Generating Multidimensional Social Network to Simulate the Propagation of Information Mathilde Forestier, Jean-Yves Bergier, Youssef Bouanan, Judicaël Ribault, Grégory Zacharewicz, Bruno Vallespir, Colette Faucher To cite this version: Mathilde Forestier, Jean-Yves Bergier, Youssef Bouanan, Judicaël Ribault, Grégory Zacharewicz, et al.. Generating Multidimensional Social Network to Simulate the Propagation of Informa- tion. the 2015 IEEE/ACM International Conference, Aug 2015, Paris, France. pp.1324-1331, 10.1145/2808797.2808870. hal-02536031 HAL Id: hal-02536031 https://hal.archives-ouvertes.fr/hal-02536031 Submitted on 7 Apr 2020 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Generating Multidimensional Social Network to Simulate the Dissemination of Information Mathilde Forestier*, Jean-Yves Bergier**, Youssef Bouanan*, Judicael Ribault*, Gregory zacharewicz*, Bruno Vallespir*, Colette Faucher** *Bordeaux University, IMS Laboratory, UMR 5218, 351, Cours de la liberation, Talence, F-33400, France fi[email protected] **Aix-Marseille University, LSIS Laboratory, UMR 7296, Avenue Escadrille Normandie Niemen 13397 Marseille, France fi[email protected] Abstract—Social simulation implies two preconditions: deter- Modeling population needs the understanding of the spe- mining a population and simulate the information diffusion within cific norms and ways the society organizes itself, e.g., an it. A population represents a group of interconnected individuals European society does not have the same features as an African sharing information. In this paper, the population we generate one. The model of the population has to respect the codes is detailed by socio-cultural features, specifically the way that of the society in order to accurately simulate the information people tend to link together. To this end, the use of a social diffusion. Furthermore, the relationships between people are network is a little bit restrictive: people are linked by only one relationship. Multidimensional Social Networks (MSN) model too complex to be modeled by one link. Social networks are, 3D social networks where each dimension represent a kind of for this purpose, a good but simple way to resume relationships relationship [1]. This architecture allows us to better represent between people. Nevertheless, social networks are generally the diversity of humans relations but also define distinctive based on only one relationship between people, or an ag- rules for the message diffusion simulation. The inner idea is gregation of several relationships into one. Multidimensional that information disseminates differently according to the links social networks (MSN) begin to emerge due to the importance through which the information propagates. So, we present in this of each relationship in the communication process [4], [5], paper the modeling of our MSN based on social science and a [6]. Flatten an MSN into a one dimension social network simulation using propagation rules set for each dimension. does not allow to consider each relation as a unique way to communicate with its own communication rules. It also I. INTRODUCTION does not allow to represent the complexity of an individual Nowadays, many researchers are interested in developing social life. In this paper, we model an MSN with the idea new and more efficient systems for social simulation. The that information disseminates differently according to the link issues explored include psychology, organizational behavior, through which the information propagates: people do not sociology, political science, economics, anthropology, geogra- receive and transmit in the same way an information according phy, engineering, archaeology and linguistics [2], [3]. Military to the person who gave them the information. Following this simulation systems have to support deeply detailed analysis of postulate, we present in this paper the general framework of individual’s behavior. The SICOMORES project, in which we our MSN in order to generate a model of a population based on are involved, is a military project whose aim is to model a several relationships. Using social science research we describe population of interconnected individuals in order to simulate some relationships representing a part of the human social life. and observe the reaction of this population face to different information. In the context of modern conflicts, the overall This paper proposes a general framework to model a maneuver is an iterative process to achieve a desired effect population based on its social structure rules and its cultural on the environment. The actions deployed in this context are features. This framework is really adaptable thanks of the divided into influence and combat actions. Combat actions is MSN architecture which allows to separate and distinguish rather limited in the context of stabilization operation, priority the relationships the one from the others, to add or delete is given to actions of influence. These actions of influence a dimension, and to set for each dimension its own features can be defined as all the intentional activities to achieve an and diffusion rules. The use of an MSN also allows to model effect on perceptions in order to change attitudes and / or the message acceptation and the transmission rules for each behaviors. In the context of stabilization phase of a conflict, the dimension. support of the population is at stake and people are the major target of such actions. Psychological Operations (PSYOPS) use media to share information customized to info-target’s The remainder of this paper proceeds as follows. In Section cultural and linguistic specific features. However, the whole II, we introduce the related work. Section III describes our population is not an homogeneous entity easy to convince. methodology: how we create nodes, what are our dimensions, It can be indifferent, opponent or ally. So, the aim of the how we linked people and so on. Then, Section IV-B presents SICOMORES project is to model a population with cultural the measures of our MSN and our proxy/sever architecture of features allowing to describe and simulate the effects of the simulation of message diffusion in an MSN. Finally, the paper operations of influence within it. concludes with directions for future work. II. RELATED WORK measure the strength of these links. As for Berlingerio et al., they analyze hubs in a multidimensional network [4]. Actually, Social network generation, or more specifically graph gen- measures from social network analysis have to be adjusted to eration, appeared in 1960 with the Erdos˝ and Renyi´ [7] From MSN. Finally, Forestier et al. propose a MSN from online this time to nowadays, researchers have looked to improve discussions where relations are from discussion structure and the graph generation to respect social network features [8]: text content. These relations help to find celebrities in the degree distribution, assortativity, clustering coefficient, small discussions [5]. So, MSN model has a really big interest in world structure [9] and group structure. Social networks are catching the social life diversity of different kinds of people. not random graph and have special features to respect in order to generate realistic population. Thus, researchers aim III. GENERAL FRAMEWORK to generate automatically social networks according to these features. Following Erdos˝ and Renyi´ [7], the dyadic models The SICOMORES project aims to simulate the actions of appeared [10], [11], the Markov random graph [12] and the influence happening in the context of stabilization phase of network generation with small world properties [13]. Since modern asymetric conflicts. These actions of influence can be 2000s, models are improved to take into account the time effect Psychological Operations (PSYOPS) or Civil Military Coop- with the notion of preferential attachment: new nodes link eration (CIMIC). Both of these operations aim to convince proportionally to high connected previous node in a richer get people’s heart and mind. These operations aim to make the richer configuration [14]. Newman et al. develop algorithms local population an ally. Military spread messages to influence to create social networks with an arbitrary degree distributions info-targets in their behavior. [8]. These algorithms allow to set a distribution degree in order to obtain nodes with a correct number of links. Nowa- In order to simulate the message diffusion, we define a days, researchers develop methods allowing to manage several multidimensional social network (MSN). Berlingerio et al. de- properties as Badham and Stocker who manage the degree fined in [1] a multidimensional network as a triple G=(V,E,L) distribution, the clustering coefficient and the assortativity [15]. where: V is a set of nodes; L is a set of labels; E is a set of labeled edges, i.e., the set of triples (u,v,d) where u; v 2 V are In a military perspective, Svenmarck et al. simulate the dif- nodes and d 2 L is a label. Also, we use the term dimension fusion of information