POLITISKY24 (Political Stance Analysis on Bluesky for 2024) is a first-of-its-kind dataset for stance detection, focused on the 2024 U.S. presidential election. It designed for target-specific user-level stance detection and contains 16,044 user-target stance pairs centered on two key political figures, Kamala Harris and Donald Trump. In addition, this dataset includes detailed metadata, such as complete user posting histories and engagement graphs (likes, reposts, and quotes).

Stance labels were generated using a robust and evaluated pipeline that integrates state-of-the-art Information Retrieval (IR) techniques with Large Language Models (LLMs), offering confidence scores, reasoning explanations, and text spans for each label. With an LLM-assisted labeling accuracy of 81%, POLITISKY24 provides a rich resource for the target-specific stance detection task. This dataset enables the exploration of Bluesky platform, paving the way for deeper insights into political opinions and social discourse, and addressing gaps left by traditional datasets constrained by platform policies.

In the uploaded files:

• The file 'Human_annotation_on_validation_users.csv' contains human-annotated stance labels for 445 validation users toward Trump and Harris, resulting in a total of 890 user-target pairs.
  The labels are divided into four stances: 1 (favor), 2 (against), 3 (neutral), and 4 (unrelated). To simplify the stance annotations provided by the large language model, the "neutral" and "unrelated" categories are combined and represented as "neither."
• The file 'LLM_annotation_on_validation_users.json' contains stance labels annotated by a state-of-the-art LLM for 445 validation users toward Trump and Harris, resulting in a total of 890 user-target pairs. In addition to stance labels, each pair includes an explanation of the reasoning, the source tweets, spans from the source tweets used in the reasoning, and a confidence score.
• The file 'LLM_annotation_on_dataset_users.json' is similar to 'LLM_annotation_on_validation_users.json but is generated for all dataset users excluding the validation set. It provides stance labels for 8,022 users toward Trump and Harris, totaling 16,044 user-target pairs.
• The file 'Main_dataset_for_stance_detection.parquet' contains up to 1,000 recent English-language posts (including both original posts and reposts) from each of the 8,022 + 445 = 8,467 users. This file was used for the stance detection task.
• The file 'Bluesky_dataset_on_us_politics.parquet' is similar to 'Main_dataset_for_stance_detection.parquet', but it contains all posts (including both original posts and reposts) from each of the 8,022 + 445 = 8,467 users.
• The file 'Like_network.parquet' captures users' interactions through likes. Specifically, it contains the number of likes each user has given to original posts made by other users. It includes likes from 8,022 + 445 = 8,467 users, but it is not limited to interactions from these users alone.
• The files 'Repost_network.parquet' and 'Quote_network.parquet' are similar to 'Like_network.parquet', but they capture users' interactions through reposts and quotes, respectively.

Bluesky Social Dataset

Pollution of online social spaces caused by rampaging d/misinformation is a growing societal concern. However, recent decisions to reduce access to social media APIs are causing a shortage of publicly available, recent, social media data, thus hindering the advancement of computational social science as a whole. We present a large, high-coverage dataset of social interactions and user-generated content from Bluesky Social to address this pressing issue.

The dataset contains the complete post history of over 4M users (81% of all registered accounts), totaling 235M posts. We also make available social data covering follow, comment, repost, and quote interactions.

Since Bluesky allows users to create and bookmark feed generators (i.e., content recommendation algorithms), we also release the full output of several popular algorithms available on the platform, along with their “like” interactions and time of bookmarking.

Dataset

Here is a description of the dataset files.

• followers.csv.gz. This compressed file contains the anonymized follower edge list. Once decompressed, each row consists of two comma-separated integers representing a directed following relation (i.e., user u follows user v).
• user_posts.tar.gz. This compressed folder contains data on the individual posts collected. Decompressing this file results in a collection of files, each containing the post of an anonymized user. Each post is stored as a JSON-formatted line.
• interactions.csv.gz. This compressed file contains the anonymized interactions edge list. Once decompressed, each row consists of six comma-separated integers representing a comment, repost, or quote interaction. These integers correspond to the following fields, in this order: user_id, replied_author, thread_root_author, reposted_author,quoted_author, and date.
• graphs.tar.gz. This compressed folder contains edge list files for the graphs emerging from reposts, quotes, and replies. Each interaction is timestamped. The folder also contains timestamped higher-order interactions emerging from discussion threads, each containing all users participating in a thread.
• feed_posts.tar.gz. This compressed folder contains posts that appear in 11 thematic feeds. Decompressing this folder results in 11 files containing posts from one feed each. Posts are stored as a JSON-formatted line. Fields are correspond to those in posts.tar.gz, except for those related to sentiment analysis (sent_label, sent_score), and reposts (repost_from, reposted_author);
• feed_bookmarks.csv. This file contains users who bookmarked any of the collected feeds. Each record contains three comma-separated values: the feed name, user id, and timestamp.
• feed_post_likes.tar.gz. This compressed folder contains data on likes to posts appearing in the feeds, one file per feed. Each record in the files contains the following information, in this order: the id of the ``liker'', the id of the post's author, the id of the liked post, and the like timestamp;
• scripts.tar.gz. A collection of Python scripts, including the ones originally used to crawl the data, and to perform experiments. These scripts are detailed in a document released within the folder.

