BWSS
2008
1st
Brazilian Workshop on Social Simulation
Bahia Othon Palace Hotel, Salvador,
Bahia, Brazil
October 30th, 2008
(co-located with SBIA 2008 -
Brazilian Symposium on Artificial Intelligence)
Social Simulation is the scientific enterprise through which computational
models and tools are employed to computationally study societies and their structures
and processes. Specially Artificial Intelligence and Multiagent Systems models
are relevant in such context.
On the other hand, Social Simulation may operate
as a source of inspiration for theories, models and techniques for AI and MAS,
as it heavily draws on the diverse social sciences (anthropology, sociology,
political science, economy, government, management, etc.).
This one-day workshop, co-located with SBIA 2008,
aims to gather researchers and students in AI, MAS and the various social sciences
where Social Simulation has been put to work, in order to further the
theoretical and practical issues concerning social simulation, and to leverage
the area of Social Simulation in Brazil.
Topics of interest:
-
theoretical foundations of social simulation
-
methods and models for social simulation
-
AI and MAS tools, techniques, models and environments for
social simulation
-
applications of social simulation (in social research,
government, management, education, etc.)
Organizing committee:
Antônio Carlos da Rocha
Costa, UCPel, Brazil
Rafael Heitor Bordini, Durham Univ., UK
Jomi Fred Hübner, FURB, Brazil
Patrícia Restelli Tedesco, UFPE, Brazil
Graçaliz Pereira Dimuro, UCPel, Brazil
Program committee:
Ana Bazzan (UFRGS, Brazil)
Antônio Carlos da Rocha Costa (UCPel, Brazil) - chair
Diana Adamatti (FTEC, Brazil)
Geber Ramalho (UFPE, Brazil)
Graçaliz Pereira Dimuro (UCPel, Brazil)
Gustavo Lugo (UTFPR, Brazil)
Helder Coelho (Univ. Lisbon,
Portugal)
Jaime Sichman (USP, Brazil)
João Balsa (Univ. Lisbon, Portugal)
Julita Vassileva (Univ. Saskatchewan, Canada)
Luis Antunes (Univ. Lisbon, Portugal)
Miguel Lozano Ibañez (Univ. Valencia, Spain)
Nuno David (ISCTE, Portugal)
Patrícia Restelli Tedesco (UFPE, Brazil)
Paulo Blikstein (Nortwestern Univ., USA)
Paulo Trigo (ISEL, Portugal)
Ricardo Rodrigo Stark Bernard (UFSC, Brazil)
Roberto da Silva (UFRGS, Brazil)
Stanley Loh (UCPel, Brazil)
GENERAL TECHNICAL PROGRAM
10:10 - Coffe break
10:30 - Technical Session 1
12:00 - Project Presentation: Cameron Griffith (Indiana Univ., USA)
12:30 - Lunch
14:00 - Technical Session 2
16:00 - Coffe Break
16:20 - Technical Session 3
18:20 - Closing Session
Paulo
Trigo, Instituto Superior de Engenharia de Lisboa, Portugal
Helder
Coelho, Universidade de Lisboa, Portugal.
Abstract. This paper proposes a multi-agent based
simulation (MABS) framework to construct an artificial electric power market
populated with learning agents. The proposed framework facilitates the
integration of two MABS constructs: i) the design of the environmental physical
market properties, and ii) the simulation models of the decision-making and
reactive agents. The framework is materialized in an experimental setup
involving distinct power generator companies which operate in the market and
search for the trading strategies that best exploit their generating units
resources. The experimental results show a coherent market behavior that
emerges from the overall simulated
environment.
João Balsa,
Fac. Ciências da Universidade de Lisboa, Portugal
Luis
Antunes, Faculdade de Ciencias Universidade de Lisboa, Portugal
Luís Moniz,
Faculdade de Ciências da Universidade de Lisboa, Portugal
Helder
Coelho, Universidade de Lisboa, Portugal.
