Continuing the discussion from Will Snapchat overtake Facebook Messenger?:
Flat Threads Are Great When the Posts Are All Worth Reading
Something unfortunate about how Discourse and other legacy forum software are designed is that they actively suppress branching, in favour of linear discussions where one thread corresponds to one extended conversation. It works exceptionally well in some rare cases where people want to read the entire thread from start to end (e.g. tech support or other sequential inquiry resembling email conversations; our use case on this forum, which has people writing lengthy, often well-cited, and low-noise mini essays; instances with very few users, typically used for project management). Discourse creator Jeff Atwood (2012) made what I think is one of the worst calls ever when it comes to community site design (mainly because of how much better it could have been):
A part of me says this is software Darwinism in action: threaded discussion is ultimately too complex to survive on the public Internet .
Atwood's perspective was informed by his experience as mostly a lurker on Reddit and HackerNews (two of the best threaded forums available at the time), Stack Exchange co-founder, and blogger. It especially makes sense for a blogger to prefer flat over hierarchical commenting, since top level comments that directly respond to the blogpost will be most likely to be immediately on topic. Likewise for Stack Exchange answers.
But that is actually the strength of branching: hierarchies enable the development of localised, makeshift contexts, within which participants may deviate from the initial "ruleset" established by the OP (which derives much of its power from being the default object to which the participants respond to). Going a level deeper within a hierarchy is like a softer version of creating a new thread, making a group chat, or a new site/platform for that matter (and these are in order of "soft" to "hard" forms of branching/forking).
In the Long Run, the Spiders Always Win
I think that in the long run, the spiders always win: the principal challenge of cyberculture is the superabundance of content and information, never its lack. The challenge is not to produce an acceptable quality of content on which the users can subsist, but to sort and deliver the most specifically valuable content to the users that want it the most. Self-organisation and production is an exceptionally effective way to solve this problem. Each unit of content being shorter, smaller, and denser means that such self-organisation can be more granular and specific, which in turn means that content can be better sorted and discovered.
So it seems to me that, in fact, chats are like spiderwebs and blogs are like roundtables: TikTok is a spiderweb and YouTube is closer to a roundtable; hierarchical threads on forums are closer to spiderwebs but never quite get there, and blogs can only be roundtables unless they are community blogs like LessWrong, which then act like forums (the software is a heavily modified Reddit fork).
The way that Messenger integrates into Facebook is powerful because there's an abundance of users on Facebook ready to self-organise. Typical group chats on Messenger are different from "Rooms" that are tied to Facebook groups, which are Facebook's counterpart to Reddit's community chats (deprecated as of 2020). I find that a mixed bag of a group is often a tolerable experience whereas a mixed bag of a chat is a horrible experience; it's much harder to skip the bad stuff to find the good stuff in chats, because everything is flat.
A Paradoxical Ideal: Hierarchical, Branching, Granular Feeds That Lead To Flat, Low Noise, Porous Silos
Hierarchies facilitate the creation of makeshift contexts which enable splintering without full commitment. Hierarchical conversations are highly effective at producing new foci around which different conversations can emerge. They allow people to ignore noise and participate selectively and locally. However, each interaction is ultimately flat: a "reply" is just pointing one post at another (there's an interesting case to consider here: imagine a forum software that lets people post multiple, branching replies to a single post, and always invites further branching. This would be a kind of many-to-one interaction that's not quite flat).
This is good, because the flatter the environment, the easier it is to read everything, and the harder it is to splinter off unnecessarily. After a certain level of focus is reached and the noise (mainly based on selecting out people who aren't interested/interesting enough), it makes sense to move from a hierarchical mode to a flat mode. Typically, people either finish off their discussions within the thread or create a new thread/group/chat.
That is to say, people start off with webs or trees and only commit to circles and lines when they're satisfied with the signal to noise ratio. This makes sense because everybody's time and energy are limited: making someone spend their time and energy on anything that isn't worth the tradeoff between their attentional resources and the interaction is a hostile act.
A good platform therefore enables the rapid, low-cost, experimental production of content and contexts whilst minimising unavoidable noise; they promote the development of flatter, increasingly low-noise environments which have branched out from the hierarchical zones, whilst balancing attrition (brain drain from the "spiderweb" to the "circles") and circlejerks. Filter bubbles are a special case of a misaligned circle, in that they are in principle a part of the "spiderweb" but act like "roundtables". They are bad because, unlike splinter groups, they weren't formed by participants of a broader topic with a deeper interest. They're made up of people who are likely to generally agree but fail to provide each other novel and unusual content or information (because they tend to be siloed).
The paradox is that this "good platform" has people starting off in the spiderweb but constantly moves them into roundtables. Presumably the spiderweb must be replenished of the users it has lost to the circles: sometimes people return to the spiderweb and repeat the process of branching and circling; sometimes the roundtables grow into spiderwebs as more and more people join them. Large, successful platforms have loops of this sort embedded into their site designs in one way or another.
Memes are so powerful because they enable users to easily disrupt ordinary hierarchies enforced by a platform's design. They can act as anchors or beacons around which a makeshift context may be formed quickly and effectively. A meme can establish localised environments within which different rules of engagement are adopted by the participants without elaborate coordination. This way of describing competition over social influence and cultural capital makes memes sound like a site of "fourth generation warfare", but I think the mechanics of online platforms explain memes much more clearly.