYBERNETICS FORUM Volume VII , Number !~Spring 1975 19

sented vigorously, if a trifle churlishly, by Olga Akhmanova, and by Madeleine Mathiot, perhaps more tellingly. Sydney M. Lamb 's chapter, "Linguistic and Cognitive Networks," should be read if only for its elegant graphic notation, which is one of the most attractive features of stratificational grammar. James D. McCawley argues for the generative semantieist position on semantic representation in linguistic models. Mathiot's paper, "The Semantic and Cognitive Domains of Language" sets out to disentangle the named domains. Her point of view coincides fairly weil with Akhmanova's in "Semantic Features," but she has much more to say.

One misses, perhaps, a spokesman from the currently dominant school of linguistics (Noam Chomsky and Jerrold Katz, for example), but the variety present is perhaps adequate to indicate some of the scope of disagreement among linguists.

The last part of Garvin's book will perhaps again seem a bit more familiar to cyberneticians; Part IV is titled "Cognition and Automata." It begins with Saul Amarel's "Problem Solvin~ and Decision Making by Computer: An Overview," in which Amarel discusses the evolution of solution languages in terms of changing their grammar and in terms of finding suitable representations for the problems.

Michael Arbib's paper should endear him to us if only for his "cyberneticians (or members of the artificial intelligentsia?)," an irresistible if tentative suggestion. Arbib sought "to give readers a feel for the different type of questions one may ask about cognition when one starts from the problern of designing a machine which can exhibit properties of cognition, or, even more interestingly, design a machine which can organize itself in interaction with its environment to exhibit properties of cognition" (p. 348).

Chapter 14, Gordon Pask's "Cognitive Systems" argues that cognitive systems are goal-directed systems. The chapter characterizes such systems in terms of TOTE (Test Operate Test Exit) units in hierarchical organizations. His discussion of learning, teaching, and conceptual genesis and evolution in these terms is followed by a considerable display of specific systems that implement various of these notions. As a fitting dessert to the banquet, Pask offers an appendix that is a reference source list for particular systems. His list illustrates at once the variety of the field and also demonstrates the validity of his caveat, "the reader is warned that even today a great deal of the material is contained in report documents that are not always readily available." Cybernetician, inform thyself!

The final chapter, "Man-like Machines: The Language Barrier," by Paul Weston, discusses some of the differences between artificial intelligence and linguistic views of human language. As a sametime computational linguist, I can only agree with Weston that some AI results should begin to affect linguistic theory. The nature of natural language has obviously not been exhaustively defined by modern, technical linguistics. To select out of the rich array that is human language only those elements that may lend themselves readily to a convenient systematization does not necessarily define what language is. I for one find it an essential property of human language that it reacts systemati-

cally to the pragmatic as well as to the semantic and syntactic components of a world of discourse.

With this I hope modest lapse into special pleading, I return to the more general question of why the Garvin book should be so rich and the Howerton book seem so thin. It is certainly not due to differences in subject-matter only, for in this respect, the overlap is considerable. It does not seem to be due to differences in editorial practice, for if anything, Howerton tries to be the more helpful. Where Garvin modestly retires after his brief "Introduction" letting the chapters speak for themselves, Howerton contributes the final chapter, a preface, and a note to each of the chapters (including one to his own), pointing out connections with other chapters and with the book as a whole.

The distinctive difference seems to be in the amount of sound theoretical underpinnings that support the individual contributions. Howerton's book not only admits a lack of theory, it sometimes even seems to suggest that such a lack does not matter and may even be desirable. Where Howerton seems to suggest that the application of some very watered-down broth of management or information theory will cause the management of information handling systems to thrive, Garvin's book is supported by a rich, thick stew of sometimes conflicting but meaty theories that range from several brands of logic, through neurophysiology, to brands of scientific anthropology, linguistics and automata theories.

