Introduction Clients in an uncertain world can benefit from being informed about the choices available to them. In some settings, such as grade school education, that process is largely dictated by an informing agent, such as a teacher. In others, key choices--such as what major to choose in college--are more likely to rest in the hands of the client being informed. In these latter settings, an interesting question becomes: should the client go for information regarding the choices that must be made? There would seem to be three obvious places that a client--henceforth referred to as an agent for the sake of consistency with standard modeling terminology--might go in search of insights. (1) The agent could seek the advice of an expert. Alternatively, the agent could seek the guidance of agents. Within the other category, however, there is a further spectrum of choices from (2) agents very similar in characteristics to the original client (homophilic agents) to (3) agents very different from the original (heterophilic agents). An extensive empirical research stream on the diffusion of innovations strongly supports the generalization that most informing takes place through networks of homophilic agents (Rogers, 2003, p. 307). From an information theory perspective, however, this preference seems odd. Presumably, both experts and heterophilic agents offer much greater potential for knowledge gain since their respective knowledge states differ most from that of the client seeking to be informed. Nevertheless the pattern of preference for homophily is observed over and over in the real world. One explanation that has been proposed for homophilic preference relates to the underlying complexity of the landscape upon which agents operate (Gill, 2010). In this conceptual scheme (Gill, 2011), an agent's state might be modeled as a collection of values (e.g., 0s and 1s), with each state being assigned its own ordinal fitness value. Where a landscape is orderly, the fitness of a particular state can be determined with a relatively compact formula. For such landscapes, general expertise with respect to the underlying formula should be of great value. At the extreme, where a landscape is maximally complex--sometimes referred to as chaotic--knowledge of the fitness of a particular state tells you nothing about any state on the landscape. On such landscapes, trial and error will likely be as good as any strategy in finding high fitness states. In between the ordered and the chaotic, the rugged landscape exhibits characteristics that are less extreme. While fitness values on such a landscape do not appear to be random, neither can they be modeled with a simple formula. On such landscapes, it is proposed that homophilic agents may prove particularly successful at searching for high fitness. The goal of the current paper is to report the results of a simulation intended to test the proposition that the benefits of homophily grow with the underlying ruggedness of the landscape. To simulate underlying landscape, the NK landscape model developed by evolutionary biologist Stuart Kauffman (1993) is used. This model was selected for two reasons: 1) it is widely used in many fields, including business, and 2) it provides a parameter (K) that allows complexity to be tuned. For purposes of comparison, an alternative cluster model of ruggedness is also tested. To simulate different agent types, algorithms defining the respective behaviors were implemented in program code. The paper begins with a brief review of the literature relating to homophily and diffusion, intended to motivate and clarify the research questions being asked. The components of the model used in the simulation are then described; appendices providing a more detailed description of the landscapes employed and the software developed for the simulation appear at the end of the paper. The results of the simulation and accompanying sensitivity analyses are then presented, followed by a discussion of their significance. …