Fig. 4: A two-dimensional representation of the latent ideology, split according to political and non-political influencers.
From: Growing polarization around climate change on social media
Triangles label the median ideological positions of accounts affiliated with specific political parties. Circles indicate the median positions of users who tweeted a particular topic, as derived using BERTopic. In Fig. 2, the latent ideology is calculated using the top 300 most retweeted accounts. Here we calculate the latent ideology twice using (1) the top 300 most retweeted accounts affiliated with individual elected politicians (x axis; political dimension) and (2) using the top 300 most retweeted accounts excluding politicians (y axis; climate dimension). The non-political axis can be thought of as the general climate dimension, whereas the political axis can be thought of as capturing the specific political groupings of the COP discussion. Some topics are merged into a single point for visual clarity.