It seems to this reviewer that the application of technologies to particular purposes is a valid undertaking, and not only because that application serves as a test of the theory. But surely there are circumstances in which an application is not appropriate. Let me for a moment caricature an application of a scientific theory so that we can see where we may look for clues suggestive of propriety .

A refrigeration engineer might apply some piece of the science of thermodynamics to home appliances to produce a refrigerator. illtimately, it makes sense to train technicians to repair and service the refrigerators. lndeed, it is quite likely that the technician who services such appliances will do a better and eheaper job of repair than will a physicist who is conversant with statistical mechanics. The technician, who has been trained as a technician (attach gauge "A" to tap "B"), needs to know little of the theories underlying the appliance. But such circumstances presuppose a definite context in which, for example, refrigerators all have pretty much the same design. If refrigerators were to be made not with compressors and condensers, but with thermoelectric components, our technician will need retraining. (What happened to tap "B"?) His new training, of course, will also be limited by the requirement that design changes do not occur. The quick-anddirty job of training technicians can often be accomplished without damaging the trainee's innocence of theory. But if refrigeration design has never become standardized, and is in fact in a period of rapid design changes, then the whole enterprise of technician training is extremely suspect.

Perhaps if Howerton's book had been less persuasive on the subject of new technologies of the near future, we might be less dubious about the virtue of what

20 ASC CYBERNETICS FORUM Volume VII, Number I-Spring 1975

amounts to a technician's approach. Surely the education of managers of information handling systems requires more than the least flexible approach. We can look forward to dark days in the library if the Howerton book is truly representative of what is happening in informa tion processing generally; fortunately, it is not, as a glance at the American Society for Information Science's Annual Review of Information Science and Technology series assures us.

Another example may drive the point harne: I hope we feel some dismay when we see the education industry gear up to adapt some one learning theory to the classroom. To caricature once more, suppose that we take as an example something like the schedules of reinforcement developed from animal behavioral studies. It seems to me that their validity as the major classroom technique should be criticized on at least the following three grounds: l) The basic psychological theory is only partially adequate; even if it held for all animals in mazes, to equate that aspect of behavior with classroom leaming is ridiculous. 2) Thus, an educational-psychology adaptation of the basic psychological theory would be inadequate even if the psychology were right. To apply the theory to a classroom forces on it a nurober of new simplifications, thereby further increasing its inadequacy. 3) The classroom teacher who applies such a doubly simplified practice to the classroom will be as culpable as is the refrigerator service man who insists on attaching as best he can his pressure gauge to the thermoelectric model. Fortunately, most teachers know enough not to believe the watered-down versions of the theory they have been told about.

On the one hand, then, information handling systems are far from standardized and are not likely to become so very soon. On the other, more and more expert management of information will be required. The implication that, therefore, one needs no solid theoretical understanding of the processes involved, I find untenable and even reprehensible, even (or partic-

ularly) in an introductory text. As Boswell quotes the good Doctor in his Life of Samuel Johnson, LL.D., (on the fifth of August, 1763) "This was a good dinner enough, to be sure; but it was not a dinner to ask a man to."

In short, a thin, watery gruel is defensible only for an invalid's diet. Although we may easily find ourselves glutted with a surfeit of too-rich stew, how much healthier it is to digest-a little at a time, if need be-the nourishing if sometimes gristly substance of a book like Cognition than it is to assume the sickly pose and frame of mind of the invalid, timidly ingesting the tasteless broth suitable only for the feebiest constitutions!

