Transforming Social Media: The Algorithmic Dilemma

There’s a growing consensus that social media platforms have deteriorated into toxic environments, rife with ideologically-driven cancel culture and rampant conspiracy theories.

Platforms like X and Facebook face criticism for fostering hatred and strife. Recent riots in the UK demonstrate how a few incendiary posts can spark widespread outrage and animosity.

In reaction, various governments are tightening restrictions on free speech. Nations such as Turkey and Venezuela have imposed bans on certain platforms, while the UK has incarcerated individuals for inciting violence, or merely expressing unpopular opinions.

However, addressing social media issues may require more than banning accounts or censoring “misinformation.”

The underlying issue appears to be the manner in which social media algorithms elevate conflict-driven content in pursuit of user engagement and advertising revenue.

“This may sound unconventional, but the current debates around free speech are a distraction,” observed Jack Dorsey, former head of Twitter, at a recent conference. “The real discussion should focus on free will.”

Creating a Marketplace for Algorithms

Dorsey posited that opaque social media algorithms distort our perceptions and impede our autonomy. He advocates for enabling users to select from various algorithms, allowing them greater control over the content they encounter.

“Allow individuals to choose the algorithm that resonates with them, or let them design their own algorithms to integrate with existing networks,” he articulated. “This could create a marketplace of algorithms.”

Although this concept is intriguing, implementing it on mainstream platforms poses significant challenges.

Challenges of Algorithmic Choice

Arvind Narayanan, a computer science professor at Princeton, has conducted extensive research into the societal impacts of social media algorithms. He agrees that Dorsey’s proposal has merit, but believes it is unlikely to materialize on major platforms.

“A marketplace for algorithms could provide an essential intervention,” he said. “However, centralized platforms restrict user control over their feeds, and the trend has been toward diminishing this control even further.”

“I anticipate that centralized platforms will resist the integration of external algorithms, much like they currently minimize user controls. This makes decentralized social media all the more crucial.”

There are initial attempts at decentralized platforms like Farcaster and Nostr, but Bluesky—a spinoff of Twitter—is the most developed and incorporates this functionality to some degree, though mostly for specialized content feeds.

Bluesky’s Algorithm Trials

William Brady, an Assistant Professor at Northwestern University, is set to trial a new algorithm on Bluesky that will provide an alternative to the platform’s main algorithm.

Research indicates that as much as 90% of the political content shared online originates from a small, motivated group of highly partisan users. “Mitigating their influence is a crucial goal,” he explained.

The proposed “representative diversification algorithm” seeks to present a broader spectrum of views rather than just extreme perspectives, aiming to maintain engagement without oversimplifying the discourse.

AI-Powered Customized Algorithms

Rick Lamers, an AI researcher, has developed an open-source browser extension designed to filter content. This tool analyzes posts from individuals you follow and automatically hides posts based on their topics and sentiment.

Lamers explained that he created this tool to enjoy content about AI on X without being subjected to contentious political discourse.

“I found myself needing something between unfollowing and engaging with all content, which led to the creation of a tool for selective content filtering.”

The application of large language models (LLMs) for social media content presents fascinating opportunities for crafting personalized algorithms without requiring platform-wide modifications.

However, reorganizing content feeds remains substantially more complex than mere content filtering, Lamers noted.

The Conflict Amplification by Social Media Algorithms

Initially, social media platforms displayed content in chronological order. This changed in 2011 when Facebook introduced the “Top Stories” feature.

Twitter followed suit shortly after, implementing algorithmic timelines that fundamentally transformed user experience.

While many users bemoan the presence of algorithms, these systems play a critical role in helping individuals navigate vast amounts of online content to discover engaging posts.

Dan Romero, founder of the decentralized platform Farcaster, noted that leading consumer applications utilize machine learning-based feeds as they align with user preferences.

“This reflects a clear consumer preference based on the time spent engaging with content,” he stated.

Unfortunately, data suggests that algorithms tend to prioritize emotionally charged and divisive content, which ultimately threatens social cohesion.

The Dynamics of Social Media Bubbles

The conventional understanding of social media bubbles—where users are exclusively exposed to like-minded content—may be somewhat misleading.

While these bubbles do exist, users are often confronted with opinions they abhor, as they’re more inclined to interact with content that provokes anger, whether through disagreement or ridicule.

This cyclical interaction reinforces their beliefs while highlighting extreme views from opposing perspectives.

Much like the tobacco industry in the 1970s, social media companies recognize the societal harm driven by a relentless focus on engagement, but the lucrative stakes may hinder meaningful change.

For instance, Meta recently reported over $38 billion in advertising revenue, largely attributed to its AI-driven ad placements, despite internal explorations of “bridging algorithms” aimed at fostering understanding among users that were ultimately shelved.

Exploring Decentralized Solutions: Bluesky, Nostr, and Farcaster

Dorsey’s ambition for meaningful transformation led to the establishment of Bluesky as a decentralized, open-source alternative to Twitter. However, disillusionment with its progress prompted him to realign with Bitcoin-enthusiast platform Nostr.

This decentralized network empowers users with the ability to choose their preferred clients and relays, potentially providing access to a diverse array of algorithms.

Nevertheless, the development of effective algorithms remains a substantial challenge, posing a hurdle for community-driven efforts.

At a recent hackathon, developers attempted to create a decentralized feed market on Farcaster, although it failed to garner significant interest due to concerns over performance and user experience.

“Crafting a high-quality machine learning feed is demanding, requiring substantial resources for optimization and real-time functionality,” Romero explained.

“To facilitate a marketplace of effective feeds, a back-end infrastructure would likely be necessary, raising privacy issues that need careful consideration.”

Ultimately, it remains uncertain whether users would be willing to invest in individual algorithms.