Search results

Influence study in combinatorial and collective decision-making, is an interdisciplinary research area combining computer science and social science, especially artificial intelligence and collective decision-making. Influence has long been studied, for instance in political science, but in the context of combinatorial and collective decision-making, this calls for a study of how influences works among multi-agents and multi-issues, how influences and decision-making are interleaved, and how the structures of influence among agents and issues produce an effect. In the thesis, we mainly performed three aspects of work:Firstly, build complex models of influence based on preference representation languages and social influence models, proposed a series of new patterns of influence to better describe the complex influences in real-world situation, and discussed a series of theoretical problems of influencing and influenced structure, influence from more than one origins, and influence with abstentions and constraints. We then test the models of influence from an exemplary perspective for interdisciplinary study, from both social science and computer science paradigms, by both qualitative case studies approach and quantitative matching algorithms approach, to provide an evaluation for the models of influence.Finally, we use the models of influence to perform agent-based simulations, by the example UN Security Council voting. We design those experiments from both social and computer science perspectives, implement it in Netlogo, and discuss the interleaved effects between new cases of influence and different SC reform schemes. ; L'objet de cette thèse (bi-disciplinaire, en informatique et sciences politiques) est l'étude de phénomènes d'influence dans des situations de prise de décision collective, impliquant plusieurs sujets interdépendants, ce qui amène à considérer des domaines dits combinatoires. L'importance de ces phénomènes d'influence est observée depuis de nombreuses années dans différents domaines (dont les ...

Influence study in combinatorial and collective decision-making, is an interdisciplinary research area combining computer science and social science, especially artificial intelligence and collective decision-making. Influence has long been studied, for instance in political science, but in the context of combinatorial and collective decision-making, this calls for a study of how influences works among multi-agents and multi-issues, how influences and decision-making are interleaved, and how the structures of influence among agents and issues produce an effect. In the thesis, we mainly performed three aspects of work:Firstly, build complex models of influence based on preference representation languages and social influence models, proposed a series of new patterns of influence to better describe the complex influences in real-world situation, and discussed a series of theoretical problems of influencing and influenced structure, influence from more than one origins, and influence with abstentions and constraints. We then test the models of influence from an exemplary perspective for interdisciplinary study, from both social science and computer science paradigms, by both qualitative case studies approach and quantitative matching algorithms approach, to provide an evaluation for the models of influence.Finally, we use the models of influence to perform agent-based simulations, by the example UN Security Council voting. We design those experiments from both social and computer science perspectives, implement it in Netlogo, and discuss the interleaved effects between new cases of influence and different SC reform schemes. ; L'objet de cette thèse (bi-disciplinaire, en informatique et sciences politiques) est l'étude de phénomènes d'influence dans des situations de prise de décision collective, impliquant plusieurs sujets interdépendants, ce qui amène à considérer des domaines dits combinatoires. L'importance de ces phénomènes d'influence est observée depuis de nombreuses années dans différents domaines (dont les ...

Influence study in combinatorial and collective decision-making, is an interdisciplinary research area combining computer science and social science, especially artificial intelligence and collective decision-making. Influence has long been studied, for instance in political science, but in the context of combinatorial and collective decision-making, this calls for a study of how influences works among multi-agents and multi-issues, how influences and decision-making are interleaved, and how the structures of influence among agents and issues produce an effect. In the thesis, we mainly performed three aspects of work:Firstly, build complex models of influence based on preference representation languages and social influence models, proposed a series of new patterns of influence to better describe the complex influences in real-world situation, and discussed a series of theoretical problems of influencing and influenced structure, influence from more than one origins, and influence with abstentions and constraints. We then test the models of influence from an exemplary perspective for interdisciplinary study, from both social science and computer science paradigms, by both qualitative case studies approach and quantitative matching algorithms approach, to provide an evaluation for the models of influence.Finally, we use the models of influence to perform agent-based simulations, by the example UN Security Council voting. We design those experiments from both social and computer science perspectives, implement it in Netlogo, and discuss the interleaved effects between new cases of influence and different SC reform schemes. ; L'objet de cette thèse (bi-disciplinaire, en informatique et sciences politiques) est l'étude de phénomènes d'influence dans des situations de prise de décision collective, impliquant plusieurs sujets interdépendants, ce qui amène à considérer des domaines dits combinatoires. L'importance de ces phénomènes d'influence est observée depuis de nombreuses années dans différents domaines (dont les sciences politiques), mais leur modélisation reste peu développée jusqu'à présent. En se basant sur un modèle existant intégrant la prise en compte de l'influence à un langage de représentation des préférences ordinal, le premier travail de cette thèse a été d'identifier des schémas d'influences pouvant survenir, entre sujets et/ou agents différents. Une deuxième contribution de cette thèse est la prise en compte de domaines partiels (permettant de représenter contraintes exogènes et abstentions, par exemple), une nécessité dans les situations envisagées. De même, différents modèles visant à capturer l'influence simultanée de plusieurs agents sont présentés et discutés. Dans un deuxième temps, l'exemple des votes au conseil de sécurité del'ONU est utilisé: l'occurence de ces schémas est envisagée à l'aide d'une analyse qualitative de différentes séquences de votes (sciences politiques), et de la recherche d'indices de schémas similaires dans les données de vote accessibles. Enfin, le modèle est mobilisé dans le cadre d'une simulation multi-agents (réalisée en Netlogo), et visant à tester différentes réformes envisagées du conseil de sécurité.

