Collaboration between Human and Artificial Societies: Coordination and Agent-Based Distributed Computing

308 Pages·1999·3.64 MB·English

by  Jean-Pierre Briot

Lecture Notes in Artificial Intelligence 1624
Subseries ofLectureNotesin ComputerScience
Editedby J.G. Carbonelland J. Siekmann
Lecture Notes in Computer Science
Editedby G.Goos,J.Hartmanis,and J.van Leeuwen
(cid:0)Berlin
Heidelberg
NewYork
Barcelona
HongKong
London
Milan
Paris
Singapore
Tokyo
Julian A. Padget (Ed.)
Collaboration
between Human and
Artificial Societies
Coordination and Agent-Based
Distributed Computing
(cid:0) (cid:1)
SeriesEditors
JaimeG.Carbonell,CarnegieMellonUniversity,Pittsburgh,PA,USA
Jo¨rgSiekmann,UniversityofSaarland,Saarbru¨cken,Germany
VolumeEditor
JulianA.Padget
DepartmentofMathematicalSciences
UniversityofBath
ClavertonDown
BathBA27AY,UnitedKingdom
E-mail:[email protected]
Cataloging-in-Publicationdataappliedfor
DieDeutscheBibliothek-CIP-Einheitsaufnahme
Collaborationbetweenhumanandartificialsocieties:coordinationandagent
baseddistributedcomputing/JulianA.Padget(ed.).-Berlin;Heidelberg;
NewYork;Barcelona;HongKong;London;Milan;Paris;Singapore;Tokyo:
Springer,1999
(Lecturenotesincomputerscience;1624:Lecturenotesinartificialintelligence)
ISBN3-540-66930-2
CRSubjectClassification(1998):I.2.11,D.1.3,C.2,D.3
ISBN3-540-66930-2Springer-VerlagBerlinHeidelbergNewYork
Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialis
concerned,specificallytherightsoftranslation,reprinting,re-useofillustrations,recitation,broadcasting,
reproductiononmicrofilmsorinanyotherway,andstorageindatabanks.Duplicationofthispublication
orpartsthereofispermittedonlyundertheprovisionsoftheGermanCopyrightLawofSeptember9,1965,
initscurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer-Verlag.Violationsare
liableforprosecutionundertheGermanCopyrightLaw.
(cid:1)c Springer-VerlagBerlinHeidelberg1999
PrintedinGermany
Typesetting:Camera-readybyauthor
SPIN10703260 06/3142–543210 Printedonacid-freepaper
Preface
The full title of the HCM network project behind this volume is VIM: A
virtual multicomputer for symbolic applications.The three strands which bound
the network together were parallel systems, advanced compilation techniques
andartificialintelligencewithacommonsubstrateintheprogramminglanguage
Lisp.The initialaimof the project was to demonstrate how the combinationof
these three technologies could be used to build a virtual multicomputer — an
ephemeral, persistent machine of available heterogeneous computing resources
— for large scale symbolic applications . The system would support a virtual
processor abstraction to distribute data and tasks across the multicomputer,
theactualphysicalcompositionofwhichmaychangedynamically.Our practical
objectivewastoassistintheprototypingofdynamicdistributedsymbolicappli-
cationsin artificialintelligenceusing whatever resources are available(probably
networked workstations), so that the developed program could also be run on
more exotic hardware without reprogramming.
Whatwehad notforeseen atthe outset ofthe project was howagentswould
unifythe strands at the application level, as distinct from the system level out-
linedabove.Itwasasaresult oftheagentinfluence thatweheldtwoworkshops
in May and December 1997 with the title “Collaboration between human and
artificialsocieties”.
The papers collected in this volumeare a selection from presentations made
atthosetwoworkshops.Ineach casetheformatconsisted ofanumberofinvited
speakers plus presentations from the network partners. The speakers submitted
draftmanuscriptsorabstractswhichweredistributedaspreliminaryproceedings
atthe meetings.The presentations frequently stimulatedplentyofdiscussion —
wreaking havoc with the schedule! Followingthe two meetings, revised versions
of the full papers were refereed by the programme committee and on occasion
by external referees (acknowledged below). These comments were forwarded to
the authors to guide them in improving their papers. It is the final results of
these endeavours that appear here.
