Determining models of influence

Michel Grabisch (1, 2) , Agnieszka Rusinowska (1, 2)

1 CES - Centre d'économie de la Sorbonne
2 PSE - Paris School of Economics

We consider a model of opinion formation based on aggregation functions. Each player modifies his opinion by arbitrarily aggregating the current opinion of all players. A player is influential on another player if the opinion of the first one matters to the latter. A generalization of an influential player to a coalition whose opinion matters to a player is called an influential coalition. Influential players (coalitions) can be graphically represented by the graph (hypergraph) of influence, and convergence analysis is based on properties of the hypergraphs of influence. In the paper, we focus on the practical issues of applicability of the model w.r.t. a standard framework for opinion formation driven by Markov chain theory. For a qualitative analysis of convergence, knowing the aggregation functions of the players is not required, one only needs to know the set of influential coalitions for each player. We propose simple algorithms that permit us to fully determine the influential coalitions. We distinguish three cases: the symmetric decomposable model, the anonymous model, and the general model. JEL Classification: C7, D7, D85

Mots clés: algorithm; social network; opinion formation; aggregation function; influential coalition

Domaines: Economies et finances

Michel Grabisch, Agnieszka Rusinowska. Determining models of influence . Operations Research and Decisions, 2016, 26 (2), pp.69-85. ⟨hal-01387480⟩