Economics and Artificial Intelligence: Proceedings of the Ifac/Ifors/Ifip/Iasc/Afcet Conference, Aix-En-Provence, France, 2-4 September, 1986

186 Pages·1987·26.93 MB·English

IFAC PROCEEDINGS SERIES
Editor-in-Chief
JANOS GERTLER, Department of Computer and Electrical Engineering,
George Mason University, Fairfax, Virginia, USA
GERTLER 8c KEVICZKY {General Editors): A Bridge Between Control Science & Technology
(Ninth Triennial World Congress, in 6 volumes)
Analysis and Synthesis of Control Systems (1985, No. 1)
Identification, Adaptive and Stochastic Control (1985, No. 2)
Large-scale Systems, Decision-making, Mathematics of Control (1985, No. 3)
Process Industries, Power Systems (1985, No. 4)
Manufacturing, Man-Machine Systems, Computers, Components, Traffic Control,
Space Applications (1985, No. 5)
Biomedicai Applications, Water Resources, Environment, Energy Systems, Development, Social
Effects, SWIIS, Education (1985, No. 6)
BARKER 8c YOUNG: Identification and System Parameter Estimation (1985) (1985, No. 7)
NORRIE & TURNER: Automation for Mineral Resource Development (1986, No. 1)
CHRETIEN: Automatic Control in Space (1986, No. 2)
DA CUNHA: Planning and Operation of Electric Energy Systems (1986, No. 3)
VALADARES TAVARES 8c EVARISTO DA SILVA: Systems Analysis Applied to Water and Related
Land Resources (1986, No. 4)
LARSEN & HANSEN: Computer Aided Design in Control and Engineering Systems (1986, No. 5)
PAUL: Digital Computer Applications to Process Control (1986, No. 6)
YANG JIACHI: Control Science 8c Technology for Development (1986, No. 7)
MANCINI, JOHANNSEN 8c MARTENSSON: Analysis, Design and Evaluation of Man-Machine
Systems (1986, No. 8)
BASANEZ, FERRATE 8c SARIDIS: Robot Control "Syroco '85" (1986, No. 9)
JOHNSON: Modelling and Control of Biotechnological Processes (1986, No. 10)
TAL': Information Control Problems in Manufacturing Technology (1987, No. 1)
SINHA 8c TELKSNYS: Stochastic Control (1987, No. 2)
RAUCH: Control of Distributed Parameter Systems (1987, No. 3)
HAASE: Software for Computer Control (1987, No. 4)
MARTOS, PAU 8c ZIERMANN: Modelling and Control of National Economies (1987, No. 5)
GENSER: Control in Transportation Systems (1987, No. 6)
KUZUGU 8c TUN ALL Microcomputer Application in Process Control (1987, No. 7)
WANG PINGYANG: Power Systems and Power Control (1987, No. 8)
BALCHEN: Automation in Aquaculture (1987, No. 9)
YOSHITANI: Automation in Mining, Mineral and Metal Processing (1987, No. 10)
GEERING & MANSOUR: Large Scale Systems; Theory and Applications (1987, No. 11)
ROOS: Economics and Artificial Intelligence (1987, No. 12)
TROCH, KOPACEK 8c BREITENECKER: Simulation of Control Systems (1987, No. 13)
KAYA & WILLIAMS: Instrumentation and Automation in the Paper, Rubber, Plastic and Polymerization
Industries (1987, No. 14)
NOTICE TO READERS
If your library is not already a standing/continuation order customer or subscriber to this series, may we recommend that you place a standing/
continuation or subscription order to receive immediately upon publication all new volumes. Should you find that these volumes no longer serve your
needs your order can be cancelled at any time without notice.
Copies of all previously published volumes are available. A fully descriptive catalogue will be gladly sent on request.
ROBERT MAXWELL
Publisher
ECONOMICS
AND
ARTIFICIAL INTELLIGENCE
Proceedings of the IFACIIFORSIIFIPIIASCIAFCET Conference
Aix-en-Provence, France, 2—4 September 1986
Edited by
JEAN-LOUIS ROOS
I.N.S.E.E., Aix-en-Provence, France
Published for the
INTERNATIONAL FEDERATION OF AUTOMATIC CONTROL
by
PERGAMON PRESS
OXFORD · NEW YORK · BEIJING · FRANKFURT
SÄO PAULO · SYDNEY · TOKYO · TORONTO
U.K. Pergamon Press, Headington Hill Hall, Oxford OX3 OBW, England
U.S.A. Pergamon Press, Maxwell House, Fairview Park, Elmsford, New York 10523, U.S.A.
PEOPLE'S REPUBLIC Pergamon Press, Room 4037, Qianmen Hotel, Beijing, People's Republic of China
OF CHINA
FEDERAL REPUBLIC Pergamon Press, Hammerweg 6, D-6242 Kronberg, Federal Republic of Germany
OF GERMANY
BRAZIL Pergamon Editora, Rua Eça de Queiros, 346, CEP 04011, Paraiso, Säo Paulo, Brazil
AUSTRALIA Pergamon Press Australia, P.O. Box 544, Potts Point, N.S.W. 2011, Australia
JAPAN Pergamon Press, 8th Floor, Matsuoka Central Building, 1-7-1 Nishishinjuku, Shinjuku-ku,
Tokyo 160, Japan
CANADA Pergamon Press Canada, Suite No. 271, 253 College Street, Toronto, Ontario, Canada M5T 1R5
Copyright © 1987 IFAC
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in
any form or by any means: electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or other
wise, without permission in writing from the copyright holders.
