Last summer, Rafa (@rafa on Farcaster) joined the Summer of Protocols to explore how protocols shape our lives. During his research, Rafa looked into how people, bots, and content are coordinating in digital spaces. Here we present an excerpt of the full research for Folklore readers. The full essay is available as a PDF here.

On the internet, we are part of swarms: networks of people, bots, and content, coordinated through algorithmic feedback loops. Swarms are harbingers of misinformation, heralds of mutual aid, and representatives of the public will. But this is not our grandparents’ crowd. Swarms are networked tempests of humans and information. Most importantly, they can act collectively without explicit protocols; they are minimally protocolized entities. In this research, I document a variety of swarms, such as the mutual-aid response to the devastation of Hurricane María, to uncover their unique methods of cooperation. To understand swarms, we also need to understand their peers within the broader category of online formations. This category includes group entities like memetic tribes and online communities that have explicit protocols which separate them from their swarm peers and makes them more explicitly manageable.

This inevitably leads us to the question: How do we steer swarms?

The sky turned lilac at the break of dawn on September 22, 2017. It was an omen of Juracán, the wind spirit, summoned by the indigenous deity Guabancex. Hurricane María, the worst storm since 1899, was about to make landfall in Puerto Rico. It was mythic: the storm crossing on to land one minute after sunrise. Thousands would die in the aftermath. Millions of Puerto Ricans living in the U.S. and around the world waited in anticipation, without a way to connect to or help their family and friends.

A Berlin-based Puerto Rican working in climate venture-building, Jorge Vega Matos fidgeted on his phone through the morning, watching his Facebook feed nervously. He dreaded the challenges his elderly parents and family might face in the coming days and weeks. Unknown to Jorge, an old acquaintance and Puerto Rican social activist in New York named Pablo Benson did the same.

There was a deafening silence online. Everyone was alone but watching together As the silence continued due to an islandwide blackout, Jorge noticed that the Puerto Rican diaspora had begun to share drop-off points on Facebook for aid supplies. Rumors of stronger-than-expected damage crawled from phone to phone, urging those with an internet connection to broadcast their fears and take action. Facebook’s ever-watching algorithm understood the opportunity for engagement and amplified their voices. A collective direction emerged as anxiety transformed into support: Let’s not delay, let’s get supplies to those in need.

There was no banner or organizing institution, just posts about where to go and visual memes with lists of useful supplies. Jorge’s feed became a cacophony of activity. The digital diaspora had mobilized into a swarm: a network of people, content, and bots kept aligned by the algorithms of Facebook’s feed. It was a symbiotic alliance between social media and the Puerto Rican community: both needed to aggregate attention.

By the end of the day in Berlin, while Hurricane María was still ravaging the island, Jorge had created a public spreadsheet to aggregate supply drop-off points circulating on Facebook. He shared the link with various mutual aid groups, in hopes of accelerating donations. In New York, Pablo saw the spreadsheet and immediately forwarded it to others in his network. It turned out that the recipe for their swarmlike collaboration had been simple: provide a platform of connection to those with a common desire. The message found its way across the Atlantic: We’re creating a team and we want to use your spreadsheet.

It was not the presence of certain protocols but their absence that gave the swarm its advantage. While government agencies and aid organizations sought approvals, individuals were able to act freely and broadcast their intentions via posts. Social media had enabled collaboration at scale through algorithms and instant messaging rather than being slowed down by explicit protocols, central planning, or strategic oversight.

The story of Hurricane María highlights the structures and affordances of online swarms. But swarms are all around us. Other instances are regularly referenced in pop culture and daily news: celebrity cancellation raids, misinformation campaigns, fandom hypes, activist rallies, and memestock frenzies.

We can see another swarm’s footprint in the paths to ruin of four banks in 2023. The episode began that year when the Silicon Valley Bank published a surprise announcement on Wednesday, March 8. In it, the bank mentioned that it was taking action to address some liquidity challenges. A frenzy of panicked founders and CEOs (sparked in part by newsletter writer Byrne Hobert’s analysis of the situation) messaged each other frantically as funds (such as Peter Thiel’s Founders Fund) advised them to withdraw their money. A torrent of provocative content flooded digital spaces. The swarm had been summoned.

Screenshots of texts and emails were forwarded from one person to another. Posts on social media and in group chats created a pattern. In turn, the algorithms identified it as highly engaging content and accelerated its reach. The impending collapse manifested itself: within a day, over $42 billion had been withdrawn. The bank was unable to react in time. By March 10, just two days later, the bank was placed under the receivership of the Federal Deposit Insurance Corporation (FDIC). Later, it became evident that the banking collapse was incited by other issues, but the swarm had accelerated its fate, striking like lightning fueled by desperate attention. Three more banks were claimed throughout the summer. Eventually, attention moved elsewhere and the swarm dissipated. The hurricane response and the bank run are examples of online swarms: networks of people, content, and bots. They allow participants to act individually, with collective impact. Through a shared orientation that emerges in algorithmic feedback loops, swarms coordinate without internally directed protocols.

Swarms mirror similar patterns we see in animals, crowds, and guerrillas. Animal aggregations such as fish shoals evade danger by finding dark waters. Army ants build bridges over tricky terrain.

Online swarms also create collective solutions, but navigate networked online worlds instead of purely physical ones. The hurricane response built bridges of information enabling a supply chain of aid. The bank runs evaded financial ruin by collectively removing themselves from the unsafe institutions. Yet unlike animal aggregations, online swarms include a diversity of agents—people, content, bots, and algorithms—each with their own objectives or programming.

It is not sufficient to characterize swarms as a basic mixed crowd. In a crowd, participants gather organically and then act as a collective. We can picture a crowd of people gathering at a sunny spot in the park.

Each person makes an individual decision to join and move from one part of the park to another. From afar, it would seem as though the crowd was working together. In reality, the sun was serving as a shepherd.

For an online swarm, the sun is the algorithm. But swarms operate with key differences. They include spectral objects like bots and content. They are digital-first, while crowds are often physical gatherings: location and mobilization are decoupled in a swarm. Additionally, swarms are deeply networked and everyone can broadcast information or send instant messages to each other. In contrast, crowd participants only communicate with their nearest physical neighbors.

At the intersection of algorithmic navigation and networked crowds, swarms take on guerrilla-like traits. They use their ecosystem as infrastructure for communication and coordination. They also form celllike teams. These teams are autonomous and self-contained, each with its preferred protocols, but globally oriented via an emergent promise.

Read the full essay at: