Looking AT the Blue Skies of Bluesky (2408.12449v2)

Abstract: The pitfalls of centralized social networks, such as Facebook and Twitter/X, have led to concerns about control, transparency, and accountability. Decentralized social networks have emerged as a result with the goal of empowering users. These decentralized approaches come with their own tradeoffs, and therefore multiple architectures exist. In this paper, we conduct the first large-scale analysis of Bluesky, a prominent decentralized microblogging platform. In contrast to alternative approaches (e.g. Mastodon), Bluesky decomposes and opens the key functions of the platform into subcomponents that can be provided by third party stakeholders. We collect a comprehensive dataset covering all the key elements of Bluesky, study user activity and assess the diversity of providers for each sub-components.

• The paper demonstrates that Bluesky’s modular structure—featuring DIDs, personal data servers, relays, feed generators, and labelers—enhances customization while raising scalability issues.
• The paper finds that initial user engagement surged post-launch, yet daily active user levels suggest competitive challenges against established platforms.
• The paper reveals that community-driven moderation via labelers promotes democratic content policing, though it introduces concerns over label standardization and trust.

Analysis of Bluesky Decentralized Microblogging Platform

The paper "Looking AT the Blue Skies of Bluesky" provides a critical and thorough analysis of the Bluesky decentralized microblogging service. This service aims to mitigate the pitfalls inherent in centralized social networking systems, such as control, transparency, and accountability issues. The authors undertake the first extensive empirical paper on Bluesky, evaluating its architecture, user activity, content moderation, and feed generation mechanisms. They leverage a comprehensive dataset to dissect the ecosystem, comprising over 5 million users and hundreds of millions of posts and interactions.

Architecture and Decentralized Components

Bluesky differentiates itself from other decentralized social networks (DSNs) through its partitioned approach, wherein core functionalities are modular and open to third-party providers. The platform breaks down into five crucial components:

1. Decentralized User Identifiers (DIDs): Users are identified via unique, cryptographically verified identifiers, enabling seamless migration across different data servers without losing social graph coherence.
2. Personal Data Servers (PDSes): Users can host data on different servers, thus detaching from dependency on any single provider.
3. Relays: These optional servers aggregate and accelerate the delivery of data from multiple PDSes to end clients.
4. Feed Generators: Independent entities can create algorithms to customize user timelines, promoting a diversity of content visibility paradigms.
5. Labelers: These entities assign labels to posts and accounts for content moderation, reflecting a flexible and distributed approach to enforcing community standards and norms.

This modular structure contrasts sharply with other DSNs, particularly those in the fediverse, which bundle functionalities into single server instances. As a result, Bluesky offers a highly customizable and pluralistic user experience.

User Activity and Growth

The dataset reveals that user activity surged following Bluesky’s launch in November 2022 and subsequent opening to the public in February 2024. Despite a growing diverse user base, including significant Japanese-speaking communities and niche language groups, Bluesky exhibits stagnation in daily active users as of May 2024. The network sustains about 500,000 daily active users, a number significantly smaller compared to established platforms such as Twitter. Insights into language-specific communities underscore how platform dynamics shift with global events and direct marketing efforts.

Content Moderation

The labeling architecture of Bluesky is innovative, facilitating both centralized and community-driven moderation. Initially, content moderation was solely managed by Bluesky PBS, but has since been opened to community-operated Labelers. This has led to a proliferation of labels and varying moderation tactics. Notably:

• Growth of Community Labelers: After opening to community Labelers in March 2024, the ecosystem expanded rapidly, with a majority of labels now issued by the community.
• Diverse Label Values: While some labels such as porn and sexual are widespread, a multitude of other labels cater to niche and specific needs, such as no-alt-text for accessibility.

The community-driven approach mirrors a democratic model of content policing but raises concerns about label standardization and trust discovery among users. The user-centric moderation model allows users to configure reactions to different labels, but there are complexities in balancing flexibility with coherence and reliability.

Content Recommendation

Bluesky’s Feed Generators foster a highly customizable timeline experience. As of April 2024, 40,398 Feed Generators exist, though not all are active. The data reveals substantial activity from these generators tailored to specific interests, languages, or communities. Feed Generators promote user engagement, and the popularity of these feeds often correlates with the creators' broader social influence.

The emergence of Feed Generator As A Service platforms underscores both the innovation and challenges in this decentralized model. Skyfeed, for example, dominates the ecosystem, suggesting centralization tendencies even in distributed settings. Furthermore, economic incentives for sustaining these services might be insufficient, posing scalability issues.

Discussion and Implications

The paper underlines several implications:

• Scalability and Centralization Tendencies: While Bluesky promotes decentralization, critical components like Firehose and AppView controlled by Bluesky PBS highlight inherent centralization challenges. The growing data loads on these components question long-term scalability.
• Economic Sustainability: The current enthusiasm-driven ecosystem could face sustainability issues without clear economic incentives. Exploring monetization methods or diversified revenue models could be essential for long-term viability.
• Compliance with Legal Frameworks: Legal and regulatory compliance, particularly around data privacy and content liability, remains a significant consideration. The Git-like data structure poses challenges for complete data removal under laws such as GDPR.
• Interoperability with Other Platforms: Enhancing interoperability, particularly with the fediverse, could expand Bluesky’s user base and leverage existing decentralized social networking ecosystems.

Conclusion

The analysis elucidates Bluesky as an ambitious frontier in the landscape of decentralized social networking. Its modular architecture, growing user engagement, and innovative moderation and recommendation models exhibit the potential for a more open and democratic social media platform. However, the challenges in scalability, economic sustainability, legal compliance, and genuine decentralization indicate that further evolution and strategic developments are necessary for Bluesky to fully realize its objectives and compete with entrenched centralized platforms effectively.