We design an experiment to study how agents share and make use of information in networks. Agents receive payo -relevant signals automatically shared with neighbours. We compare the use of information in different network structures, considering games in which strategies are substitute, complement, and orthogonal. To study the incentives to share information across games, we also allow subjects to modify the network before playing the game. We find behavioural deviations from the theoretical prediction in the use of information, which depend on the network structure, the position in the network, and the strategic nature of the game. There is also a bias toward oversharing information, which is related to risk aversion and the position in the network.