Abstract. We propose the use of a multiagent-based
social simulation approach to help politicians in their task of choosing the
best policy/decision when facing complex scenarios. We show how we can use MABS
tools an techniques to experiment/rehearse with different scenarios and how
this methodology can help decision makers. We use these ideas to develop an
application to the problem of GHGs (Greenhouse Gases) emission reduction. We
explain why this is a critical problem and stress the importance of our
approach to deal with it. We focus here in one of the most relevant instruments
that has been used so far -- a market based approach, more specifically the
``carbon market.'''' We illustrate how this type of instruments can be
rehearsed, by means of a simulation implemented in the NetLogo framework.
Inacio Lanari Bo, Universidade de São Paulo,
Brazil
Jaime Sichman, USP - Universidade de São Paulo,
Brazil.
Abstract. This work
presents the results of experiments made with a spatial evolutionary model of
agents playing the n-Players Prisoner''s Dilemma, using two different ways to
represent the agent''s strategies: finite automata and adaptive automata. Since
adaptive automata can represent complex strategies that cannot be represented
by finite automata, comparative analysis of the co-evolution of strategies
using both representations may lead to a better understanding of the role of
complex strategies in evolutionary games. Here are presented the differences
observed on the total utility obtained by the agents, the speed in which they
converge to a nearly-stationary state, and the characteristics of the
prevailing strategies.
Graçaliz Dimuro, Universidade
Católica de Pelotas, Brazil
Antônio Carlos da Rocha Costa, UCPEL, Brazil.
Abstract. In this paper we
review two approaches to the regulation of agent interactions based on Piagets
theory of social exchanges. These approaches model a social equilibrium
supervisor, that, at each time, recommends certain exchange actions to the
agents, in order to lead the interaction towards the equilibrium, regarding the
balance of the exchange values involved in the exchanges. One approach uses a
centralized supervisor, that has access
to all agents internal state, and give recommendations to lead the agents to an
equilibrium in their exchanges. This centralized supervisor uses a Qualitative
Interval Markov Decision Process (QI-MDP), to determine the best recommendation
for the agents. The other approach is a decentralized one, in which each agent
has an equilibrium supervisor internalized in it. In this model, each
supervisor in each agent has access to the agents internal state where he is
in, but is unable to access the internal states of the other agents. In order
to give exchange recommendations to the supervised agent, the internalized
supervisor uses BDI (Beliefs, Desires, Intentions) plans derived from the
optimal interaction policy provided by a Partially Observable Markov Decision
Process (POMDP). We present a qualitative comparison between the two
approaches, aiming at the identification of which features of each approach can
be used to improve the other one.
Pedro Andrade, INPE, Brazil Antonio Monteiro,
Instituto Nacional de
Pesquisas Espaciais, Brazil
Gilberto Câmara, Instituto Naiconal de
Pesquisas Espaciais, Brazil.
Abstract. This work
presents a contribution towards generalizing the representation of geospatial
entities and their relations for simulating complex spatial systems using the
agent-based approach. We analyse the works in the literature, and argue that
each of the four types of relation is necessary. These relations can be grouped
in two classes, placements and neighbourhoods, with likenesses and differences
between them. Given that, we define requirements for representing geospatial
entities and their relations, and then study six toolkits for ABM (Netlogo,
OBEUS, Repast, Swarm, GRSP, and TerraME), analysing their capabilities to
address the proposed requirements. Finally, we present our current work and
future directions on the development of the TerraME toolkit.
Antônio Carlos da Rocha Costa, UCPEL, Brazil
Fernanda Jeannes
Ulisses Cava, UCPel, Brazil.
