On Monday, March 4, 2024, CITP and DeCenter co-hosted a workshop on the topic of decentralized social media. The invite-only workshop brought together a diverse group of leaders and innovators from the growing decentralized social media sphere, including scholars, engineers, and administrators who actively study, build, and manage decentralized social networks such as Mastodon, Bluesky, and Nostr. The workshop included a keynote on decentralized online governance by Nathan Schneider, a panel on content moderation led by Aaron Shaw, and multiple “unconference” breakout sessions ranging from “Comparative Analysis of Protocols” to “Funding Community-owned Decentralized Social Media,” where participants split into different tables to discuss specific topics of interest. Researchers and practitioners came together to explore the possibilities of a social media landscape that is collectively owned and governed. The discussions articulated the existing challenges and identified potential solutions to foster a more democratic, secure, and user-centric future for social media.
Here, we summarize a few key takeaways and questions raised from the discussions around five emergent themes:
1. Community and Platform Governance
Discussion in both the keynote and unconference sessions revolved around the governance of online communities in decentralized social media platforms. Participants identified the need for flexible governance structures to accommodate the diverse needs and functionalities of different services within the decentralized social media ecosystem. This flexibility should allow for experimentation and democratic rule-making, ensuring that governance adapts to changing needs and innovations. Also, participants noted the problematic power dynamics on platforms like Nextdoor, leading to segregation and a lack of true local engagement. The discussion highlighted the sparse network issue and the lack of personalized content, contributing to a weakened sense of community.
The discussion also included potential solutions, such as stewardship-first approaches that emphasize community values and collective responsibility over technological solutions. Drawing inspiration from existing governance models of institutions such as W3C and ALA, conversations emerged around the pros and cons of institutionalization. Participants recognized the drawbacks, such as burnout and potential obsolescence, while also exploring the concept of intentionally ephemeral institutions or coalitions. This involved discussing planned “spin-down” cycles, wherein platforms and associated assets are provisioned and decommissioned within fixed time intervals, such as every five years.
Participants also raised open questions about the practice of governing online communities in decentralized social media:
- In the pursuit of democracy, how to make sure minority voices are amplified to avoid being drowned out?
- Given the varying norms, values, and technical capabilities of different communities, how feasible is it to implement uniform governance structures across diverse decentralized platforms?
- How do we balance the decentralization of governance with some form of centralization of the underlying technology to facilitate threat analysis and cross-platform sharing?
2. Platform Trust and Safety
Participants highlighted the importance of developing adaptable trust and safety measures for decentralized social media, aiming to avoid market dominance by a few platforms to maintain true decentralization. They also delved into data governance issues, emphasizing the need for ethical data use that respects user consent and privacy, amidst the challenges posed by evolving social norms and legal standards. These discussions underscored the need for transparent and accountable data governance mechanisms that balance privacy protection with beneficial data use, advocating for governance models that consider the interests of tech developers, users, and regulators to ensure responsible data handling.
Decentralized social media has opened up opportunities for developing third-party curation and moderation services/microservices, potentially outsourced for trust and safety management. Nonetheless, privacy and sustainability present significant challenges in this context.
Here are some open questions that emerged from the discussion focus on the practical implementation of the ideals in decentralized systems:
- How can privacy and autonomy be maintained without hindering the potential insights gained from data analysis?
- What innovative consent mechanisms can be developed to ensure users’ preferences are respected in complex digital ecosystems?
- How can UX design support users in making safe decisions when multiple options are available?
3. Content Moderation
Content moderation in decentralized social media platforms differs from traditional centralized platforms, presenting opportunities and challenges.
Among these, the time and labor required for moderation stand out, highlighting the need for automated tools that support community-centric operations. Furthermore, the decentralized nature of these platforms often results in a lack of institutional support, making information sharing and exchange problematic in some instances.
Some specific issues, such as CSAM (child sexual abuse material) and the use of generative AI, were salient topics during the workshop. The CSAM discussion centered around the issue of how to scan for CSAM, raising critical questions about coordinating scanning tools with platform protocols and whether single or multiple entities should provide these tools. For generative AI, participants posed concerns regarding misinformation and privacy violations while recognizing some positive applications for using generative AI to enhance customization and personalization of content moderation. There was also some skepticism about the severity of generative AI-enabled misinformation and its potential harms; for example, participants questioned whether text-to-image generative models have made things much worse when products like Photoshop have already existed for a long time and could be used to propagate similar misinformation.
