Andy Updegrove recently wrote about the "death of archeology" by the hand of Wikipedia. Essentially, Updegrove argues that Wikipedia eliminates the task of 'digging up' clues about prior societies and their self-conceptions by immediately providing this kind of material in abundance, conviently in an easily searchable form.
Instead, he argues for the development of anthropological tools to make sense of this wealth of information which can yield useful results if properly read. Part of this analytical toolset is provided by wikipedia itself, e.g. by offering articles on the same topic in its different language versions, allowing for inferences about the culturally specific understanding of a topic, say, a particular culturally sensitive videogame like America's Army.
Comparing the article categories, tags, choice of information etc. of the US vs. German Wikipedia version could, regardless of categories/structures predefined by the Wikipedia architecture, provide some insights into the linguistically and, probably, culturally specific understanding of the topic (or at least the respective modes of knowledge production in that culture...); Michel Foucault would be delighted. For instance, the concept of propaganda exhibited in the respective English and German Wikipedia articles is profoundly different. While the German article reports on the US Army allegedly "admitting" propagandistic intentions connected to the development of AA, the English article has a less 'essentialist' understanding of propaganda, considering the term rather as a historically defined, rhetorical construct used by critics of the game.
Updegrove furthermore suggests that a company like Google should partner with Wikipedia to archive the entire dataset at regular intervals, in this case less than anually. Apart from the fact that this would probably not appear very lucrative and hard to implement for a technology partner in itself, I wonder whether it would be an adequate strategy given the aforementioned particular modes of storing and 'harvesting' anthropologically relevant data.
Such a centralized, top-down approach might defy the whole concept of 'digital anthropology' as Updegrove himself outlines it, as opposed to, say, decentralized policies such as adding more comprehensive timestamping options in Wikipedia architecture which, not right from the start but over time, will lead to most entries being collectively 'updated' to use the new timestamping format. After all, not the entirety of information available is useful for retracing public conceptions of particular topics or artifacts but plausible criteria for selecting and linking a handful of information snippets.
P.S.: Wikipedia itself hosts a list of studies on the topic, including patterns of using it for more advanced purposes.