One parting shot at the technology and economics of technical book publishing. At $12.95 and $22.95 respectively, Management costs something like 5.8 and Cognition about 5.4 cents a page. These prices are higher than those of equivalent single Xeroxed copies at current prices. Surely there must be a more economical way to manufacture and market books. The advantage of printing, as I understand it, is its economy. If we end up paying relatively more for a mass-produced book, in terms of both money and delays, than for individually copied recent manuscripts, the advantages of printing become dubious indeed. A few possible alternatives are hinted at in the Howerton book; as a Society, we might consider a course of action such as that taken by the Association for Computational Linguistics, whose American Journal of Computational Linguistics reproduces by microfiche the articles, and uses small cards for the abstracts and notices to readers. As an impoverished academic, I for one would welcome less expensive and more rapid means of communication among us; even had I access to a generaus expense account, I would welcome less expensive but faster communication. In our FORUM, I hope we are moving in that direction. But consider: I am writing this page on October 10, 1974. I wonder when you are going to read it.

ASC CYBERNETICS FORUM Valurne VII, Nurober i-Spring 1975 21

ASC Forum N otes

Conclusions from the ASC Conference on

Communications and Control in Social Processes

Conference participants, on the assumption that a cybernetic orientation could contribute to the solution of contemporary problems of scientific and public policy, gathered at a concluding seminar out of which came the following recommendations:

1. Many emergent social problerns arise from circular flows of information or are the product of unrestrained mutual causation. Their solution should be sought in the underlying control processes involved.

Common to many current social problems is their sudden emergence. Population explosion, information overload, pollution, racial strife, the energy shortage, for example, are often perceived as "crises," not so much because they are entirely unpredictable but because of the failure of the social sciences to recognize their initially slow, but usually exponential, development before critical thresholds are reached. Research methods in the social sciences are largely geared to identify single, or at best multiple, causes of events but are relatively powerless when it comes to understanding the circular causal fabric of many social phenomena. Control engineering, on the other hand, has been preoccupied with morphostatic mechanisms seek:ing simple stabilities. Attempts at compensatory controls-as exemplified by responses to higher crime rate with more policemen-are likely to produce other unanticipated social problems; not to mention the high costs involved and limited resources available for such efforts.

It is suggested that social problems ernerging from the "vicious circles" of mutual causation, from the self reinforcing nature of social prejudices and from circular information flows can be anticipated by adequate techniques of analyzing positive feedback processes and that solutions to those problems might be found in the rearrangement of interdependent variables.

2. A better understanding of Social change requires a greater focus on structural formulations of social processes.

On the one band, most social systems models assume continuous linear changes. Merely increasing the number of variables in those models-a recent trend is exemplified by many global economic models or world models

of resource uses, etc.--does not change this fundamental limitation. On the other hand, many social changes occur in steps and might be considered qualitative in nature; revolutions, wars and catastrophies being extreme cases in point, more moderate examples are found in the social-structual adjustments to modern technology.

For example, the structural changes we are witnessing seem to be facilitated largely by the new technologies of communication and of information processing around which new industries, new institutional complexes, new organizational forms and new social practices have grown.

While linear models of social systems might prove appropriate under morphostatic conditions, they have proven incapable of understanding how quantitative changes evolve into structural changes; how morphogenesis takes place in society. To understand, predict and perhaps control ongoing, social processes, non-linear models of structural change are needed and should be developed.

3. In policy decisions and in social research, attention should be directed to the social consequences of differential accessability and unequal distribution of information.

While most decisional formulations assume knowledge to be equally distributed and sufficient for rational solutions to be forthcoming, differential access to modern information processing technologies and telecommunications has greatly enlarged the inequality in the distribution of information in society. And, by eroding the traditional socio-cultural controls on how knowledge is to be acquired and applied, this technology has favored the emergence of socio-technological controls that are hardly rationally comprehended or sufficiently understood. This unintended shift is manifest in numerous concerns such as the fear of the invasion of privacy, the fear of the increased institutional ability to

22 ASCCYBERNETICS FORUMVolume VII, Number I-Spring 1975

confine the flow of public information, the fear of loss of democratic control.

Attention should be given to the unanticipated social changes in community and in social structure caused by unequal access to, and distribution of, information, to the political, social and scientific value of information, to the costs of selective information processing including the costs of violating privacy, and to the interests that the selective dissemination of information might serve.

