🚧 Under development 🚧
The goal of bsky-net is to benchmark belief dynamics models, assessing their accuracy in predicting actual beliefs and enabling comparisons between different models.
Concretely, bsky-net is a temporal graph dataset of user connections, communications, and beliefs over time, using real data from the Bluesky social network. These three components enable more accurate depictions of network structure, timing of belief updates, and measurement of model accuracy, respectively.
bsky-net uses a newly-available, nearly-complete record of over 1 billion interactions on the Bluesky social network. As of 8/27/24, the data includes:
- 6.2M users
- >300M posts (including quotes, replies)
- >1B likes
- >100M follows
- Among other events, like reposts, blocks, etc.