First edition 1987
Library of Congress Cataloging in Publication Data
IFAC/IFORS/IFIP/IASC/AFCET Conference on Economics
and Artificial Intelligence (1986: Aix-en-Provence, France)
Economics and artificial intelligence.
(IFAC proceedings series; 1987, no. 12)
Includes indexes.
1. Economics—Data processing—Congresses.
2. Artificial intelligence—Data processing—
Congresses. I. Roos, J.-L. II. International
Federation of Automatic Control. III. Title. IV. Series
HB 143.5.144 1987 330\028'563 87-6943
British Library Cataloguing in Publication Data
Economics and artificial intelligence:
proceedings of the IFAC/IFORS/IFIP/IASC/
AFCET conference, Aix-en-Provence, France,
2-4 September 1986.—(IFAC proceedings
series; 1987, no. 12).
1. Economics—Data processing
2. Artificial intelligence
I. Roos, J.-L. II. Series
330'.028'563 HB 143.5
ISBN 0-08-034350-3
These proceedings were reproduced by means of the photo-offset process using the manuscripts supplied by the
authors of the different papers. The manuscripts have been typed using different typewriters and typefaces. The
lay-out, figures and tables of some papers did not agree completely with the standard requirements: consequently
the reproduction does not display complete uniformity. To ensure rapid publication this discrepancy could not be
changed: nor could the English be checked completely. Therefore, the readers are asked to excuse any deficiencies
of this publication which may be due to the above mentioned reasons.
The Editor
Printed in Great Britain by A. Wheaton &f Co. Ltd., Exeter
IFAC/IFORS/IFIP/IASC/AFCET CONFERENCE ON
ECONOMICS AND ARTIFICIAL INTELLIGENCE
Organized by
Association Française pour la Cybernétique Économique et Technique
(AFCET)
Sponsored by
International Federation of Automatic Control (IFAC)
Co-sponsored by
International Federation of Operational Research Societies (IFORS)
International Federation for Information Processing (IFIP)
International Association for Statistical Computing (IASC)
Association Française pour la Cybernétique Economique et Technique
(AFCET)
International Scientific Committee
J.-L. Le Moigne, France (Chairman) E. Kuh, USA
Y. Anzai, Japan J.-C. Lauriere, France
P. Bourgine, France J. Lesourne, France
J. Browne, Ireland P. Mertens, FRG
J.-M. Chambers, USA M. Muller-Merbach, FRG
M. Chein, France D. Pascot, Canada
J.-C. Courbon, Switzerland J. Pi trat, France
G. Doumeingts, France A. Rolstadas, Norway
P. Falster, Denmark J.-L. Roos, France
W. A. Gale, USA M. Shakun, USA
J.-D. Goldhar, USA L. Siklossy, Netherlands
M. Gondran, France H.-A. Simon, USA
H.-W. Gottinger, FRG J.-P. Van Gigch, USA
D. Hertz, USA R. Welsch, USA
P.-E. Johnson, USA Z. S. Zannetos, USA
A. Kay, USA M. Zeleny, USA
National Program and Organization Committee
P. Bourgine (Chairman) C. Ernst
R. Arrus C. Giraud
J.-M. Attonaty J. Granier
M. Aucoin B. Guilhon
B. Baixe J.-C. Hubert
J.-C. Bertrand A. d'Iribane
E. Bianco J. Laporta
J.-L. Bodin P. Levine
J.-F. Boisvieux J.-P. Marciano
J. Boussard A. Moisan
G. Bramoulle B. Munier
N. Cot J.-C. Rault
G. Charbonnel Ph. Renard
J.-P. Dalloz A. Recoque
M. Dappe D. Retour
A. Dussauchoy E. Sanchez
M. Egea T. Tabourier
H. Tardieu
Copyright © IFAC Economies and Artificial
Intelligence, Aix-en-Provence, France, 1986
PREFACE
Can we meet the challenge? Is It possible to summarize some oh the key arguments which can be considered
as the, basic characteristics oh a rather now mating oh some rather old disciplines. The, ^ÂMt interna
tional Coherence on Economico and Artlhiclol Intelligence create an excellent opportunity to observe
and to understand the. nature oh this rather new product which seems to emerge, i^om this melting pot :
ahter all, it AS not so ire.qme.nt to meet during three day* people, coming firom 25 di^ere-nt countries,
working in six or seven rather dl^exe-nt technical and scientihic areas,... who agree to speak altogether
and who consider that they are. able. oh some mutual understanding : their languages appear to be. rather
common although they don't perceived themselves initially as belonging to a common culture, : statistician,
management scientist, computex scientist, economist, systems engineer, and some, othexs including
economics and neuro-scientist !
Each participant oh the, Aix-e.n-Prove.ncz''s Conference, has pxobably his original interpretation oh this
challenge.; we, cannot - and we, don't want to - propose, a sort oh general and common position, summarizing
and uni^ormizing... the. basic characteristics oh our emerging scientific f^ield. Many members oh
the. International Scientihic Committee and oh the, National Program and Organization Committee have, help
us, through many letters and discussions, to design, formulate., redesign and rehormulate the, pattern
which support the, iinal program and latex the. proceedings oh the, conference,. But we, are. not able, to
summarize, in some, words their collective, opinion. We have, the. he.ellng that, ih we, cannot collectively
meet the. challenge,, we. can, at least, meet it personalty : the, had that the. chalxmen oh the, two commitles
initially appointed by IFAC [in close, cooperation with 1E0RS, and later with IEIP and JACS), had to work
closely altogether during more, than two years to prepare, the. coherence. (?) gives us a unique, opportunity
to try, day ahter day, to design our own synthesis.
