Artificial intelligence and its collective ideation: doing a media archaeology on Wikipedia. Datasets

Datasets for the paper Stanusch, N, Rogers, R, Sánchez Querubín, N (2026). “Artificial intelligence and its collective ideation: doing a media archaeology on Wikipedia.” Preprint. The data collection notebooks are available under the DOI 10.5281/zenodo.20446278 revisions_dictionary_w_mentions.pkl This is a joblib-serialized (a library to serialize data, as the pickle library was crashing) python dictionary. The key of the dictionary is a wikipedia page name. The value is an descending-ordered array of a "revision" dataclass. The dataclass holds the time of the revision and the mentioned wikipedia pages in the revision. historical-wiki-grouped-251211.gephi Gephi graph of the above data, showing communities that emerged, each in a different color. Each node represents a wikipedia page and each edge is a mention. This is a dynamic graph, which means that one can select a range of time that applies to the graph, in which case only the existing edges in that timeframe are shown. The size of each node is dependent on its "in-degree" value, which reflects the number of pages mentioning that node. We defined a community based on its modularity class. This approach results in 8 communities emerging. We then proceed to isolating each community and re-run a modularity class on that community, resulting in subcommunities emerging. Each community is labeled by a number from 1 to 8. We reflect the label of the respective community in their subcommunities by multiplying each original community label by 10 and adding to it the subcommunity's own label. Each subcommunity is labeled 0 to 8, with some communities having far fewer subcommunities forming. This means that community 34 is the 5th (as we label the subcommunities starting from 0) subcommunity in the 3rd community.