A Blue Start: A large-scale pairwise and higher-order social network dataset

Smith, Alyssa Data Collector (Northeastern University)
Landry, Nicholas Principal Investigator (University of Virginia)
Kumar, Sagar Data Curator (Northeastern University)
Foucault Welles, Brooke Principal Investigator (Northeastern University)
Amburg, Ilya Data Curator (Pacific Northwest National Laboratory)

2025-05-09

v1

Published

This dataset has two levels of access: Public, Login Required, Controlled Download, and Virtual Data Enclave.

• Public files can be downloaded directly from the dataset record. Typically, documentation files such as READMEs and codebooks are made public.
• Controlled Download files have access restrictions due to the sensitivity of the data or data sharing requirements. These data require an application. Once your application is approved, ICPSR will provide the data files for approved researchers to download and use in their own computing environment. Click the “Apply for Access” button below to apply through our application portal.

National Science Foundation GRFP, 1938052

Smith, Alyssa, Landry, Nicholas, Kumar, Sagar, Foucault Welles, Brooke and Amburg, Ilya. A Blue Start: A large-scale pairwise and higher-order social network dataset. Inter-university Consortium for Political and Social Research [distributor], 2025-05-09. [doi forthcoming]

Datasets
Bluesky

2023-01, 2025-02

2025-01-20, 2025-02-07

Bluesky

application programming interface (API)

web platform data

Our intent when collecting this dataset was to understand how the Bluesky network’s growth is shaped by starter packs, if at all. It consists of all available starter packs discoverable via querying individual users' lists of starter packs in January/February 2025, as well as all following data discoverable via querying individual users' following ties in January/February 2025. Each starter pack's contents (i.e. user list) and metadata are recorded in the dataset.

For each account DID (distributed identifier) in the PDS (Personal Data Sever) directory file we obtained in January 2025, we used the app.bsky.graph.getActorStarterPacks public API endpoint to obtain all starter packs associated with that account. Next, for each starter pack, we queried the app.bsky.graph.getList public API endpoint with the DID (also referred to as a URI, or Uniform Resource Identifier) of the list of entities associated with that particular starter pack. Note that entities in a starter pack are not necessarily users; they can be any entity that can be "followed" on Bluesky (e.g. a list or a feed). Querying the app.bsky.graph.getList endpoint allows us to page through and record the entire list of entities in the starter pack. We used the Python asyncio and aiohttp packages to efficiently download data from the API while respecting rate limits.

For each account DID in the PDS directory file mentioned above, we used the app.bsky.graph.getFollows public API endpoint to obtain all users followed by that account. Querying this endpoint allowed us to page through and record all accounts followed by each account. Ultimately, this gave us a snapshot of all following relationships for all accounts in the PDS directory file we used. We used asynchronous programming, as explained above, to efficiently download the data.

All Bluesky users in the PDS directory, all users they follow, and all starter packs they created as of late January/early February 2025.

See codebook for more details.

totaluniversecompleteenumeration

individual

Analysis code will be available.

Python

~250 GB of storage necessary for the raw network data.

Need to obtain a PDS directory list of all user DIDs.