May we, present it through hour key propositions?
PROPOSITION nb. 7
intelligible., understandable? Can we. avoid such
questions each time. we. discuss the, good use. oh
[perhaps rather too provocative ?).
reason in human ahhaixs"? (2)
Artlhicial intelligence is born - hon- an impor
tant share, - h^om Economics, between 7936 and 1'Jhe.n we. consider the, history oh economics and
operations research and oh cybernetics and
7956.
artlhicial intelligence., between 7936 (3) and
7956 [the. ohh^clal birthdate oh A.I., at the.
We argue, that it was the. questions raised by
Vaxthmouth summer seminar), we. can see. the. role.
some, young economisti* during the, Great
oh this permanent re.hle.xlon on the. complexity oh
depression, which lead them, through a natural
the, use. oh human reason in human ahhairs, not
intellectual evolution, to design and to
initially related to the selh standing development
developp this new discipline, called - since,
oh cybernetics and computer-science.
7956 - Artlhicial Intelligence,.
Vor sure, the impressive article oh E.J. Varela
One, oh them, today worldwide, known as Nobel
in this volume will highlight this point, artl
Prize., Herbert A. SIMON, has perfectly well
hicial intelligence and cognitive sciences also
described this cultural progression : to think
about the. ressources allocation's decisions, borrow to neurocybernetics and connexlonism. But
was to think about the. decision-making process-, not only, and perhaps, not mainly.
to think about the, dealston-making process was
to think about the. thinking process; and to (2) Can we suggest that one of most recent H.A.
think about the. thinking process was to think Simon's Book, titled : "Reason in Human Affairs"
about the. intelligence, process : how can we. (Stanford University Press, 1983), will become
hormallze. it enough in order to perceive, it as the bedside book of most of artificial intel­
ligence researchers as of politics, economics
and management scientists?
The idea of this conference is born during
the preparation of a doctoral seminar devo­
ted to economics and artificial intelligence The first papers of Alan Turing, H.A. Simon
mainly with the direct contribution of V, and CE. Shannon were published in 1936.
r
J.L. ROOS and of Pr. B. MUNIER
Vili Preface
Also can we consider today that the too {ami-
liar concept o{ "Application o{ A.I. to Econo PROPOSITION/ nò. 3
mics, or to O.P., or to management and social
sciences", wilt probably be s clero sing [a sort
The conjunction o{ Economic* and Artl{lclal
o{ Incestuous relationship!). U/e cannot apply
Intelligence constitutes an excellent common
A.I. to Economics as we try to apply Mathematics
{ramework supporting the. development o{ s~öme
to Economics or to social sciences, because A,I.
dl{{erent organizational engineerings which
and Economics are at the same level, or belong
ani..becoming'more and more Imposant in the
to the same tree. Any good progress o{ A.I.
day to day ll{e o{ social complex organizations.
will be a progress {or Economics, and any good
progress {or Economics will be a progress {or
Moot o{ the participant* o{ the Aix-en-Provence1
A.I. That I* to say that the conjunction
Con{erence perceived Initially themselves as
[Instead o{ the application) o{ Economics and
belonging to rather dl{{erent areas o{ sclentl{i
A.I. [and computation sciences) wilt be a priori
and technological research :
{rutt{ull {or both disciplines.
* Organizational and Managerial Information
Systems [Design and development),
PROPOSITION nb. 2
% Decision Support Systems,
* Vota bobe, and knowledge baòe , Management
The conjunction o{ Economics and Artl{lclal
Systems and related area* o{ Conceptual
Intelligence can be a triggering {actor {or
Modeling
a new development o{ Operations Research.
* Model* and Method* o{ Stati*tic* and Economic
Since ten year* at least [4), Operations
E o reca* ting
Research try to renew lt*el{. during the
sixties, this young discipline became acade
* Planning and Control Systems {In Industrial
mically respectable with the strong support
a* In Administrative Organization*).
o{ applied mathematics : the "applications"
o{ mathematics to economics [management and
At the end o{ the Conference, each o{ them was
social science* was Initially rather {rult{ul;
rather positively impressed by the {act that
and O.K. was able to Irrigate In a very
they had been able to communicate altogether
e{{ective way the {leid* o{ economic* and
without any di{{lcultles : it was not a Babel
management. But l{ Ô.R. was "giving", It was
Tower! Thli probably means that the conceptual
not "receiving" - or It was not able to
reference on which the conference was designed,
receive : searching too exclusively the "good"
constitutes an elective common and solid scien
algorithms [optimization algorithms), it
tific and technological {ramework, on which
was not able to hear the message proposed In
1958 by H.A. Simon and A. Uewell -."Heuristic those various {leid* can develop themselves
Problem Solving : the next advance In and exchange easily with the other* : the
Operation* Research" (5). This stagnation o{ conjunction o{ Economics and Artificial Intel
O.P. [in term o{ sclenti{lc development and ligence build today more than the Cybernetics
renewal) during the seventies I* now well o{ the sixties, a solid scientific foundation
perceived and understood by most o{ the {or the development o{ those new organisational
researchers and practltlonner* : they have engineerings. And those engineerings are,
the strong {eeling that the general problema through their own development* and Interactions,
tic o{ Arti{iclal Intelligence will propose contributing to the development o{ this basic
an excellent {ramework {or re-assesslng the conjunction.
