Networked problem solving underlies the formative dynamics of a variety of complex systems. Ranging from cellular communication to human organization, collectives engage in goal-oriented behavior and adapt to maximize their rewards. In this setting, it is important to consider not just what network ties exist, but also what behavior or exchange is facilitated by the interactions these ties represent. When considering a network tie between two problem solvers, optimal group-level outcomes often roughly depend on two overarching concepts: cooperation and coordination. The first two projects of this dissertation use stylized models of networked problem solvers to demonstrate how preferential interaction (i.e. adapting network ties to maximize individual rewards) can lead to system dynamics that are initially unintuitive. In one setting, preferential interaction is shown to harm cooperative prosocial behaviors, while in the other it promotes coordination and robustness. The first agent-based model demonstrates the conditions under which spiteful behavior could emerge. Spite is a behavior that harms both the actor and recipient, and therefore its emergence is evolutionarily counterintuitive. The model demonstrates that if agents are able to interact preferentially, spiteful behavior can gain a relative advantage and become the norm in a population. A second agent-based model examines the formation of signaling conventions under preferential interaction. Unexpectedly, we found agents using a novel form of signaling where communicating partners failed to reach a consensus on the meaning of each signal. Instead, agents solved the signaling problem by using the network structure to ensure two distinct types of complementary signaling strategies interacted. Finally, the third project reports on an online survey and experiment conducted on an online labor market platform. This project examines the psychological willingness of freelancers to cooperate in temporary teams, and how our results can inform platform design aimed at strengthening the social ties between freelancers. Overall, our results suggest that our freelance participants display traits that are more consistent with belonging to a shared, coherent group than would be expected given existing theory.