The WELL is one of the longest-running communities on the web, started in 1985 to cater to an emerging crop of Silicon Valley tech workers and former hippies living in the Bay Area who were interested in exploring the new and exciting networking technology known as the internet. It was an exemplar of early internet communities, chronicled by Howard Rheingold in his influential book The Virtual Community as a salon hosting good conversation between all sorts of people from far-flung places. Much like today’s message boards or Facebook groups, it was divided into rooms based on interest and topic.

It was also, as described by historian Fred Turner, a site that kept its lights on thanks in large part to the community of Grateful Dead fans who paid for subscriptions to connect there. On the WELL, a Deadhead could find a robust network of fellow Grateful Dead fans discussing news about the band, the state of Deadhead culture, and tape traders mailing each other cassette bootlegs of Dead shows. Tape trading was abundant in the Grateful Dead fandom, as the band gave its blessing to a thriving culture of fans taping their live performances, as long as no one profited from those recordings.

On the WELL and other early experiments in bulletin boards, Deadheads were early pioneers in using the internet to share music. Deadhead music-sharing culture would eventually extend to torrent technology with sites like etree.org and even the Internet Archive. Along the way, Deadheads were collectively innovating a new method of social network: discovery logic.

Discovery logic, in broad strokes, is defined by four user behaviors—each of which was a hallmark of these early online Deadhead communities:

• Curation. The fans put together a huge catalog of available recordings, organized by date and documented with notes like set lists and recording techniques for each individual bootleg.
• Review. Fans often included notes about what made a specific show unique or some anecdotes from seeing that show in person.
• Recommendation. Often, specific shows from that larger curated collection of all available shows would be recommended, usually because it had a particularly good performance of a song or was one of the better shows in a given month of the year, helpful for a band with as deep and deeply checkered a live archive as the Dead’s.
• Discussion. Through forum posts and user comments, users would weigh the merits of particular shows, entire years of performances, specific keyboard players, and so on.

Applied to social media, discovery logic has an important fifth characteristic: user identity. That identity is defined through those other four behaviors. For example, on Goodreads, a discovery-based social network for books, someone who gives all of Octavia Butler’s and N.K. Jemisin’s books five stars is self-identifying as a fan of science fiction written by black women, while someone who has been discussing the merits of various travel books about Brazil might be revealing that they just came back from vacation.

This performance of identity is different from the Instagram or TikTok model. Instead of posting their own original content, like a photo or video, discovery logic users create some record of things they like. In the case of Last.fm, a service that tracks every song a user listens to, a series of graphs are generated showing other users how often someone listens to specific artists or songs. If you see that I listen to a lot of Prince, Dolly Parton, and Gregorian chant, you might either friend me because you, too, love a wide range of vocal styles, or because my taste is so bizarre that you need to see what else I’m interested in.

Some especially visible examples of discovery logic are communities built around user reviews like Goodreads, MyAnimeList, Letterboxd, and RateYourMusic. Each of these offers the same basic affordances across different types of media. Goodreads lets users track, rate, and discover books; MyAnimeList does that for anime television; Letterboxd for movies; and RateYourMusic for music. On Wikipedia, these are grouped as “social cataloging applications,” which highlight a communal approach to knowledge- and taste-making. Discovery logic doesn’t just appear in social networks focused on media: Fragrantica and MakeupAlley are long-running and incredibly popular communities for users to review and compare notes on perfume and makeup, respectively.

Perhaps the most familiar discovery logic platform of all is not one built for entertainment but for food: Yelp. Effectively, Yelp is a platform that aggregates user reviews, which are often a star rating and a short written blurb. It helps you find what to eat for dinner. But even though Yelp is powered by its users, it doesn’t center social interactions like media-based discovery platforms do.

On the surface, Yelp satisfies all five criteria for discovery logic. It curates restaurants in your area, lets users review and discuss those restaurants in the comments, provides recommendations based on star ratings, and gives users the ability to build out profiles identifying them as prolific Yelpers. But since the platform’s emphasis is ultimately on curating and recommending, the social component is incidental—participatory yet not interactive. You can have friends but there isn’t much to do with them. This shallow sociality is a relic of, as the Initiative for Digital Public Infrastructure’s Ethan Zuckerman puts it, “the social mediafication of everything.” You are unlikely to follow an unknown person on Yelp because of their eclectic taste in restaurants, while that’s exactly what social discovery communities around music, movies, or other things are for.

Discovery logic is often but not always social. It exists in a continuum of spaces, from the most to least community-centered. The most social discovery spaces might be forums like Okayplayer—founded in the early 2000s by Questlove, drummer of the popular hip-hop act The Roots—which acted not just as a space for fans to discuss and discover alternative hip-hop in the early 2000s but one where musicians connected to collaborate. The least social discovery space is arguably Spotify, where the majority of the social interaction happens off-platform, with users sharing playlists with each other by other means like text message, Discord, and social media. Spotify effectively outsources sociality to other technologies.

Another way to understand the discovery landscape is as a Venn diagram of community, content, and review. There are some services like Netflix that mainly focus on delivering content, ones like Yelp that are dedicated to crowdsourcing reviews, and ones like the Grateful Dead section on the WELL that build community around a specific shared interest.

Discovery as a technology can inhabit any of the circles in that Venn diagram. Truly social discovery logic, though, will only occupy the intersections. Consider Erowid, which innovated a community approach to reviewing drug experiences, creating a massive trove of trip reports that leveraged the sociality of the internet without being explicitly interactive, like the participatory social media we know today. Erowid doesn’t provide access to any of those drugs, but it does provide content and review, helping users potentially discover (or avoid) substances of interest.