{uture development o{ O.R.; they have perhaps
to avoid the naive temptation to substitute Instead o{ applying model* designed elsewhere,
Applied Arti{iclal Intelligence to Applied they are designing themselves their own,
Mathematics to solve the many problems o{ specific and multiple model*, re{ereeing to the
Operation Research. But, l{ they remember same concept*. o{ knowledge representation, o{
that "the next advance In O.R."... Is In formal reasoning,o{ conceptual modeling or o{
lt*el{, and not at a metalevel, they wilt complexity modeling (6).
develop a new conjunction o{ O.P. and A.I.
Instead o{ an "application". During the Alx-
en-Provence's Con{erenee, we had the {eeling PROPOSITION nb. 4
that most o{ them will do so.
Complex problem solving Is no longer a question
o{ optimizing technics : it becomes a question'
on. Intelligent design.
The "paradigm o{ Homo-Economlcus" has {or a
long time dominated most o{ our method* o{
economics and management problem solving : the
richness o{ the use o{ human reason was reduced
to the strict constraint o{ a {ormai logic,
[4) Remember the well known article of R.L. and the variety o{ human and social behavior
Ackoff : "The future of O.R. is past"! was reduced to a mo no criteria system, such as
(Journal of O.R. Society, Vol.30, nb 2, "to optimize a utility or a cost {unction". Vor
pp 93-104, Pergamon Press Ltd, 1979). sure, since hi* doctoral thesis, {orty year*
ago, H.A. Simon [amongst some others) suggest*
(5) H.A. Simon and A. Newell : "Heuristic
Problem Solving : the next advance in
(6)See for instance : United Nation University :
Operations Research". In Operation
"The Science and Praxis of Complexity", Tokyo,
Research, Vol.6, nb 1, Jan-Feb. 1958,
Japan, 1985
pp 1-10.
Preface
ix
to rethink ou/L problem solving methods, reheering common epitimo logical foundation : let us cali
to the "paradigm oh Homo-Co gitans" : human them, with H.A. Simon : The Systems Sciences,
reason monk in a "bounded rationality" environ the Sciences oh the Artificial .
ment, but is able not only "to design" and "to
invent", but also to develop many horms oh
Paul BOURGINE and Jean-Louis LE MOIGNE
intelligence : the {vu>t steps o\ a problem
solving process can be a problem-finding and a Chairman oh the Chairman oh the
problem netting process, and this intelligent Motional Program International
pK.oo.ebt> can be hormalized enough to become and Organization Scientific Committee
understandable. And the òecond steps axe to Committee
recognize the multiple criteria^ through which
human beeing tò able to consider any problem :
the search hor "satisdicing solutions" is not
only mone economic, it is also, and mainly, more
(7) A. Newell : "Heuristic programming : ill
elective. In 1968, A. Newell summarized those
structured problems", in "Progress in
methods in an a/iticle which hat> indJjiectly ins
Operation Research, Vol. Ill", J. Aronofsky
pired the design oh the Aix-en-Provence''s
ed., pp 360-413. (Publications in Operations
Coherence : the {^eaòibitity oh the methods
Research, nb 16, 1969, J. Wiley & S., N.Y.
booed on the Paradigm oh Homo-Cogitano was
clearly underlined just as the beginning oh the
impressive di^uòion oh the computer-bo*ed
modela in complex organization : "Heuristic
programming : ill structured problem" (7) : one
oh the more practical conclusions oh the confe
rence it> perhaps that the initial proposition
oh A. Newell and H.A. Simon was correct : we ACKNOWLEVGEMENT
can renew our problem solving technics, when we
agree on the ^act that heuristic problem solving
are as rational as algorithmic problem solving. Each scientific conference needs so many helps
Here is one oh the basic foundation oh Artificial coming ^rom so many people, so many institutions,
Intelligence and cognitive sciences-, here also so many countries that we cannot expect to
is one oh the basic foundation oh Economics express all the thanks we must say to each oh
viewed as ... the elective use oh reason in them. The reader may particularly remember the
human avoirs! Ih we now read most oh the papers decisive contribution oh the sponsoring orga
oh the proceedings having in mind this obser nizations : IFAC, IVORS, IVIP, IACS, NASA,
vation, we s halt see the elective emergence oh with AECET and the University oh Aix-Marseille
programmed heuristics in the operational models III, as oh most oh the members oh the Interna
we are all developping fior assessing and solving tional and hlational Committees .We wish to express
the complex problem oh our modern organizations . our personal gratitude to the editor oh the
proceedings, who has worked so hard since more
than two years to promote and to organize the
coherence, Jean-Louis ROOS [hrom the Vrench
"Time is now"! Institut National de la Statistique et des
Etudes Economiques), and to Françoise LABAN,
Thu> conclusion is suggested by I.S. lannetos in doctoral Student attached to the "Groupe de
his rich paper. The ßinal meeting oh the Aix- Recherche en Analyse de Systeme et Calcul
en-Provence}s Conference, chaired by Pr. V.B. Economique" (Université d1Aix-Mars etile 1J1,
Hertz, gives us the strong Meeting that we are EaculXe d'Economie Appliquée, CNRS 935, headed
collectively reaching a sort oh maturation : by Pr. B. MUNI ER), who had in charge the burden
Economics and Artificial Intelligence, Operation oh the local secretariat oh the coherence,
Research and Cybernetics, Management Sciences behore, during and ahter the h^^i CECOIA meeting.
