Network science has provided useful answers to research questions in many fields, from biology to social science, from ecology to urban science. The first analyses of networked systems focused on binary networks, where only the topology of the connections were considered. Soon network scientists started considering weighted networks, to represent interactions with different strength, cost, or distance in space and time. Also, connections are not fixed but change over time. This is why in more recent years, a lot of attention has been devoted to temporal or time-varying networks. We now entered the era of multi-layer networks, or multiplex networks, relational systems whose units are connected by different relationships, with links of distinct types embedded in different layers. Multiplexity has been observed in many contexts, from social network analysis to economics, medicine and ecology. The new challenge consists in applying the new tools of multiplex theory to unveil the richness associated to this novel level of complexity. How do agents organise their interactions across layers? How does this affect the dynamics of the system? In the first part of the thesis, we provide a mathematical framework to deal with multiplex networks. We suggest metrics to unveil multiplexity from basic node, layer and edge properties to more complicated structure at the micro- and meso-scale, such as motifs, communities and cores. Measures are validated through the analysis of real-world systems such as social and collaboration networks, transportation systems and the human brain. In the second part of the thesis we focus on dynamical processes taking place on top of multiplex networks, namely biased random walks, opinion dynamics, cultural dynamics and evolutionary game theory. All these examples show how multiplexity is crucial to determine the emergence of unexpected and instrinsically multiplex collective behavior, opening novel perspectives for the field of non-linear dynamics on networks.

Using an adaptation of O’Connor and Shumate’s (2018) theoretical propositions, this research examines interorganizational communication through the lens of multidimensional networks. Twitter data was crawled from a selection of emergency management organization accounts to measure affinity, representational, flow, and semantic networks. These data included the organizations’ followed accounts, retweets, replies, and mentions. A thematic analysis of the organizations’ mission statements was also conducted in order to inform the examination of the semantic networks. The results show a significant relationship between the number of accounts an organization follows and the likelihood of having its message shared. This research provides a further theoretical application of a network analysis method of studying interorganizational communication as well as a practical application for organizations seeking to increase their engagement on Twitter.

This study discovers sex differences in multiplex links on formal and informal networks of Swedish local elite. Elites are widely known to have an immense influence on a country’s politics and governance, and proportional representation of women in elite positions is an indicator of democratization and gender equality. Sweden has long been known for democratic and gender equal regulations, and women occupy more elite positions relative to other countries, yet they are still heavily underrepresented in the elite. Previous research on Swedish local elite revealed that women in the elite do not differ from their male peers in terms of local network properties on formal and informal networks; however, the circumstances on the multiplex links are unknown. In this study, multiplexity approach is adopted as it is known for allowing to capture social processes in social network analysis, which could otherwise be overlooked. The formal and informal networks of the community elite from four mid-sized municipalities in Västra Götaland region in Sweden are transformed into multiplex networks, and they were analyzed for local network configurations by using exponential random graph model (ERGM) estimation method. The findings showed that women in the community elite tend to have more multiplex relationships than men; however, they significantly lack valuable brokerage positions on the multiplex level compared to men. Male closure on the multiplex level was found to be higher than females at a partially significant rate, and gender-based homophily on multiplex networks was not found to be statistically significant. One implication of the study is women’s position and integration in the community elite do not appear identical to men’s, and women’s access to social capital in the elite networks is more constrained than it was presumed previously. Another implication is that special attention should be paid to multiplexity in social network analysis research, as it is a valuable tool for improving the apprehension of social mechanisms.

Il est bien établi dans la littérature en criminologie que les pairs délinquants peuvent avoir un effet néfaste sur les comportements d’un jeune. L’analyse des caractéristiques de l’entourage social est donc essentielle à une compréhension globale des conduites individuelles. Puisqu’il est impossible pour un jeune, délinquant ou non, de se distancer complètement du monde conventionnel (Sykes et Matza, 1957; Warr 2002), il importe de considérer le chevauchement des relations conventionnelles et délinquantes pour saisir l’ampleur du phénomène de l’influence social. De surcroît, le réseau social des jeunes ne se limite pas à leurs amis, les membres de la famille, les collègues de classe et de travail pouvant aussi avoir une influence sur les comportements. La présente étude propose une analyse de l’entourage social de 237 jeunes âgés de 14 à 24 ans, fréquentant les organismes communautaires au Québec. Les résultats révèlent que: 1) la participation à un délit chez les jeunes en communauté est fréquente, 2) les caractéristiques du réseau social, reflétant l’enchâssement social, ne se trouvent pas révélatrices de la participation à un délit, 3) côtoyer les membres de son réseau social en grande intensité réduit de manière significative le volume de délits de marché commis dans une année, et ce, même en contrôlant la présence de contacts délinquants dans le réseau, 4) la présence de contacts délinquants dans plus d’une sphère relationnelle composant le réseau social permet de créer un index de dispersion de la délinquance reflétant ainsi l’enchâssement criminel des jeunes et finalement 5) plus les contacts délinquants sont dispersés à travers les sphères relationnelles, plus le risque de participation à un délit augmente. Toutefois, la dispersion des contacts délinquants dans le réseau social ne prédit pas la fréquence des délits commis. À des fins d’intervention, connaître la dispersion de la délinquance dans le réseau social peut aider à la prévention des comportements délinquants. / Abstract: It is well established in the criminology literature that delinquent peers can have a detrimental effect on young people’s behaviors. The analysis of social environmental characteristics is therefore essential to fully comprehend the unraveling of individual actions. Since it is impossible for a youth, whether he offends or not, to completely isolate himself from the conventional world (Sykes et Matza, 1957, Warr 2002), the overlapping of conventional and delinquent relationships needs to be considered to grasp the extent of the social influence phenomenon. In addition, youth’s social networks are not limited to friends, family members, classmates and co-workers all have a potential influence on their behavior. This study conducted an analysis of the social environment of 237 youth aged 14-24, attending community organizations in Quebec. The results reveal that: 1) crime participation in a youth community sample is frequent, 2) social network’s relational characteristics reflecting social embeddedness are not indicative of crime participation 3) high intensity interactions with social network members significantly reduces market crime commission rates, even when controlling for the presence of offenders in the network, 4) presence of delinquent contacts in multiple social network’s relational spheres allows to create a delinquent dispersion index in order to reflect youth’s criminal embeddedness and finally 5) the more offenders are scattered across different relational spheres that make up the youth’ social network, the greater the risk of involvement in crime is. However, this measure of network delinquency does not predict offending rates. For intervention purposes, looking at network’s delinquent dispersion index might improve prevention of delinquent behaviors.

No description available.