An Overview of Social Simulation Research in Brazil Jaime Sim˜ ao Sichman Intelligent Techniques Laboratory (LTI) Universidade de S˜ ao Paulo (USP) S˜ ao Paulo, SP, Brazil jaime.sichman@poli.usp.br Antˆ onio Carlos da Rocha Costa, Diana Adamatti, Grac ¸aliz Pereira Dimuro Laboratory of Environmental and Social Modelling and Simulation (LAMSA) Universidade Federal do Rio Grande (FURG) Rio Grande, RS, Brazil {ac.rocha.costa, dianaada, gracaliz}@gmail.com Fernando Buarque Computational Intelligence Research Group (CIRG) Universidade de Pernambuco (UPE) Recife, PE, Brazil fbln@ecomp.poli.br Pierre Bommel Centre de coop´ eration internationale en recherche agronomique pour le d´ eveloppement (CIRAD) Montpelier, France bommel@cirad.fr Abstract—Presented as a panel at the Third Brazilian Workshop on Social Simulation (BWSS 2012), held in Curitiba, Brazil, on October 20th. 2012, this paper presents an overview of the current research on social simulation in Brazilian research groups. Keywords-artificial intelligence, multi-agent systems, multi- agent-based simulation, agent based social simulation. I. I NTELLIGENT TECHNIQUES LABORATORY (LTI/USP) Created in 1997, the Intelligent Techniques Laboratory (LTI) is a research group that belongs to Escola Polit´ ecnica (EP), one of the oldest and more prestigious Engineering schools of Brazil. The laboratory is composed of 3 faculty members and more than 20 students, including undergradu- ate, graduate and pos-doc students. LTI’s interest is mainly Artificial Intelligence, and its multi-agent systems (MAS) team is leaded by Prof. Jaime Sim˜ ao Sichman. Its webpage may be found at http://www.lti.pcs.usp.br. With other colleagues, Prof. Sichman was one of the founders of the Multi-Agent-Based Simulation (MABS) domain and the corresponding workshop series [13] [15], that has been occurring as an associated event of ICMAS and AAMAS conference since 1998. According to Castelfranchi [2], research in multiagent systems may have two complementary perspectives: social resolution and social simulation. In the first, the design of certain computer techniques, like coordination and negotia- tion, are the end of the research, aiming to solve a class of problems which involves autonomous and distributed entities. On the second perspective, computer techniques are means to develop and test social theories. However, even if it seems more natural that Multiagent Based Simulation (MABS) should be used in the second perspective, it is also an important step to produce effective distributed problem solving systems. In the LTI, we have been using MABS techniques to tackle these both MAS perspectives. Regarding the social simulation perspective, our group have been investigating several mechanisms of partnership and coalition formation, as well as the effect of social control, like reputation, in this phenomena. A first work in this perspective was the PartNET++ [8]. It consists of an experimental MABS tool that uses a new model based on hyper-graphs for understanding partnership formation among heterogeneous agents. Based on a previous tool, the agents of the system have goals to achieve and actions that lead to these goals, which must be performed by other agents. When choosing preferred partners, the agents may have different strategies (utilitarians, substantialists and misers). In PART-NET, partnerships were restricted to two agents, and the authors have shown that some social hy- potheses were validated, i.e., that in heterogeneous societies, utilitarians should have the best net benefit, followed by substantialists, then my misers. By using the PartNet++ simulator in several new experiments, it was shown that results that are valid for partnerships between two agents can be generalized for multiple agents. On the other hand, we have experienced in the last decades a rapid increase in the number of available online e-services. Agent-based computing has been advocated as a natural computational model to automate the interaction with those services, thus enabling the formation of multiagent systems. In these latter, agents may use trust and reputation as the main control mechanism and they usually exchange such information in order to accelerate reputation evaluation. However, due to the semantic heterogeneity of the different reputation models, agents interaction about reputation has to deal with interoperability issues. We have developed an architecture, called SOARI [10], that enables the semantic 2012 Third Brazilian Workshop on Social Simulation 978-0-7695-4942-2/12 $26.00 © 2012 IEEE DOI 10.1109/BWSS.2012.31 18 2012 Third Brazilian Workshop on Social Simulation 978-0-7695-4942-2/12 $26.00 © 2012 IEEE DOI 10.1109/BWSS.2012.31 18