and Computer Sciences, Planning and Control
Theories and Statistical Sciences... are now
mature enough to identify and to design their Paul BOURGINE and Jean-Louis LE M0IGNE
Copyright @ IFAC Economics and Artificial KEYNOTE ADDRESS
Intelligence, Aix-en-Provence, France. 1986
TRENDS IN
COGNITIVE SCIENCE AND TECHNOLOGY
F. J. Varela
CREA, Ecole Polytechnzque
and Institut des Neurosciences, CNRS-Universite de Paris VI
1. INTRODUCTION that cognition can be m a s co mputatlons of symbollc
representations Clearly this orientation could not have emerged without
the basis laid during Stage I The main dlfference is that the original
The science and technology of cognition (STC) is a hybrid of several tentative idea is here promoted to a full blown n~~0the515w. itn a strong
disciplines. each one having its own flavor and commitments, in strong desire to set its boundaries apart from Its broader, exploratory, and
resonance with the others (Fig 111 interdicipltnary roots where the social and biological msciences figurJed
preeminently wlth all thelr multlfarious complexlty is a
convenient label for this large but well-delineated orientation, that has
STC is a little over 40 years old It Is & establlshed as a mature motivated many sclentific and technological developments since 1956, in
science with a clear sense of direction and a large number of the areas of cognltive psychology, linguistics, a large part of
researchers constituting a communlty, as is the case of, say, atomic neurosciences and, of course, artificial intelligence
physics or molecular biology Accordingly, the future development of
STC is far from clear, progress in the field is based on darlng conceptual
bets, somewhat llke trylng to put a man on the moon without knowing
where the moon Is h ~ ~ * ~
The cognitivist research programme can be summarized as
The purpose of this paper 1s to bare outlines of the UCEDLstate of answers to the following questions
affairs of STC We shall do so by depicting STC as having four tiers
which are conceptually quite distinct, and having emerged In roughly
successive moments of time over these 40 years These four stages are Questlon '1
the following What is cognition?
Answer
Information processing Rule-based manipulation of symbols
Stage I The foundational years (1943-53), which will not Question' 2
concern us here2 , HOW doe5 it work?
Stage 2 The cognltlvist paradlgm, Answer
Stage 3 The alternative to symbol manipulation, Through any devlce whim can support and manipulate aiscrete pnysical
Stage 4 The alternative to representations elements the symbols The system interacts only with the form of the
symbols (their physlcal attrlbutes), not their meaning
Through thls four-tiered description we will examine the basis or (luestlon '3
what is already established as a clearly outlined paradlgm (Stages I and How do I know when a cognitive system is functioning adequately?
2),a nd the essential fact that thls established paradigm is in the Answer
process of making way for new, emerging perspectlves (Stages 3 and 4) When the symbols approprlately represent some aspect of the real
This challenging heterodoxy has the potential for deep changes world, and the Information processing leads to a successful solutlon of
the problem glven to the system
Not only has thls bold research programme become fully
established, but it is now -with cognitive Science itself Few
2. STAGE TWO: THE COGNlTlVlST HYPOTHESIS. among its active participants, let alone In the public at large, are
sensitive to elther its roots, or its current challenges and alternatives
"The brain processes informatlon from the outside worle Is a household
-*-
phrase understood by everybody To say that this statement is, at best,
misleadlng, sounds odd, and the ensuing conversation will inmrneaiately
be labelled as belng 'phllosophical' This is a blindness in contemporary
Just as 1943 was clearly the year in which the cybernetics phase
common sense Introduced in our culture by the cognitivist paradigm. and
was born, so was 1956 clearly the year which gave birth to the second
the greatest danger for being unable to perceive a broader vision and
phase of STC During this year, at two meetings held at Cambridge and
future for STC
Dartmouth, new voices like those ot Herbert Simon Noam Chomsky,
Marvin Minsky and John McCarthy put forth ideas which were to become
the major guidelines for modern cognitive science3
w-
The central lntution is that Intelligence (includinq human
lntelligencei so resembles a computer in its essential characteristics
The manlfestations of the cognltivism are nowhere more visible
than in artificial intelllgence (Al) which is the-oor the
cognitivist hypothesls Its complementary endeavour Is the study of
I I isuasi~let oarg,e tnatcognirivescienoe (CS) is8 separatediscipline.n avino artificial natural, biologically Implemented cognltlve systems, most speclally
intellipsnce (A ;m its tecnnolcqical pole um- tnw tnw sno.ia not impw twiner m we man Here, too, computatlonally characterizable representations have
anfBfEecmtsnaell ntheer ep ahrwtlcsivpearti.n ignd 0i.r scipvliimne st,h inecaluu~dliinmgio~ialt sseicdlf i stinction c.rsmpr IMn this and been the main explanatory tool Mental representations are taken to be
2 FM IhlsSIepsaeour reOsntmllectiveworkonthe neglecledhislwyoiBerlyn/bernetics
andself-orwlzatlon publlshedas MlersCQL4 NQ7-9 and In partlcular lhearllcleby
J.P.Dupuy, "L'skfde lapremiere~bernetlque"N, Q7,p p.7-140. Thsmlyoltmr useful 3 &a H.brdner, wcit, Chwler 5 for thls pricd
scurcalsSHalrns, ~nv~Alwmdnnmd~~fWi~~MIT1P98r0e.Tshesr,em tbmk 4 for this Section I hwe profiled much from D.Andlrr's. '"Ccqvlivism--ffthcd% w
by H.Owdner, ~ ~ ~ i ~ s ~ ~ i ~ A n i s t ~ y o f Btm~ic B~&si, t i ~ R ~ oithuerwt1ri.A~ new, p hsss?",P roc.Gmferencs ~minNM~ofJ&I?IIl~Altmns,Ju19m8,4 ,
1985, dlscussss thls perlcdonly In every superficial way. ( forthmrnrng).