AbstractThe New Development Bank (NDB) and the BRICS Contingent Reserve Arrangement (CRA) were built by Brazil, Russia, India, China, and South Africa (the BRICS countries) in response to the 'disappointing' Bretton Woods Institutions, including the World Bank and the International Monetary Fund (IMF). The main difference between the Bretton Woods Institutions and the BRICS institutions is that the latter adopts an equal‐weight system instead of a weighted voting system in the NDB (all five BRICS countries have equal shares and voting weights regardless of the major differences in their economic weights) and uses consensus instead of majority voting rules to make decisions regarding critical issues in the CRA. In fact, both the equal‐weight system and consensus rule are unprecedented in the history of multilateral financial institutions. The World Bank, the IMF, the Asian Development Bank and even the Asian Infrastructure Investment Bank (AIIB) all adopt weighted voting systems and majority voting rules. We explain why the BRICS countries adopt these unprecedented designs through precise power measurements and carefully designed experiments. First, we measure the voting power of each member of the BRICS' NDB and CRA under the current decision‐making systems, using this information as a control group. Then, we remeasure the voting power of each member of the NDB under the assumption that their shares and voting weights are allocated according to their economic weights; additionally, we remeasure the voting power of each country of the CRA under the assumption that majority rules are adopted when decisions are being made on critical issues. Moreover, we establish evaluation indices of power equality and power equity based on classic and variant Gini coefficients and Lorenz curves and use them to assess and compare the power structures of the BRICS' NDB and CRA in the contexts of different voting weight reallocations and different voting rules redesigns. We use these indices to precisely analyse whether the power structures of the BRICS institutions become more equable or equitable in the examined contexts and to investigate the relationship between equality and equity in this setting.

AbstractAs the first multilateral development bank (MDB) initiated by China, the Asian Infrastructure Investment Bank (AIIB) has been growing rapidly, and within three years of its establishment, has become the second largest multilateral development bank after the World Bank. With the assistance of the supercomputer (Weiming‐1) and combing the method of generating functions, we measured precisely the voting power of each member of the AIIB (analysis which has been done for the first time for an organization with more than 80 members and more than 1 million distributed votes). We find that a host of member countries other than China have outsized power to prevent action on significant matters, but their power to initiate action are quite modest and closely distributed. Furthermore, the power structures in the AIIB are relatively even or 'democratic' on substantive matters when compared to procedural matters. More interestingly, we analyzed the dynamics of voting power of major members like China, India, Russia, Germany, the UK and France with the admission of new members. While the voting weights of all existing members decreased monotonously and proportionally, the voting power of some members in some cases even increased. We find that the requirements of voting rules, and the size of new and existing members, all affect how members' power evolve. Lastly, we measured the balance of power and the decision‐making efficiency of the AIIB as a whole organization beyond individual members and discussed relationships between them. Our analysis aims to provide insights into the allocation of voting weights, the design of voting rules and importantly, the choice of membership expansions for growing multilateral development banks.