Citation

If used for research purposes, please cite the following paper describing the dataset details:

Andrea Failla and Giulio Rossetti. "I'm in the Bluesky Tonight: Insights from a Year's Worth of Social Data." PlosOne (2024) https://doi.org/10.1371/journal.pone.0310330

Right to Erasure (Right to be forgotten)

Note: If your account was created after March 21st, 2024, or if you did not post on Bluesky before such date, no data about your account exists in the dataset. Before sending a data removal request, please make sure that you were active and posting on bluesky before March 21st, 2024.

Users included in the Bluesky Social dataset have the right to opt-out and request the removal of their data, per GDPR provisions (Article 17).

We emphasize that the released data has been thoroughly pseudonymized in compliance with GDPR (Article 4(5)). Specifically, usernames and object identifiers (e.g., URIs) have been removed, and object timestamps have been coarsened to protect individual privacy further and minimize reidentification risk. Moreover, it should be noted that the dataset was created for scientific research purposes, thereby falling under the scenarios for which GDPR provides opt-out derogations (Article 17(3)(d) and Article 89).

Nonetheless, if you wish to have your activities excluded from this dataset, please submit your request to blueskydatasetmoderation@gmail.com (with the subject "Removal request: [username]"). We will process your request within a reasonable timeframe - updates will occur monthly, if necessary, and access to previous versions will be restricted.

Acknowledgments:

This work is supported by :

• the European Union – Horizon 2020 Program under the scheme “INFRAIA-01-2018-2019 – Integrating Activities for Advanced Communities”,
  Grant Agreement n.871042, “SoBigData++: European Integrated Infrastructure for Social Mining and Big Data Analytics” (http://www.sobigdata.eu);
• SoBigData.it which receives funding from the European Union – NextGenerationEU – National Recovery and Resilience Plan (Piano Nazionale di Ripresa e Resilienza, PNRR) – Project: “SoBigData.it – Strengthening the Italian RI for Social Mining and Big Data Analytics” – Prot. IR0000013 – Avviso n. 3264 del 28/12/2021;
• EU NextGenerationEU programme under the funding schemes PNRR-PE-AI FAIR (Future Artificial Intelligence Research).

This dataset consists of all starter packs and all following network data available on Bluesky in January and February 2025. Starter packs can be created by any Bluesky user. They are lists of users and curated feeds with a minimum of 6 and a maximum of 150 users, curated by the starter pack creator. The creator typically names them and provides a description. Other users can use a single click to follow all users in the starter pack, or they can scroll through a specific starter pack to decide who to follow within that starter pack. In our dataset, all DIDs (persistent, unique identifiers) are anonymized with a non-reversible hash function; users in the network, as well as users who created starter packs, or appear in starter packs, are identified by their hashed DIDs. Similarly, starter packs themselves are identified by their hashed identifiers.

First, we include the Bluesky following network as it appeared in late January/early February 2025. This shows all available directed following relationships on Bluesky. We also include a network dataset of starter packs with information on creators and starter pack members. This is intended for users who wish to undertake a computational analysis of the networks created by starter packs or starter packs’ influences on networks.

This repository contains the datasets required to reproduce the results presented in the paper "The Rise of Bluesky."

• profile_creations.parquet.zip: Profile creation dates for each user (Fig.1a).
• user_group_date_intervals.pickle.zip: The dates corresponding to each user group (Fig.1a).
• total_engagement_by_day_by_user.parquet.zip: Daily total activity per user group (Fig.1b, d).
• active_user_count_per_day_0_8.parquet.zip: Daily number of active users (Fig.1e, f).
• summary_stats.zip: Files containing daily network statistics such as average degree and node count (Fig.1g).
• group_degree_distributions.zip: Daily user group out-degree distributions (Fig.1g).
• clustering_coef.parquet.zip: Clustering coefficients for follower network for each day (Fig.1h).
• gini_and_kappa.zip: Daily Gini and Kappa statistics (Fig.1i).

Due to its large size, the dataset used to construct the follower network in Fig. 1c is not included here. However, it may be made available upon request under exceptional circumstances.