April 1999 Julian Padget
VIM Network Coordinator
VI
Organization
Program Committee
Ulises Cort´es Universitat Polit´ecnica de Catalunya
Julian Padget University of Bath
Christian Queinnec Laboratoire d’Informatiquede Paris 6
Hans Voss GMD – German National Research Center for
InformationTechnology
Carles Sierra Institut de Investigacio´ en Intel.lig`encia Artifi-
cial, CSIC
Referees
Russell Bradford John Fitch Josep Puyol
Harry Bretthauer Enric Plaza Vi¸cenc Torra
EmmanuelChailloux
Acknowledgements
WeareparticularlygratefultoUlisesCort´es,MiquelSa`nchez-Marr´e,JavierB´ejar
and other members of the Departament de Llenguatges I Sistemes Informa`tics
(LSI) at Universitat Polit´ecnica de Catalunya (UPC) for the organization of
the first workshop in Lanjaro´n and to Vincenzo Loia and other members of
the Dipartimento di Informatica` ed Applicazioni at Universita` degli Studi di
Salerno and staff of the Centro Universitario Europeo per i Beni Culturali for
the organization of the second workshop at VillaRufoloin Ravello.
Our thanks are due to the program committee and especially the other ref-
erees for helping to bring this volume to fruition. We also wish to record our
appreciation for the support received from Alfred Hofmann of Springer Verlag
in realizing the publicationof this volume.
Finally,we express our thanks to the European Commission for funding the
project through their Human Capital and Mobility programme under contract
CHRX-CT93-0401,1994-97inclusive.
Table of Contents
I Languages and Systems
A Classification of Various Approaches for Object-Based Parallel and
Distributed Programming ...................................................3
Jean-Pierre Briot (Laboratoire d’Informatique de Paris 6, FR)
Rachid Guerraoui (E´cole Polytechnique F´ed´erale de Lausanne, CH)
Towards Meta-Agent Protocols ............................................30
Andreas Kind and Julian Padget (University of Bath, UK)
Examples of Fuzziness in Compilers and Runtime Systems .................43
Angela C. Sodan (GMD FIRST –Berlin, DE)
Towards Rigorous Compiler ImplementationVerification ...................62
Wolfgang Goerigk and Friedemann Simon (Institut fu¨r Informatik und
Praktische Mathematik, Christian-Albrechts-Universita¨tzu Kiel, DE)
Shiftingthe Focus from Control to Communication:the STReams
OBjects Environments Model of CommunicatingAgents ...................74
Stefano A. Cerri (Universita` di Milano, IT)
Direct Manipulation,Scalabilityand the Internet .........................102
Don Cruickshank and Hugh Glaser (University of Southampton, UK)
II Agents and Capabilities
Towards the Abstraction and Generalization of Actor-Based
Architectures in Diagnostic Reasoning ....................................115
Stefano A. Cerri (Universita` di Milano, IT)
Antonio Gisolfi and Vincenzo Loia (Universit`a di Salerno, IT)
Converting Declarative into Procedural (and Vice Versa) ..................132
Iain D. Craig (University of Warwick, UK )
Reflective Reasoning in a Case-Based Reasoning Agent ...................142
Miquel Sa`nchez-Marr`e, Ulises Cort´es and Javier B´ejar (Universitat
Polit`ecnica de Catalunya, ES)
Ignasi R.-Roda and Manel Poch (Universitat de Girona, ES)
ModellingRationalInquiry in Non-ideal Agents ...........................159
Antonio Moreno (Universitat Rovira i Virgili,Tarragona, ES)
Ulises Cort´es and Ton Sales (Universitat Polit`ecnica de
Catalunya, ES)
VIII Table of Contents
On the Process of Making Descriptive Rules ..............................182
D. Rian˜o (Universitat Rovira i Virgili, ES)
III Coordination and Collaboration
A Service-Oriented Negotiation Model between Autonomous Agents ......201
Carles Sierra (Institut de Investigacio´ en Intel.lig`encia Artificial,
CSIC, Bellaterra, ES)
Peyman Faratin and Nick R. Jennings (Queen Mary and Westfield
College, University of London, UK)
Competing Software Agents Support Human Agents ......................220
Sabine Geldof (Vrije Universiteit Brussel, BE)
Walter Van de Velde (Riverland s.a., BE)
CoordinationDeveloped by Learning from Evaluations ....................234
Edwin D. de Jong (Vrije Universiteit Brussel, BE)
Rules of Order for Electronic Group Decision Making – A Formalization
Methodology .............................................................246
Henry Prakken (Utrecht University, NL)
Thomas F. Gordon (GMD – German National Research Center for
Information Technology, DE )
Broadway:A Case-Based System for Cooperative InformationBrowsing
on the World-Wide-Web ..................................................264
Michel Jaczynski and Brigitte Trousse (INRIA Sophia Antipolis, FR)
Towards a Formal Specification of Complex Social Structures in
Multi-agent Systems ......................................................284
Juan A. Rodr´ıguez-Aguilar, Francisco J. Mart´ın, Pere Garcia,
Pablo Noriega and Carles Sierra (Institut de Investigacio´ en
Intel.lig`encia Artificial, CSIC, Bellaterra, ES)
Author Index .............................................................301
Introduction
The objective of this introduction is to outline the origins and development
of the VIM project in order to give a background to the papers which follow.