4. In planning policy, systems considerations should be given preference over sectional decision making.

Most policy planning is oriented to the maximization of separate goals in separate sectors such as health, education, pollution, international relations and the economy. These sectors are far from independent, however. Decisions in one sector are communicated and may · alter the conditions for decision making in another. And, what serves one sector weil may be detrimental to another. Sectional decision making, while politically expedient ignores the vitallinks that hold highly developed societies together and results in a sub-optimal quality of life.

Systems conceptions for policy decision that take account of these possibly complex interdependencies should be developed and cybernetics with its focus on feedback communication processes should be employed to achieve this end.

5. Contextual effects of technology transfer from industrial to less industrialized countries should be examined cybernetically.

Most decisions on the technology across national or cultural boundaries assume similar and desirable effects,

whereas evidence indicates that despite well intended technical aid programs, differences in standard of living between affluent and poorer countries in most cases are increasing. In addition, the political stability that such technology requires, tends, in fact, to be diminishing;."rö· a degree better than chance, technology transfer h·as aided military dictatorships and the Suppression of individual freedoms.

These discouraging facts are born out of a lack of understanding of the culturally rooted knowledge base of societies other than our own, the communicaiton networks that transform such knowledge into social action, and that channel social action back into knowledge and, the variables that are and that are not controllable in less industrialized countries. Technologies ate the product of fertile socio-institutional environments with which they interact and through which they grow. Decisions on technology transfer should be preceeded by an examination of how these "transplants" are taken by the "organism," the nature of the interactions they facilitate and the reorganizing processes they may set in motion. The development of adequate models of the social and political functions of technologies should be encouraged to make technology transfers more meaningful in the terms set forth by the aided countries themselves.

Klaus Krippendorff (Chairman)

F rederick Betz

Fred Haber Paul R. Kleindorfer

Noah S. Prywes Henry Teune

University of Pennsylvania Philadelphia, PA 191 74


Letters to the Editor

Editors Note: This is the first, and last, time that an exception will be allowed to the ASC CYBERNETICS FOR UM'S policy barring publication of unsigned and unattributable contributions.

- The Editor

Dear Editor:

The Club of Washington categorically denies Dr. Kopstein's petulant allegations. The truth is that the Society's membership, far from suffering resentfully a takeover by an elite group, has been remarkably supportive in turning its collective head the other way, and in energetically repressing its own views, if any.

We judge that this energetic abnegation is characteristic of two dasses of members. The first dass indudes those who have something substantial to contribute, either to the thought or the administration ( or both) of the Society, but who feel far too busy with other affairs to spare the effort or time. The second dass comprises those who have nothing to contribute but their dues, which they gladly pay for the privilege of basking in the reflected glory of our more distinguished and active members, as exemplified in The Club of Washington.

To the latter group, which has neither intellectual nor administrative resources, we say thanks for the financial support. To the former group, wherever they may be in the outlands, we extend a whole-hearted invitation to plug in to the nexus. Telecommunications have shrunk the world, while ground transportation in the Washington, D.C. area has made a myth of Kopstein's hypothecated proximity.

At the risk of falling to argument ad hominem, we call attention to the fact that Kopstein's address in the Autumn FORUM masthead was in the Washington, D.C. area while for the Winter issue, when his editorial app~ared, he had removed hirnself from the "nerve center." Will he, therefore , cease tobe a member of The Club?

As for Kopstein's recommendation that we consider now Margaret Mead's proposal of a switch-off mechanism for the Society, it takes little cognizance of the fact that we are now an organism, an open-loop system, with self-preservation and inertia potent vectors.

So long as The Club of Washington and the rest of the membership do their bit, and inflation or other external factors do not deal us a fatal blow, the Society will continue to subsist. Its Ievel of negentropy will be driven by those others, not now of The Club.

The Club of Washington

Editor's Note: We are pleased and proud to print the following Ietter to ASC President Roy Herrmann. Recognition of the significance of cybernetics is implicit in ASC's election to the AAAS, and, we feel, long overdue.

