Community detection is a fundamental challenge in the analysis of online networks. However, there is a lack of consensus regarding how to accomplish this task in a manner that acknowledges domain-specific, substantive social theory. We develop a typology of what social phenomena communities of hyperlinked actors may signify—topical similarities, ideological associations, strategic alliances, and potential user traffic—and offer recommendations for community detection grounded in these concepts. Testing procedures on a hyperlink network of the food safety movement, we demonstrate that the handling of tie directions and weights as well as algorithm choice influence which communities are ultimately detected in such a network.