1
2 F. J. Varela
occurrences of a formal system and the minds activity is #hat qives continued search for -processing algorithms has met with little
these representations their attitudinal color beliefs desires plans and success because the entire computational philosophy runs precisely
so on Here therefore unlike Al we find an interest in what the counter it
cognitive systems are really like and it is assumed that their coqnitive
representations are something & the system (they are said to 00 The second important limitation is that symbolic processing is
intentional) b.uLu&I the loss or mdlfunctlon of any part of the cognitive system
implies a complete breakdown of the system In contrast a distributed
For example subjects exposed to geometric tigures ana asked to operation is highly desirable, so that there is at least a relative
rotate them in their head consistently reported that the ditficulty of the equipotentiality and immunity to mutilations
task depended on the number of degrees of freedom in which the figure
had to be rotated That is everything happens as though we a have mental These two disappointments 81th cognitivism can be seen as one
space where figures are rotated like on a television screens In due time the architectures and mechanisms are tar from biologiral indeed the
these experiments produced an explicit theory postulating rules by which most ordinary visual tdsks done even by tiny insects are done faster
the mental space operates slmilar to those used on computer displays than IS physically possible when simulated in a sequential manner
Operating on stored data These researchers proposed that there is an Similarly, the resiliency of the brain to damage without coinpromising
interaction between language-like operations and picture-like all of its competence, has been Known to neurobiologists tor a long time
operations, and together they generate our internal eye6 Thls approach
ha5 generated an abundant llterature for and against', and every level of W h o t # Q O y ~ ?
the observations has been given alternative interpretations However,
the study of imagery is a perfect example of the way the cognitivist The above suggests that instead of starting with symbols, one
approach proceeds when studying mental phenomena could Start with simple components which would connect to one another
In dense ways In this approach each component operates only in its
J.lb+ldh~&lbthobWlh environment but because of the network quality of tne system there is
~cooPeratiowlh ich emeraeSspontaneously when the states of all
Another equally important effect of cognltivism is the way it has participating neurons reach a mutually satlsfactory state without the
shaped current views about the brain Over the years almost all of need of a central processing unit to guide the entire operation9
neurobiology (and its huge body of empirical evidence) has become
permeated with the information-processing perspective More often than a-+----%3*-
not, origins and assumptions of this perspective are not even
questioned8 Furthermore. there is also increasing evidence that self-
organization is at the core of the brains operation This is hardly
An outstanding example are the two decades of studies on the surprising if one looks at the details of the brains anatomy Although
visual cortex, an area ot the brain where one can easily detect electrical neurons do have distinct responses to specific features of the Visual
responses from neurons when the anlmal is presented with a visual stimuli as mentioned above, this is valid in an anesthetized animal with
image It was reported early on that it was possible to classify cortical an artificially slmplified (internal and external) environment When
neurons like feature detectors, responding to certain attributes of the more normal sensory surroundings are allowed, and the animal Is studied
object being presented its orientation, contrast, velocity, color, and so awake and behaving, it has been shown that their response
on In line with the cognitivist hypothesis, these results are normally characteristics directly depend on neurons localized far from those
seen as giving biological substance to the notion that the braln picks up neurons in their inmediate surroundings, and the stereotyped neuronal
visual information from the retina through the feature speclfic neurons responses previouslj described become highly context sensltiveio
in the cortex, and the information is then passed on to later stages in Neurons must be studied as members of large ensembles which are
the brain for further processing (conceptual categorization, memory constantly disappearing and arising through their cooperative
associations, and eventually action) interactlons To describe the brain as a computer, with a directed flow
of sequential information carried by individual neuronal activity seems
aw.rrtLkrp+w
inadequate today
STC as cognltivism is clearly a constltuted and well-deflned
h - v
research programme, complete with prestigous institutions, journals,
applied technology and international commercial concerns By and large The brain is once more the main source of metaphors and ideas for
most of the people who work within Al (and IT) would subscribe -- other fields of STC in thls alternative orientation Simply put, Instead of
knowningly or unknowlngly-- to cognltivism After all, if ones bread and starting from abstract symbolic descriptions, one starts wlth a whole
butter consists in writing LISP codes or findlng neurons for well- army of simple Stupid components, which, appropriately connected, can
defined tasks, how could it be otherwise? In thls paper we wish call have interesting ghhl properties These global propertles are the ones
attention to the depth of t h i s s o c l a l f r o ma large sector of that embody the cognitive capacities The entire approach depends then
the research community of STC Our orientation is to examlne the on introducing the appropriate connections, this is usually done through
foundations of cognitivist STC, so that the basis for dissent are also a rule for gradual of connections starting from a fairly arbitrary
