1,814,682 articles were selected from the NELA-GT-2018 dataset (a dataset from a variety of sources with labels pertaining to quality and ideology). A CSN (described above) was used as a relationship graph - a TF-IDF matrix was generated for all articles in the dataset, and the cosine similarity between each article vector pair was computed. Pairs with a cosine similarity greater than 0.85 were extracted to create a directed graph with edges whose thickness indicates the similarity between two groups. Quality scores and political leaning scores were then generated for each article using the labels in the dataset. User profiles were generated with a variety of trusted sources. The simulation was ran over a set time period and the changes in news consumption habits were measured.