Computational interaction frames [Elektronische Ressource] / Michael Rovatsos

Sujets

• Informatik
• Soziologie

Informations

• Publié par: technische_universitat_munchen
• Publié le: 01 janvier 2004
• Nombre de lectures: 31
• Langue: Deutsch
• Poids de l'ouvrage: 2 Mo

Extrait

Computational Interaction Frames
Michael RovatsosInstitut fur¤ Informatik
der Technischen Universitat¤ Munchen¤
Computational Interaction Frames
Michael Rovatsos
Vollstandiger¤ Abdruck der von der Fakultat¤ fur¤ Informatik der Technischen
Universitat¤ Munchen¤ zur Erlangung des akademischen Grades eines
Doktors der Naturwissenschaften (Dr. rer. nat.)
genehmigten Dissertation.
Vorsitzender: Univ.›Prof. Dr. Alois Knoll
Pruf¤ er der Dissertation:
1. Univ.›Prof. Dr. Dr. h.c. mult. Wilfried Brauer
2. Univ.›Prof. Dr. Hans›Dieter Burkhard
(Humboldt›Universitat¤ zu Berlin)
Die Dissertation wurde am 24. Juni 2004 bei der Technischen Univer›
sitat¤ Munchen¤ eingereicht und durch die Fakultat¤ fur¤ Informatik am
14. Dezember 2004 angenommen.Zusammenfassung
In dieser Arbeit werden Interaktionsrahmen als neuartiger Ansatz zur sozialen Inferenz
und zum Umgang mit sozialer Interaktion in Multi-Agenten-Systemen vorgestellt. Das
soziologisch inspirierte Konzept der Interaktionsrahmen kann verwendet werden, um
Kategorien von Interaktionsmustern zu reprasentier¤ en. In Multi-Agenten-Systemen
¤operierende und interagierende Agenten konnen mit Hilfe dieser Rahmen ihre Inter-
aktionserfahrungen aufzeichnen und darauf basierend strategische Kommunikation-
sentscheidungen treffen.
Solch ,,sozial intelligente” Rahmung, die Rationalitat¤ im entscheidungstheoretischen
Sinn mit empirischen Methoden zum Erlernen von Gesetzma igkeiten¤ bezuglich¤ der
Kommunikation in einem Multi-Agenten-System kombiniert, ist besonders fur¤ offene
Agenten-Gesellschaften geeignet. In solchen Gesellschaften kann die Einhaltung einer
a priori festgelegten Semantik von Nachrichten und Kommunikations-Protokollen nicht
vorausgesetzt werden. In Ermangelung absoluter Sicherheit in Bezug auf das Verhalten
¤Anderer in zukunftigen Interaktionen, versucht der rahmenbasierte Ansatz, Unsicherheit
durch Anspassung an beobachtetes Verhalten zu minimieren. Dies stellt eine erhebliche
¤Verbesserung gegenuber der strikten Einhaltung vorde nier ter Protokolle und sogenan-
nter ,,Conversation Policies” dar, die – zumindest wenn man Agenten-Autonomie ernst
nimmt – eine zu gro e Einschrankung¤ darstellen kann.
Es wird zunachst¤ die abstrakte Architektur InFFrA vorgestellt, die auf den Konzepten
von Interaktionsrahmen und Rahmung beruht und als Meta-Modell fur¤ konkretere
Agenten-Designs auf Basis dieser Konzepte verwendet werden kann.
Diese abstrakte Architektur wird durch das formale Modell einer konkreten, direkt im-
¤plementierbaren Instanz von InFFrA erganzt, die den Anforderungen des Meta-Modells
genugt.¤ Fur¤ dieses formale Modell kann auf der Grundlage des allgemeineren Konzeptes
¤der empirischen Semantik von Agenten-Kommunikation eine formale Semantik fur rah-
menbasierte Kommunikation abgeleitet werden. Fur¤ diese formalisierte InFFrA-Variante
werden daruber¤ hinaus Entscheidungsmechanismen de nier t und Lernalgorithmen in
Anlehnung an Theorien des hierarchischen ,,Reinforcement -Lernens entwickelt.
Eine konkrete Implementierung der formalisierten Architektur wird verwendet, um die
Leistungsfahigkeit¤ rahmenbasierter Agenten in einem realistischen Anwendungsszenario
aus dem Bereich der agentenbasierten Webseiten-Verlinkung zu evaluieren. Die experi-
¤mentellen Befunde stellen die Leistungsfahigkeit des Ansatzes unter Beweis und zeigen,
dass Interaktionsrahmen ein machtiges¤ Werkzeug zum Umgang mit sozialer Interaktion in
offenen Systemen darstellen.
Die breite Anwendbarkeit von Interaktionsrahmen wird schlie lich durch die Darstel-
¤lung weiterer Anwendungsmoglichkeiten aufgezeigt.vi ZusammenfassungAbstract
This thesis introduces interaction frames as a novel approach for social reasoning and in-
teraction management in multiagent systems. Interaction frames are a sociologically in-
spired concept that can be used to represent categories of interaction patterns. Agents
operating and interacting in multiagent systems can employ these frames to record their
interaction experience and to make strategic communication decisions based on this ex-
perience.
This kind of socially intelligent framing that combines decision-theoretic rationality
with empirical methods for learning the regularities of communication processes in a mul-
tiagent system is particularly well-suited for open agent societies. In these societies, the
adherence to an a priori semantics of messages and protocols cannot be
taken for granted. In the absence of absolute certainty about the ways others will behave