Dear J?r. Herrmann:

I am pleased to report that the AAAS Council, at its meeting on J anuary 31, elected the American Society for Cybernetics as an affiliate of the AAAS. We Iook forward to fruitful interaction between the members of the Society and the AAAS. Toward that end, you will soon receive a request from Mrs. Gwen Huddle, Director of our Membership Recruitment Department, that she is permitted to send your members invitations to join the AAAS ....

For your information about relations between AAAS and its affiliates, I endose a guide . Under separate cover we are sending you the AAAS Handbook.

I hope you will now Iet us know in which AAAS sections, up to five, the Society would like to enroll. .. We will submit your request to the section committees of the sections you indicate for their approval. .. and will later ask you to appoint a representative to the section committee of each section in which your enrollment has been approved.

.. .If there are further details you would like to have about relations between AAAS and its affiliates, please let me know.

William D. Carey Executive Officer American Association for the

Advancement of Science

24 ASC CYBERNETICS FORUM Volume VII, Number I - Spring 1975

Conference Calendar

30 April-2 May 1975 47th NATIONAL ORSA MEETING, TIMS 1975 NORTH AMERICAN MEETING, Palmer House , Chicago, IL. Meeting Chairman: Professor P. Pierskalla, IE/MS Department , Northwestern University , Evanston, IL 60201.

11-16 May 1975 1975 SESA SPRING MEETING of the Society for Experimental Stress Analysis on "Down-to-Earth Stress Analysis." Sheraton-Chicago Hotel, Chicago, IL.

15-17 May 1975 FOURTH ASIS MID-YEAR MEETING-INFORMATION ROUNDUP, sponsored by ASIS Pacific Northwest Chapter. A continuing education session on microform and data processing in the library and information center, to be held at the University of Portland , Portland, OR. Contact: Skip McAfee, jr. , ASIS, 1155 Sixteenth Street, NW, Washington , DC 20036; phone (202) 659-3644.

5-8 June 1975 SEVENTH ANNUAL MEETING OF CHEIRON; The International Society for the History of Behavioral and Social Sciences. Write to: W.E. Marshall, Department of Psychology , Carleton University, Ottawa, Ontario, Canada KIS .

16-20 June 1976-Call for Papers INTERNATIONAL FEDERATION OF AUTOMATIC CONTROL (I F AC) will hold a Symposium on " Large Scale Systems Theory and Applications" in Udine, Italy . Special attention will be given to papers whose main contribution lies in the application of the theoretical tools and techniques to the solution of practical problems arising in: industrial process control ; information processing ; bioengineering; economic systems ; environmental dynamics and control ; agricultural systems ; national, regional and urban planning; transportation systems; power systems ; systems for management and administration.

Only unpublished papers (max. 20 typewritten double-spaced pages) should be submitted in triplicate by 1 July 1975, addressed to G. Guardabassi,

Istituto di Elettrotecnica ed Elettronica. Politecnico di Milano, Piazza Leonardo Da Vinci , 32, 20122 Milano (Italy).

21 June-18 July 1975 FIFTH ANNUAL WORLD GAME STUDIES WORKSHOP, sponsored by Earth Metabolie Design, Inc. of New York Haven, CT and the University City Science Center of Philadelphia, to be held in Philadelphia , PA. Topics on "Comprehensive Anticipatory Design Science" - the effective application of the principles of science to the conscious design of our environment will be presented by R. Buckminster Fuller and others. Contact: Workshop/Earth M.D. , Box 2016 Yale Station, New Haven, CT 06520.

29 June-2 July 1975 INFO/EXPO 1975 INTERNATIONAL DATA PROCESSING CONFERENCE AND BUSINESS EXPOSITION. Data Processing Management Association, Mariott Motor Hotel, Atlanta, GA. Dr. Carl Hammer is a member of the Program Advisory Committee.