clarified Essential dissent from the established views on STC today initial state Although several such rules are available today, by far the
takes two basic forms most explored is Hebb s Rule In 1949 D Hebb suggested that changes in
connectivity in the brain could arise from the degree of coordinated
A critique of symbollc computations as the appropriate carrier activity between neurons if two neurons tend to be active together,
for representatlons, their connection is strengthened, otherwise it is dimlnished Therefore
A crltlque of adequacy of the notlon of representations as the the system 5 connectivity becomes inseparable vom its historv 0 f
building block for STC transrormatlon. and related to the klnd of task deflned for the system
Since the real action happens at the level of the connections, the name
These two lines of dissent are examined in the next two Sectlons (neo)connectlonlsmh as been proposed for this direction of research11
Connectionist models provide, with amazing grace, a working
model for a number of Interestingcoanltlvecapacities,su ch as rapid
3. STAGE THREE: THE SELF-ORGANIZATION ALTERNATIVE. recognition, assoclative memory and categorical generalizatiori For
instance, take a total number of N simple McCulloch-Pitts type neurons,
*.urotiwLbuwm- connect them reciprocally, and provided them with a Hebb-type rule
Next present this system with a succession of patterns at some of its
nodes When the system is presented again with one of these patterns,
The motivatlon to take a second look into self-organization was it recoanim it, in the sense that it rapidly adopts a unique global
based on to two widely acknowledged deficiencies of cognitivism The coherent state (an attractor) The recognition is excellent provided that
first IS that symbolic information processing is based onseauential the number of patterns presented is not larger than about 0 IS N
rules, applied one at the time This von Neumann bottleneck is a Furthermore, the system performs a correct recognition even if the
dramatic limitation when the task at hand requires large numbers of
sequential operations (such as image analysis or weather forecasting) A
9for e~lensireaix~~iono~lniSpoinlO'vDi.~rnsoe.cen~e 'and.PD~~N\EIE;
Aufofpmiwm Delam,w,e~oo/iiwe!m a. 9,il Paris 1983
5 RShepardandJtletzler, S2im171 701-3 1971 10 For relerenm on tnis $eF Vwela, .ivinp w.m 01 mnm rnaxinp A rnia3e *w aDprosn
6SKasslyn, PSYC/M/Rw88 46-66,1981 lone~rmiewca, in P .ivingstone (Ed ), c+-&&W#iw&P, rocof 1neSlanfora
SEeBBnBrSmS212535-81, 1979 nternationalS ympmim.AnrnaLibrl. 1984. DD 208-224
8 This Is thecpening iineofa popular textbmk in neurmience 'The brain ism unresting 11 Felman and D Ballaro. Canneclionisl rnmels md lhair properties', aP?,liva.%m
asssmblyof calls that mntinually rgaives inforrnatlon,e laboratesandp ercaives It, and 6 205-254. 1982 Forex~ensivedzc~srmrmcurrentrnmelsinlhisairktionssas
makesdeclsions'.SKuffler andJ Nichols, Frmh&wrmtOBr8ln, Sineuer Associates, Bmton, McClellana an0 D Rmelharl (Em ), P~al4IDislr~Pu(arPSrl~LU1I6~sm I&
1976,p3 Hims/rx/umo/@mi/m 2 vols Mil Press. I986
Trends in Cognitike Science and Technolop 3
pattern is presented partially, or the system is partially mutilated scale technological investments, and a smaller, young group of
The exploslve interest over the past 3 yearsI3 in this type 01 researchers and daring commercial ventures, moving in a different
model is ]ustifled on several counts First, cognitivist AI and direction
neuroscience had few convincing results on the kind of cognitive
performance just described Second, these models are immediately But, for two good reasons this would be an adequeate description
comparable to biological systems, since both are intrinsically parallel of STC only on a superficial level First, because most researchers
(each neuron works according to its own local context with no overseer) within this alternative orientation do not find it impossible to look for a
and highly distributed (the global attractor is not due to one localized synthesis bet ween cogni t ivism rn connect ioni sm they might be
element but to the contribution of many, and is thus resilient to the loss complementary top-down and bottom-up approaches respectively The
of a few) In particular, this means that one can work with a degree of conceptual status of such synthesis is unclear and so far there are no
integration between Al and neuroscience hitherto unthinkable Finally concrete examples An alliance under some form of compromise between
the models are general enough, since expressed in rriathematical terms a less orthodox cognitivism, relaxed to include low-level parallel
to be applied, with little modlfication, to problems as varied as parallel dlstributed processing provided by self-organizational approaches is a
vision or speech recognition likely possibility, especially in engineering-oriented A1 given its
opportunistic mood This potential complementati on w I I I undoubtedly
produce vislble results, and might well become the dominant trend for
many years in STC
A second reason why such a view of STC would be superficial is
This alternative orlentation--connectionist, selt-organization,
because in both orientations (and hence in a future synthesis1 some
assoclationist network dynamical-- is young and diverse Most nf those
essential dimensions of cognition would be S~.Umlssing
who would enlist themselves as members hold widely divergent views
on what STC 1s and on its future Keeping this disclaimer in mind here
are the alternative answers to the previous questions -culunMb-apnao
Question *l The central dissatlsfaction of what we here call t h e w
What is cognition? alternative is simply the complete absence of common sense in the
Answer definition of cognition so far Both in cognitivism and in present day
The emergence of global states Attractors in a self-orqanizinq system connectionism it is still the case that the criteria for coqnition is a
succesful representation of an external world, which is pre-qiven,
Question *2 usually as a problem solving situation However, our Knowledge activity
How does it work7 in everyday llfe reveals that this view of cognition is too incomplete