in future interactions, the frame-based approach relies on observation and adaptation of
one’s own behaviour to the observed patterns of interaction to reduce uncertainty.
This constitutes a signi cant improvement over the use of pre-speci ed communica-
tion protocols and conversation policies in a hard-wired fashion which can be too limiting,
at least if we take agent autonomy seriously.
We present an abstract social reasoning architecture called InFFrA that is based on the
concepts of interaction frames and framing and that can be used as a meta-model for con-
crete agent designs.
This abstract architecture is supplemented by the formal model of a concrete, ready-
to-implement instance of the meta-model that complies with InFFrA requirements. For
this formal model, we also establish a formal semantics based on a more general model of
empirical semantics for agent communication. Furthermore, we de ne decision-making
procedures for this formal version of InFFrA and develop learning algorithms that borrow
from the theory of hierarchical reinforcement learning.
An implementation of the formal social reasoning architecture is used to evaluate the
performance of frame-based agents in a realistic application scenario taken from the do-
main of agent-based Web linkage. The experimental results prove the effectiveness of our
approach and show that interaction frames can be successfully used as a powerful tool for
reasoning about interaction in open systems.
Finally, the broad applicability of frames is illustrated by a discussion of various further
applications.viii AbstractAcknowledgements
Above all, I wish to thank my supervisor Prof. Wilfried Brauer who has contributed to this
thesis in many ways. First of all, by providing a highly productive and pleasant working en-
vironment with enough room for every idea to grow. His unswerving belief that students
and collaborators should be given as much freedom as possible to pursue their own re-
search directions has been con r med by the high quality of the research done in his group
– I hope that this thesis is yet another piece of work that has bene ted from this liberal
spirit. Also, his advice and his often unconventional outlook on things, stemming from his
tremendous experience in computer science and arti cal intelligence, has been immea-
surably valuable and helpful over the years.
Also, I wish to thank Prof. Hans-Dieter Burkhard for his willingness to act as a second
referee and to devote a lot of time to reading and commenting on the thesis. The feedback
he provided on earlier drafts signi cantly contributed to improving the structure and over-
all presentation. As it is not an easy task to deal with material based on the research one
was not directly involved in, I appreciate his effort very much.
A number of people have actively contributed to the research described in this thesis
as close collaborators. Gerhard Wei , leader of the AI/Cognition group has contributed
through his wide and profound knowledge of the eld of multiagent systems and also by
being open to new ideas (at peak times, we would come up with the idea for a new eld
of research every other week). Matthias Nickles, my co-investigator in our Socionics re-
search project had a similar in uence , albeit by expressing views that were often very dif-
ferent from mainstream multiagent systems research. His often radical ideas lead to end-
less but very fruitful discussions that helped shape many novel concepts and perspectives
on seemingly clear and well-understood research issues.
A major part of the research presented here was carried out in the project “ Con ict res-
olution and structural change: social theory as a ‘construction manual’ for adaptive mul-
tiagent systems” which made up part of the Socionics priority program. I am grateful to
the German National Research Foundation DFG for funding this program and our project
(and for paying my salary, of course). In the context of this project, our sociological collab-
orators from the Technical University of Hamburg have introduced us to social theory, but
also to new ways of thinking computer scientists are rarely confronted with. Kai Paetow
and Marco Schmitt deserve my gratitude for introducing me to the works of G. H. Mead
and E. Goffman, but also for backing our interdisciplinary work with in-depth sociological
expertise. And for being very patient with answering each and every question of mine.
Several students have directly contributed to the work presented here. Marco Wolf re-
alised the rst real system based on the principles of interaction frames and framing. Felix
Fischer assisted in the development of the formal semantics of the InFFrA architecture.
He also