20-26 August 1975 THIRD WORLD CONGRESS OF THE ECONOMETRIC SOCIETY, University of Toronto , 150 St. George Street, Toronto 5, Canada. Program Committee Chairman: Professor Mare Nerlove, Department of Economics, Northwestern University, 629 Noyes Street, Evanston, IL 60201.

25-29 August 1975 THIRD INTERNATIONAL CONGRESS OF CYBERNETICS AND SYSTEMS, World Organization on General Systems and Cybernetics , Bucharest, Roumania.

3-8 September 1975 INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE (IJCAI), Tbilisi, USSR. General Chairman : Professor Erik SandewalL Papers to be submitted to Program Chairman: Professor Patrick H. Winston , The Artificial In telligence Laboratory , 545 Technology Square, Cambridge, MA 02139.

ASC CYBERNETICS FORUM Volume VII, Number i - Spring 1975 25

23-25 September 1975 1975 INTERNATIONAL CONFERENCE ON CYBERNETICS AND SOCIETY, IEEE SMC Society, at The Hyatt Regency Hotel, San Francisco, CA. The conference will cover the full ra~ge of the field of systems sciences and engineering. A special theme, however, will be "Technological Forecasting and Assessment: 197 5-2000."

26-30 October 1975 THIRTY-EIGHTH ASIS ANNUAL CONFERENCE, Sheraton Boston Hotel, Boston, MA. Theme: Information Revolution. Meeting will examinee the effects

of the emergence of new computer network, communication and reprographic technologies as weil as the accelerated growth of the research literature. Contact: Skip McAfee, jr., ASIS 1155 Sixteenth Street, NW, Washington, DC 20036; phone: (202) 659-3644. For information on submitting papers contact: ASIS 1975 Technical Program Committee, P.O. Box 224, Needham, MA 02194.

17-19 November 1975 NATIONAL ORSA/TIMES JOINT MEETING on "OR/MS and logistics", MGM Grand Hotel, Las Vegas, NV 89109.

26 ASC CYBERNETICS FORUM Volume VII, Nurober I - Spring 1975

About the Authors MICHAEL ARBIB ~- Michael Arbib was born in England,

raised in New Zealand and Australia, and received his Ph.D. from MIT. After this, he made lecture tours of the United States, Europe and the Soviet Union, as weil as teaching for a term each at Imperial College in London and at the University of New South Wales in Australia. He then spent five years

at Stanford University before moving to the University of Massachusetts at Amherst. He is currently Chairman and Professor of Computer and Information Science, Adjunct Professor of Sociology, and Director of the Center for Systems Neuroscience. He has published many books including: Brains, Machines, and Mathematics; The Metaphorical Brain (this book received the ASIS A ward for Best Information Seiences Book of 1973); and Theories of Abstract Automata.

ROLAND FISCHER

Roland Fischer is a cartographer of inner space and a biologist of the fleeting momen t. A multidisciplinary psychopharmacologist with a Ph.D. from Basel (Switzerland), he was the first to initiate and edit an In tl. Conference on In terdisciplinary Perspectives of Time [Ann. N.Y. Acad. Sei. 138, Art. 2,

367-916 (1967)] and to treat consciousness as a scientific topic (Science 174, 897-904 ).

From his 240 papers 130 were published while a professor of exptl. psychiatry and pharmacology at Ohio State University (1958-70). Now, he is a Research Coordinator at the Maryland Psychiatrie Research Center in Saltimore and a lecturer and Clin. Professor with The Johns Hopkins and Georgetown University Medical Schools.

FREDERICK KILE Dr. Frederick Kile is a Lutheran Minister (American Lutheran Chursh) and a systems engineer. He received his Th.D. from the University of Marburg, Germany; his B.D. from the Luther Theo1ogica1 Seminary; and his M.S. in electrical engineering and B.S. in Mathematics from the University of Wisconsin. Dr. Kile is presently

Research Consultant for Social Modeling for the Aid Association for Lutherans. He has written extensively on the subject of cybernetics applied to social concerns.