Answer precisely the greatest ability of all living cognition is, within broad
Through a device bullt from a large number of neural-like elements limits, to the relevant issues to be addressed at each moment of our
wlth local rules for individual operation, and rules for changes in life They are not pre-given, but m t e d0 r brouaht forth from a
connectivity background, and what counts as relevant is what our common-sense
sanctions as such, always in a contextual way
Questlon *3
How do I know when a cognitive system is functioning adequately? This is a critique of the use of the notion of representation as the
Answer core of STC, since only if there is a prf-glven world can it be
When the global states (attractors1 can be made to correspond to a represented If the world we live in is brought forth rather than pre-
cognitive capaclty, leading to a successful Solutlon to a requlred task given, the notion of representatlon cannot ?ave a central role any longer
The depth of the asssumptlons we are touching upon here should not be
-*- underestimated, since our western tradition as a whole has favored
(with variants of course) the understanding of knowledge as a mirror of
nature Instead one touches here the entlre phenomenon of WrDretatlon
One of the most interestlng aspects of thls alternatlve approach to understood as the activity of enactment or bringing forth, to which we
STC is that symbols play little if any role anymore This entails a are alludtngi4 Since we are concerned with the dominarice of usage,
radical departure from a basic cognitivist assumption the physical instead of with representatlons, it ISa ppropriate to call this alternative
structure of symbols, their form, is forever separated from what they approach to SIC .q&Jy.g!s
stand for, their meaning This separatlon between form and meaning was
the master stroke that created the computational approach, but It also In recent years, however, some researchers W S T Ch ave put
implled a (perhaps fatal) weakness when addressing cognltlve phenomena forth concrete proposals, taklng this critique from the phllosophical
at a deeper level How do symbols aQue their meaning? Whence thls level into the laboratory and re-understanding of Al Thls Is a more
extra activlty which is, by construction, not in the cognltlve system? radical departure from STC than ever before, one that goes beyond the
themes discussed during the formative period At the same time, it
In sltuations where the unlverse of possible Items to be naturally incorporates many of the tools developed withln the
represented is constrained and clear-cut (such as when a computer Is connectionist context, as we shall presently see
programmed, or when an experiment Is conducted wlth a set of
predeflned visual stlmull), the assigment of meaning is clear Each m+-w-n--&93
dlscrete physlcal Item within the cognltlve system Is made to
-to an external item (its referential meaning), a mapping The assumption in STC has all along been that the world can be
operatlon which the observer easily provldes Remove these constraints, divided into regions of discrete elements and tasks to which the
and the form of the symbol is that is left, and meaning becomes a cognltlve system addresses itself, acting within a given space of
ghost, as it would if we were to contemplate the bit patterns In a Problems vision, language, movement Although it is relatlvely easy to
computer whose operating manual was lost define all possible states in the space of the game of chess, it is has
proven quite unproductive to carry this approach over into, say, the
In the connectionist approach, meaning is linked to the overall space of moblle robots Of course, here too one can slngle out dlscrete
performance (say In recognltion or learning) Hence, meanlng relates to items (such as steel frames, wheels and windows in a car assembly)
the global state of the system, and ISn ot located In a particular But it is also clear that while the chess world ends neatly at some polnt,
symbols The form/meanlng dlstlnction at the symbollc level dlsappears, the world of movement amongst objects does not It requires our
and reappears In a different garb the observer provldes the continuous use of common sense to conflgure our world of obJects 16
correspondence between the systems global state and the world It Is
supposed to handle Thls is, then, a radlcally different way of obtalnlng
Just like the meaning of a wora in natural language, tile
(and studying) representatlons As to how this alternatlve approach categorization of our @&,udw orld does not have sharp boundaries In
aladndgrueasgsee si s,o tahte pr,r ehselgnht,e vr,a gcuoeg nlOtlbvvelo cuasplya,c fitoiers t hseu cahd eapst sth tlhnlksin Igs oorn ly sfalocwt, pitr oigsr feasisr, tiot sdaayw tnheadt obny mthaen y1 A9l w7o0rk~aef~tres r t htwato e dveecna dthees osifm hpulmesbtl ingly
further terrain to be conquered, for its critics, a sign of its limitations cognitive action requires a seemingly infinite amount of knowledge,
which we take for granted (in fact it is so O~VIOIJS as to be invisible), but
4. STAGE FOUR: THE ENACTIVE ALTERNATIVE. which must be spoon-fed to the computer The cognitivist hope for a
r-wcur-
i4Mait influential in thlsrespact istheworkofHGGadamer. IruthandHefm Mury
Prsa, 1975 For8ciear lntrcductiontohermeneutlcs~PalmerW, rmmfi&,
It ISte mpting to see today s STC as consisting of two warring Nothwestern Univ Prsa, 1979 The formulation of thls Sectlon ow= a great &Is to the
factlons a mainstream paradigm with academic preferences and large- influence of F.Flores,s ~e'T .WinncqredandF Flores. U n ~ s t ~ i ~ ~ D u f ~ s ~ ~ i f i ~ ~
AnewF~n~ti~for~i~,ANebw lJeerxse,y , 1986
12 J HoDfleld. Pru?Nef/AcalSO (USA), 79 2554-2556, 1962 i5Thenameisfarfrmb einganestabilshedom.W esuwt it heremostlyforpeda)glcal
~3SOmeofthenamesgivento~ntheticp/stemswithalternative~isig tenl ling the rms,u ntfl a better one is prop&
Connection Machine, theButterfly Machine,t he Boltzman MEhine, theCmmicCube, the
cedn Porject, thesis1 Languw 16 Ses P Blere. "The profmWs challenw", A/IY@wM Winter 1895, pp.60-70.
-
EAI B