Collections

Discovery logic is at its most social when it inherits traits from gift logic, as outlined by Casey Fiesler in her own entry to this series of articles. In her piece, Fiesler talks about communities of fans creating and sharing original work, and this very dynamic appears in discovery logic too: fans creating and sharing their knowledge of a given subject. Often they do this by building out collections of organized content, which can either be created by one user or by a group of them.

From 2007 to 2016, perhaps the most audacious and comprehensive discovery network of all operated for invited members only, called What.cd. It was an awe-inspiring catalog of recorded music, with a lively forum culture where fans would discuss music and a collections feature called “collages” where users would collate albums based on genre, geography, year, or some sort of aesthetic commonality. Collections are a key part of a discovery platform, and the most exciting collections are ambitious cooperative curation efforts to pool knowledge. In the case of What.cd, for example, users created incredibly comprehensive databases around specific artists, fringe genres, or already-existing collections like Pitchfork’s “Top 100 Albums of the 1990s.”

So why was What.cd so short-lived? It was a private torrent tracker and an incredibly large one at that. Its well-developed discovery technology was tied directly to unauthorized sharing of MP3s, and it shut down when its servers were seized by the French authorities. The tragedy of What.cd is that it would have been useful to users even without the links to torrents but, sadly, the connection to a technology for music “piracy” meant that the baby got thrown out with the bathwater. There was no opportunity for What.cd users to unlink that community from the huge trove of copyright content—it simply disappeared alongside those files.

That does not mean that user interaction is dead in discovery. In fact, Letterboxd, a movie-based discovery platform, has made social interaction via collections the central affordance of its platform. These collections are nearly identical to What.cd’s collages and allow users not only to have discussions on pages for individual movies but on pages for collections too.

Those collections can be defined by anything from a genre to a group of movies referenced by a director in a particular interview to a specific trope. Letterboxd’s collection affordance is so robustly social that it even facilitates a unique brand of play. Users will create hilariously specific collections based around some odd detail shared by a handful of movies, like “Movies that were released in 2021 in which Benedict Cumberbatch puts on an American accent and has a Strange relationship with a boy in his late teens named Peter who has a dead dad, telling the boy at some point in the film not to call him ‘sir.’”There are, in fact, two films in the collection.

Technology

Some kind of database of content is always core to discovery logic. In the case of file-sharing communities, these would be actual digital versions of the media. But in all other cases, that content is some sort of sorted metadata. Letterboxd is built around data from an open-source alternative to IMDB called The Movie Database, or TMDB. Goodreads is built around a proprietary corpus of data aggregated from publishers, online bookstores like Amazon, and the Library of Congress.

The other key component of this is a robust user profile, which often includes a catalog of all the media that that user is willing to share and reviews of that media. Users often communicate with each other through comments on specific items or collections of media.

Metadata is uniquely important to discovery logic, since so much of the social interaction on these platforms is built around organizing data. Typically, specific works get a single entry, and metadata are treated like tags.

For example, on Letterboxd, the film “Chunking Express” gets its own page that includes links to pages for its directors Wong Kar-wai, the year it came out (1994), its user-identified genres and themes, every notable member of its cast and crew, and so on. This is crucial for the social interaction on the platform, as it lets users do things like create a list of all romance movies from 1994 set in Hong Kong. Then, in the discussion on that collection, they can dissect which movies qualify, debate which is actually the best movie in the collection, and recommend, say, a romance movie set in Hong Kong from 1995.

Revenue model

Discovery platforms are often funded by a mix of paid advertising and user subscriptions. In the case of private torrent trackers, donations are often solicited in bitcoin to help support operating costs.

Governance

Centralized moderation remains the model of discovery platforms, likely descending from traditions of message board moderators. Discussion is inherently constrained to the relevant topic for that platform, providing some kind of guardrails for discourse. Governance effectively sets the bounds of that conversation. For example, on Goodreads there are clear rules against hate speech and bigotry, as well as established norms that keep conversations focused on the books themselves. Hence, on the site entry for Mein Kampf you find neither antisemitic rants nor essays denouncing its author. Even so, as with any moderation rules, there is still a fair amount of trolling and line-toeing that barely fit within the bounds of said governance and norms.

Seemingly conflicting ideologies are embedded in discovery logic. On one hand, there is classical cyberlibertarianism that dates back to the WELL, taking the principle of individual liberty and extending it to information, boiled down to the idea that information wants to be free and should move unchecked. Such is the raison d’être not just behind ambitious cataloging projects, but What.cd’s file distribution and Erowid’s imperative to expand knowledge and know-how about illicit substances. The very term “cyberlibertarianism” was coined by Grateful Dead lyricist and early WELL member John Perry Barlow.

Yet discovery platforms are also consummately voluntarist, communitarian projects. Together, users are creating vast corpora of knowledge, making a huge amount of movies, books, music, anime, or even restaurants more accessible through review and categorization. The thing that elevates Letterboxd from simply being a repackaging of TMDB metadata is the fact that it exists as a social space where users, together, create context for artistic work.

This is where the aforementioned gift logic comes into play: discovery platforms are powered by passionate affinity groups that rally around work they want to support, in turn supporting each other’s passion. Much like with the Okayplayer example, such support can extend to creators too. What results is the “cycle of gifting” that Feisler sketches out in her piece, where participants reward each other for meaningfully taking part. That cycle makes discovery logic transformative, as the incentive structure does not simply reward engagement but creates a safe place for participants to try new ways of expressing themselves.