Then,foreach paper, we giveabriefsynopsis and relate ittothe larger picture.
The network brought together 15 partners researching the areas of parallel
and distributed systems, compilationfor parallelexecution and artificial intelli-
gence. Our originalgoalwasto demonstrate AI applicationsrunning on distrib-
uted architectures and across local and wide-area networks. The VIM partners
were:
– AI Lab Vrije Universiteit Brussel (BE)
– Christian Albrechts Universit¨at zu Kiel (DE)
– GMD FIT.KI St.Augustin (DE)
– GMD FIRST Berlin (DE)
– Universitat Polit´ecnica de Catalunya,Barcelona (ES)
– Institut de Investigacio´ en Intel.lig`encia Artificial (CSIC), Bellaterra (ES)
– Universitat Rovira i Virgili,Tarragona(ES)
– ILOG s.a.,Paris (FR)
– INRIA, Rocquencourt (FR)
– University of Bath (GB)
– University of Southampton (GB)
– University of Warwick (GB)
– CNR Napoli(IT)
– Universita` di Pisa (IT)
– Universita` di Salerno (IT)
What the network had not foreseen was how the originalgoal of distributed
AIapplicationswas goingto be satisfied inquite different ways fromthose orig-
inallyenvisaged, through the topic of agents. This emerged as a strong collabo-
ratorytheme inthe second halfofthe project, largelyinspired by the modelling
ofthe Spanishfishmarket(lalonja)asamulti-agentsystem.Thefirstversionof
thiswasdevelopedbetween CNRNaplesandIIIABellaterrain1995andseveral
complete rewrites followed, as understanding of the architecture of agents and
the architecture of platforms for electronic commerce increased.
In Part I, we have grouped together the papers on languages and systems,
laying out some of the technology that was researched and developed in the
framework of the project in order to support multi-agent systems. In Part II,
we present material focussing on the internal architecture of agents and how
various classical AI techniques are being adapted for an agent context. Finally,
in Part III, we explore architectures within which agents interact, looking at
both general problems, such as negotiation, coordination and decision-making
and application-derived problems in informationfinding through the web.
X Introduction
I Languages and Systems
Briot and Guerraoui: A Classification of Various Approaches for Object-Based
Parallel Distributed Programming Jean-Pierre Briot gave this invited presenta-
tionatthefirstworkshop,surveyingthevastrangeofobject-orientedlanguages,
focussing in particular on those aiming to support parallel and distributed pro-
gramming.The authors identify three novel categories by which to classify sys-
tems: applicative, where distribution and concurrency are provided through li-
braries (see for example work done in EuLisp as part of this project [10]), inte-
grative,where these facilitiesare mergedwithobject behaviour(see forexample
the Meroon system [11],also related to the VIM project and other work in Eu-
Lisp [9]) and reflective, where they are intimately bound up with the protocols
of the language (see also the next paper [5]).
Kind and Padget: Towards Meta-Agent Protocols Since the first papers on
ObjVlisp and later the Common Lisp Object System, those involved in object-
oriented language design have looked to meta-object protocols to provide an
extra dimensionofflexibilityinapplicationsdevelopment. The ideaput forward
in this paper is an exploration of a similar notion for agent-oriented program-
ming. One of the most difficult requirements to satisfy in agent construction
is the need to adapt: not only is change intrinsic to the kinds of problems to
which agent-oriented approaches are applied, it is also a necessary consequence
of agent mobility [2]. Although change can be avoided or at least minimized,
it is a fundamental property of the potentially most fruitful agent application
areas.Theintentionset forthinthis paperistoseparate outtheconcerns ofthe
application from the meta-level while maintaining a unified framework in the
reflective style identified in the preceding paper.
Sodan: Examples of Fuzziness in Compilers and Runtime Systems Althoughthe
title refers to compilers and runtime systems, the thesis behind the paper is of
the wider applicability of fuzziness to complex problems and, linking it to this
project,howfuzziness couldbeappliedtodistributed systems andhence agents.
Specifically,the approach has been used in the configuration of distributed pro-
grams for multiple heterogeneous target environments, where fuzzification per-
mits the combination of qualitative and quantitative attributes and just plain
incompleteknowledge,tomakeareasonable choice inahigh-dimensiondecision
space.
Goerigk and Simon: Towards Rigorous Compiler Implementation Verification
The matter of security — in its many forms — is a critical factor in the de-
velopment of web-based programs and even more so when the application is
electronic commerce. One aspect of security is to decide whether the results of
a compiler can be trusted and the technologies presented in this paper outline
how a completely verified compiler has been developed for a subset of Common
Lisp. Hence, it provides a blueprint for the development of verified compilers
andafoundationforsecure distributed computingwhere correctness certificates