When deciding whether to adopt a new technology or to join a social movement, agents make rational decisions that take into account the decisions of their peers. We propose a game-theoretic approach to diffusion on networks, encompassing many situations that have previously been studied separately. In the model, starting from a seed adopter, existing adopters decide to invite their neighbors to adopt a new good. Agents value the number of adopters (accomplices), and the number of neighbors of adopters (witnesses). We find that whether agents like/dislike additional accomplices/witnesses impacts the resulting level of diffusion. We rank these preferences in by infectiousness. We characterize the impact of changes to network topology on the level of diffusion and identify key players.