CHRISTOPHER LONGYEAR

Dr. Christopher Longyear is Associate Professor of English at the University of Washington at Seattle. He received his B.S. in Engineering Physics from Lehigh University and his M.A. and Ph.D. in English Language and Literature from the University of Michigan. While a Research Fellow at Harvard University, Dr. Longyear worked with

Warren McCulloch. Dr. Longyear was a member of the professional staff at the General Electric Advanced Study Center in Santa Barbara where he worked with Fred Thompson on naturallanguage question-answering systems. He has been an advisor on applied linguistics to the government of Pakistan.

RICHARD E. SCHNEIDER Member of the technical staff at The MITRE Corporation, Washington, D.C. Since 1961, has worked on a wide range of computer hardware and software applications.

He is in terested in several fields within Automata Theory, especially brain research. Last summer he hosted a conference on Biologically Motivated Automata Theory

at MITRE. Does adjunct teaching in Switching and Automata Theory.

Earned a BSEE and MSEE, is a member of SIG/ ACT in ACM, and a member of IEEE Computer Group.


ASC PUBLICA TIONS ORDER FORM American Society for Cybernetics • 1025 Connecticut Avenue, Suite 911 • Washington, DC 20036

Piease send me the following ASC publications. My check, payable to ASC in the amount of $ _____ is enclosed.

No. of Copies

___ 1967:

___ 1968 :

___ 1969:

___ 1970:

___ 1971:

Proceedings

1st Annual Symposium, "Purposive Systems"

2nd Annual Symposium, "Cybernetics and the Management of Large Systems"

3rd Annual Symposium, "Cybernetics, Simulation, and Conflict Resolution"

4th Annual Symposium, "Cybernetics, Artificial lntelligence, and Ecology"

Fall Conference, "Cybernetics Technique in Brain Research and the Educational Process"

Journal of Cybernetics-Annual Subscriptions;

___ 1971: Vol. 1, 1-4

___ 1972: Vol. 2, 1-4

___ 1973: Vol. 3, 1-4

ASC FORUM-Annual Subscriptions

___ 1972: Vol. IV, 1-4*)

___ 1973: Vol. V, 1-4*)

___ 1974: Vol. VI, 1-4

*Single copy: $2.50

Total Cast @$12.50 = $ ___ _

@$12.50 = $----

@$12.50 = $----

@$12.50 = $----

@$ 7 .50 = $----

@$20.00 = $ ---

@$20.00 = $ ---

@$20.00 = $----

@$10.00 = $ ___ _

@$10.00 = $---

@$18.00 = $ ---

TOTAL=$====

Name _______________________________________ _

Address ------------------------------------------------------------------

Zi p Code ---------------

MEMBERSHIP AND RENEWAL APPLICATION

GENTLEMEN:

Please consider my application for membershipfrenewal in the American Society for Cybernetics. Annual dues are $20 for members and $5 for students. Dues include the quarterly ASC CYBERNETICS FORUM, and the quarterly Journal of Cybernetics.

Return tagether with your check (payable to American Society to Cybernetics) to:

Mr. Gary D. Bearden, ASC Suite 911, 1025 Connecticut Ave., N.W. Washington, D. C. 20036

NAME ADDRE~S~S---------------------------------------------------------------

CITY _____________ STATE ----------- ZIP -------------

TITLE/OCCUP ATION ------------------------------------------------------

ORGANIZA TION/ AFFILIATION ----------------------------------------------

SIGNATURE ----------------------------------------------------------

Statement of Editorial Policy

The ASC CYBERNETICS FORUM is an internationally distributed quarterly publication of the American Society for Cybernetics. lt is published to promote the understanding and advancement of cybernetics. I t is recognized that cybernetics covers a very broad spectrum, ranging from formalized theory through experimental and technological development to practical applications. Thus the boundaries of acceptable subject matter are intentionally not sharply delineated. Rather it is hoped that the flexible publication policy of the CYBERNETICS FORUM will foster and promote, the continuing evolution of cybernetic thought.