Abstract. This paper
presents some issues concerning the idea of taking equation-based simulation
models (EBM) as formal, mathematical specifications of agent-based simulation
models (ABM) for Social Simulation. The paper tries to identify some of
benefits that such point of view could bring to the development of ABM, as well
as some of the preliminary problems that an ABM developer has to face, in order
to guarantee the credibility of his model. The paper first identifies the
benefit of the support for the formal verification of the correction of the
ABM, given the correction of the EBM. Then, it identifies a preliminary set of
crucial implementation details that are not covered by EBM specifications, and
that the simulation developer has to define to make the ABM operational, under
the constraint of keeping the ABM model compatible with the EBM. In particular,
the two issues of the local vs. global scope of the agents' perceptions and of
the degree of personalization of the agents are identified as crucial ones.
Some preliminary concepts are proposed to help to solve those two problems.
João Ferreira, ISEL, Portugal
Paulo Trigo, Instituto Superior de Engenharia
de Lisboa, Portugal.
Helder Coelho, Universidade de Lisboa,
Portugal.
Paulo Urbano, Faculdade de Ciências da
Universidade de Lisboa, Portugal
João Balsa, Fac. Ciências da Universidade de
Lisboa, Portugal
Luis Antunes, Faculdade de Ciencias
Universidade de Lisboa, Portugal
Luís Moniz, Faculdade de Ciências da
Universidade de Lisboa, Portugal.
Abstract. Interactions
between agents in multi-agent systems have an inherent potential for conflict
and must be coordinated. The adhesion to uniform behaviour is one way of coordinating
actions: social conventions and lexicons are good examples of coordinating
systems, where uniformity promotes shared expectations of behaviour and shared
meanings. This paper deals with the emergence of a uniform collective choice, a
consensus, inside a population of artificial agents. The efficiency of the
formation of a uniform shared decision or choice is an important issue. The
nature of interactions and also the nature of society configurations may
promote or inhibit consensual emergence. We study the efficiency of uniform
decision formation along two dimensions: rules of interaction and network
topologies. We compare two different interaction behaviours: the well know and
used Highest Cumulative Reward, which is a kind of reinforcement learning
behaviour, against a recent behaviour named Recruitment based on Force. We also
compare those interaction behaviours along five types of social link networks:
fully connected, regular, random, scale-free and small-world. Simulating
Argumentation about Exchange Values in
Márcia Franco, Universidade Federal do Rio
Grande do Sul, Brazil
Antônio Carlos da Rocha Costa, UCPEL, Brazil
Helder Coelho, Universidade de Lisboa,
Portugal.
Abstract. This article presents
two dialogue protocols to support argumentation about the exchange values
involved in multi-agents interactions. In this work, we adopt the Piaget''s
theory of social exchanges, where as interaction is an exchange of services
between agents, such that they agents assign subjective, qualitative values to
the actions and objects that they exchange during the interaction. The
protocols enable the agents to argue with each other about the exchange values.
The agents present characteristics related to the social power and personality
traits. We show how these characteristics are considered in the reasoning of
the agents, and how it can influence the actions of the agents during the
interactions.
Marcelo Pita
Fernando Buarque de Lima Neto, University of
Pernambuco, Brazil.
Abstract. This paper
simulates the dynamics of a virtual disease disseminated on a population made
of intelligent agents (called Jivas) in the Vidya multi-agent system platform.
Vidya was originally proposed as a strategy and god game in which player is
responsible for supplying intuitions to a Jivas clan, aiming at helping it to
survive in a complex and adaptive environment. This environment is inhabited
not only by Jivas, but also by many other types of artificial lives that
compete by natural resources. This combination of characteristics motivated us
to use the Vidya game as a social simulation and multi-agent systems platform,
where fairly complex social phenomena and strategies can be investigated. In a
previous work the Vidya platform has been used to simulate egoism and altruism,
revealing interesting coherent dynamics for artificially motivated social
behaviors. In this paper we have incorporated in the Jivas intelligent decision
module
one additional decision variable,
which is the presence (or lack) of disease. Results obtained in simulations
have shown emergent change in social behaviors of Jivas social aggregation,
mostly related to the disease contamination and dissemination. We argue that
this platform may be considered in the future as a base for plausible
agent-based social simulations that might be used for supporting public
policies in heath care and other relevant social areas.