How does network structure affect how teams find solutions and how do solution strategies spread? We investigate collective problem solving on a rugged landscape with dyads as the problem-solving unit, i.e. participants are paired with network neighbors to seek a solution to the problem. Inspired by the 80’s sci-fi movie Close Encounters of The Third Kind, we set up a networked multiplayer game where participants had to signal peace to invading aliens from space by finding the four-tone melody signaling peace. The game proceeded through consecutive rounds, and in each round of the game, participants were matched in dyads and through online chat and a shared virtual keyboard had to jointly construct a melody. Melodies translated to points according to a rugged landscape score system. We compared two network topologies: a lattice network (participants only play with immediate neighbors), and a fully connected network (see Figure 1). Furthermore, we manipulated the nature of the problem so that the maximum possible score was either known to the participants or not. Based on experiments with ~400 participants we analyze and compare exploration behavior and how good solutions diffuse in the two network types under known and unknown maximum score. As a benchmark we also compare the two networks to isolated pairs of individuals.