The CYBERNETICS FORUM is designed to provide not only cybernetics, but also intelligent laymen, with an insight into cybernetics and its applicability to a wide variety of scientific, social and economic problems. Contributions should be lively, graphic and to the point. Tedious listings of tabular material should be avoided .

The Editors reserve the right to make stylistic modifications consistent with the requirements of the CYBERNET/CS FORUM. No substantive changes will be made without consultation with authors . They further reserve the right to reject manuscripts they deem unsuitable in nature, style or content.

Opinions expressed in aritcles in the CYBERNETICS FORUM do not necessarily reflect the opinion of the CYBERNETICS FORUM or its editors, or the American Society of Cybernetics or its directors and officers. All material published in the CYBERNETICS FORUM is Copyright by the American Society for Cybernetics who reserve all rights .

lnstructions to Authors

Papers already published or in press elsewhere are not acceptable. Foreach proposed contribution, one original and two copies (in English only) should be mailed to The Editor, ASC CYBERNETICS FORUM, Suite 911, 1025 Connecticut Avenue, N.W., Washington, D.C. 20036. Manuscripts should be mailed flat, in a suitable envelope. Graphie materials should be submitted with suitable cardboard backing.

Types of Manuscripts: Three types of contributions are considered for publication: full-length articles, brief communications of 1,000 words or less, and letters to the editor. Letters and brief communications can generally be published sooner than full-length manuscripts. ßooks, monographs and reports are accepted for critical review. Two copies should be addressed to the Editor.

Processing: Acknowledgment will be made of receipt of all manuscripts. The ASC CYBERNETICS FORUM employs a reviewing procedure in which all manuscripts are sent to two referees for comment. Whcn both referees have replied, copies of their comments are sent to authors with the Editor's decision as to acceptability. Authors receive galley proofs with a five-day allowance for corrections. Standard proofreading marks should be employed.

Format : Manuscripts should be typewritten double spaced, on white bond paper on one side only, leaving about 3 cm (or 1.25 inches) of space around all margins

of standard letter-size paper. The first page of the manuscript should carry both the first and last names of all authors and their affiliations including city, state and zip. (Note address to which galleys aretobe sent.) All succeeding pages should carry the last name of the first author in the upper right-hand corner.

Style: While the CYBERNETICS FORUM demands a high standard excellence in its papers, it is not a scholarly or technical journal. Authors should avoid mathematical formulae and long lists of references or footnotes . Titles should be briefand specific, and revealing of the nature of the article. Acknowledgments and credits for assistance or advice should appear at the end of articles. Subheads should be used to break up- and set off- ideas in text.

Graphie Materials : All artwork submitted must be in finished form suitable for reproduction (black on white) and !arge enough so that it will be legible after reduction of as much as 60%. Photographs should be black and white glossy no !ess than 5" X 7".

About the Authors : A brief biography (less than one page), along with a small photograph, must be sent with all manuscripts. This will be included in the "About the Authors" section of each issue.

Manuscript Return : Authors who wish manuscripts returned must include a stamped, self-addressed envelope along with their manuscripts .

SUITE 911 1025 Connecticut Ave., N.W. Washington, D. C. 20036

Nonprofit Org.

U.S. POSTAGE

PAID

Washington, D.C.

Pennit No. 44085

TIME VALUE

CYBERNETICS FORUM - univie.ac.at · CYBERNETICS FORUM A PUBLICATION OF THE AMERICAN SOCIETY FOR...

Documents

Transcript of CYBERNETICS FORUM - univie.ac.at · CYBERNETICS FORUM A PUBLICATION OF THE AMERICAN SOCIETY FOR...