Additionally, there was a strong advocacy for more effective content curation to boost community engagement. Participants critiqued the “For You” feed approach for leading to context collapse and diminishing meaningful community interactions, expressing interest in exploring solutions to these issues on decentralized social media platforms.
Some of the open questions include:
- What could be the potential impact of generative AI in low-salience contexts, such as local elections, where resources for misinformation campaigns were previously very limited?
- What’s the impact on users when generative AI enables “epistemic one-stop shopping,” i.e. allowing users to curate information according to their values and preferences, potentially leading to value filter bubbles?
4. Artifact Design and Development
In the realm of decentralized social media, protocols play a crucial role, with various ecosystems developing around them. Popular protocols in decentralized social media include ActivityPub (e.g., Mastodon), AT Protocol (e.g., Bluesky), and Nostr (e.g., Damus). During the workshop, participants from industry and academia brought up topics about protocol design and development. Participants were interested in misalignments between protocol designers and users. Designers and engineers shared insights and lessons learned, especially about the mechanism of data deletion in these protocols. For instance, BlueSky designed its protocol to make data unremovable, but it turned out that users preferred the content to be removable.
A pivotal question emerged in the discussion: Will a single protocol eventually dominate, or will there be interoperable bridges between different protocols? While some bridges currently exist to facilitate coordination, the variation in content moderation policies across platforms complicates the moderation of content that spans multiple protocols. Participants also noticed some tensions between maintaining a decentralized ethos and the practicalities of swift decision-making in protocol development. Thus, Nostr introduces a model of nominated veto power to balance diverse developer inputs and maintain protocol integrity. Moreover, inclusivity and diversity in protocol development were highlighted as challenges, emphasizing the need for broader participation beyond the developer-centric community. For example, how might individuals without a coding background influence protocol development? It remains a challenge to build for minority users when the majority of the developers are not members of said minority groups.
Participants also delved into issues of identification, authentication, and authorization, discussing the pros and cons of existing implementations such as WebFinger for identification, OAuth2.0 for authentication, and blocklist for authorization. User experience (UX) design was another area of focus, with conversations about meeting the unaddressed needs of users in decentralized environments and exploring potential solutions through design innovations.
Yet, some questions remained open for further debate, including:
- What kind of decentralization do users need and want?
- Do we need a universal representation of identifiers across protocols? If yes, how can we enable it?
5. Funding
Discussions on funding decentralized social media platforms focused on creating sustainable models that support both user engagement and platform maintenance without compromising the ethos of decentralization. A central theme was the exploration of innovative monetization strategies that go beyond traditional advertising or donation models, including transactional systems that offer fair compensation to creators (e.g., through digital assets or tiered access similar to Patreon models). Participants debated the integration of cryptocurrency to facilitate decentralized payments, despite concerns about user accessibility and maintaining true decentralization. The conversation also addressed the practicality of ads, with a focus on designing accountable advertising models that users would feel comfortable with and the potential for a dedicated ad layer in decentralized protocols. Furthermore, the discussion highlighted various funding sources, from community-driven micro-payments and membership fees to leveraging smart contracts for sponsorships. The aim was to ensure platform sustainability, balancing the need for financial support with the desire to keep platforms accessible and true to decentralized principles, while also considering the human costs involved in maintaining these ecosystems.
Some of the open questions include:
- How do we ensure the sustainability of decentralized platforms beyond immediate financial considerations, addressing issues like volunteer burnout?
- What are the trade-offs and accessibility barriers of different funding approaches, such as through advertising, foundations, membership fees, and the government?
- Can we use regulatory compliance as a potential funding lever for threat analysis and moderation efforts?
In summary, the workshop covered topics that are frequently discussed in relation to centralized social media platforms, but present new challenges and opportunities in decentralized settings—for example, online governance, trust and safety, and content moderation. While pointing out the harms stemming from the current social media landscape, participants maintained a hopeful outlook on the potential design and future of decentralized social media. Additionally, it highlighted unique issues specific to decentralized environments, like protocol development and innovative funding models. We hope the discussions and questions from this workshop will chart out new paths for practitioners and researchers alike to explore the increasingly rich landscape of decentralized social media.
This report was prepared by Yuhan Liu (Princeton University), Varun Rao (Princeton University), Owen Zhang (Princeton University), Ryan Liu (Princeton University), Priyanka Nanayakkara (Northwestern University), Zilin Ma (Harvard University), Kevin Feng (University of Washington), Zhilin Zhang (Oxford University)
Speak Your Mind