Information Sharing to Reduce Misperceptions of Interactions Among Complementary Projects: A Multi-Agent Approach

Agents who invest periodically in two complementary projects i and j try to minimize shortfall due to misperceptions concerning the interaction α between i and j. Previous studies have analytically solved such problems but they have been limited to two agents making one decision. We set out with the hypothesis of a large number of deciders sharing information with their nearest neighbors in order to improve the understanding of α. After each period of time, they exchange information on their real payoff values which enables them to choose the best neighbor expected perception of α in order to minimize their shortfall. To model this situation, we used an agent-based approach and we considered that the payoff information transmission was more or less efficient depending on the difficulty to assess the real values or when agents voluntarily transfer wrong data to their neighbors. Our simulation results showed that the total shortfall of the network: i.) declines when α is overestimated, ii.) depends on the initial agent's opinions about α, iii.) evolves in two different curve morphologies, iv.) is influenced by the quality of information and can express a high heterogeneity of final opinions and v.) declines if the size of the neighborhood increases, which is a counterintuitive result.