Much of the appeal of Twitter in its early days was what the community made of it. @mentions, #hashtags, and even a bird as the avatar of Twitter—all came from community members and outside developers. Third-party clients reached much deeper than Twitter’s own Web app and basic platform apps. Software makers offered alternative interfaces, cleverly constructed bots tweeted information from automated systems, researchers extracted insights from trawling the data, and more.
But as Twitter focused on increasing engagement in the service of making money—and as the company flails wildly in the insanity era of Elon Musk—many of the unusual uses of Twitter have fallen away. (The most recent Twitter implosion entry: users posting just one or two tweets being told erroneously, “You are over the daily limit for sending Tweets,” an error previously reserved for posting what seems like a thoroughly excessive 2400 tweets per day.)
The rise of Mastodon, with its open-source, federated nature, has gotten me thinking (see “Mastodon: A New Hope for Social Networking,” 27 January 2023). If developers can break free from thinking of Twitter as the apex of microblogging, we could see tools that would radically change and improve our ways of interacting. Here are a few ideas that have occurred to me; share yours in the comments below and on Mastodon, where we can hope they’ll catch the eye of developers.
Interact with Mastodon More Like Slack or Discord
I can’t take credit for this one. I saw the basic idea roll through from a developer when I was first playing with Mastodon, but now I can’t seem to construct a search that finds that post. (Search is effectively local and unfederated.) My version of the idea is that, instead of a firehose timeline where every post appears in chronological order, a Mastodon client would present posts in a hierarchical, columnar fashion, much like messages appear in Slack or Discord channels. (Columnar Mastodon clients already exist, but they merely devote a different column to different timelines: your feed, notifications, messages, and profile, say.)
The first column would contain a list of people you follow, sorted alphabetically, with an option to pin favorites at the top and another option to show only those people who had new posts. The second column would naturally show posts from the selected person, and the third column would contain replies to the selected post. Below those three columns, an optional content pane would show the first linked Web page from the selected post; if it had multiple URLs, clicking another one would replace the previous page. (Bonus points for anyone who wants to create a mockup of this app interface.)
The big win of this technique would be that you could more easily pick and choose who you wanted to read at any given time, just as you can focus on particular people or channels in Slack, Discord, and other chat environments. Providing a Web view of linked content could also speed up reading by eliminating the need for a trip out to your browser for every link.
Limit Reading Time
Social media is an endless scroll. That’s one of my main complaints about it. There’s no way to “finish” reading unless you follow only a handful of people who post sparingly. Even TidBITS Talk is finite, despite its roughly 1400 posts per month. I read every post there, and if I skip a day or two, I can catch up easily. I can devote significant time to TidBITS Talk because it’s a key part of the overall TidBITS publishing model and thus is part of my work day. But I don’t have time or attention to spare for random content from social media, even if some of it comes from friends or other fascinating people. (Family doesn’t intersect with social media for me because we’ve successfully centralized all family communications in Slack, which is also finite; see “Fed Up with Facebook? Move Your Family to Slack,” 12 February 2019.)
I’d like to see a Mastodon client designed with “finishability” in mind. You could set a certain amount of time per day that you wanted to read Mastodon, and the app would manage that. It might prioritize what you see by preferring posts from certain people or that are connected to messages you replied to, favorited, or boosted. Then you would move on with your day without feeling like you were missing out because the next post might be life-changing (hint: it never is). Screen Time might seem like a solution, but I find it frustrating because it’s an arbitrary cut-off after X minutes, and I know more posts are waiting.
Instead, my hypothetical Mastodon client could estimate how long it would take to read a post, make some assumptions about whether you’d also read replies, build in some time for clicking out to external pages, and then display a selection of posts. The technology for this shouldn’t be difficult. Many blogs and publications now display a reading time estimate so you can get a sense of whether you would be committing to 4 minutes or 25 minutes before you start. It could also observe your reading speed and adjust accordingly. The trick is that when you say that you want to read 10 minutes of Mastodon, the client would have to make it seem like you had actually finished at the end. A nice “You’re done!” animation would be welcome.
In the meantime, you can simulate a finite Mastodon by tuning lists or using hashtags. In both cases, filtering by a set of people or a topic dramatically reduces what appears. These stop-gap measures for “finishing” work for now because fewer people post on Mastodon and tend to post less frequently.
Could Local Content-Selection Algorithms Serve Users?
Unlike Facebook and Twitter, which employ algorithms that choose which posts you see based on what they think will cause you to keep reading and replying, Mastodon apps generally present posts solely in chronological order, with the newest first. It’s sad that a simple chronological list is a refreshing change from the manipulation of social media algorithms, but algorithms don’t have to be evil. The desire to bring the best and most interesting posts to the fore isn’t problematic on its own; the concern is that these black box algorithms are designed to serve the needs of their corporate masters, not users.
Without a profit-driven company at the heart of Mastodon, I’d suggest that the time is right for Mastodon clients to revisit content-selection algorithms, but with two key differences.
- The algorithm should be designed to serve the user, and only the user, in ways specified explicitly or implicitly by the user. (The lack of a corporate overlord also means there’s finally room to revisit the discussions surrounding Asimov’s Three Laws of Robotics in the context of software agents.)
- The algorithm’s rules should be at least visible to the user, and preferably actively editable. It shouldn’t be difficult to specify people or topics that interest you more than others.
Tasking a client-based algorithm with what posts to show would also increase the utility of a feature that limited reading time because the algorithm could select only those posts most likely to be of interest. Mastodon servers are unlikely ever to adopt an API-accessible algorithmic feed, so we’re reliant on app developers experimenting in this area.
Add Categorization to the Mix
Adding categories to posts would let you see discussions that are topically related, something that’s not possible with social media now. Let’s face it: Twitter and Mastodon both suffer from an extremely short attention span when you look solely at the timeline. Posts jump from topic to topic haphazardly.
Chronological ordering works well on a service like Strava, where every post is a friend’s workout, but it’s just schizophrenic with quickly updating microblogs. Your feed might include someone linking to a long read about an important but depressing topic, others heatedly discussing last night’s sportsball game, a comedian trying out new material, and extreme weather warnings from a bot. They’re all interspersed with one another, forcing you to make mental context jumps every 20 seconds. That can’t be good for our brains.
The popularity of hashtags and trending topics on Twitter shows the extent to which people like to stay on one subject for a while before moving to another. I’ll admit that whenever I do end up loading the Twitter page, I can’t avoid looking at the What’s Happening box. I’m very seldom interested in anything there, but on the rare occasions I click through, I appreciate that all the tweets are inherently related.
We’re treading into the world of AI here, but if an algorithm could automatically categorize Mastodon posts, it would enable users to focus on specific topics in the Mastodon content stream. With a Slack-like Mastodon client, categories could appear alongside people, much like Slack channels mix with direct messages. You could pin favorite categories, but others would come and go as associated content ebbs and flows. Because machine learning benefits from training, an app that started with loose categories and let you provide feedback could rapidly produce worthwhile results.
It’s still early days, and most Mastodon clients I’ve seen are nearly identical. I get that developers are still polishing their interfaces and figuring out how to handle Mastodon’s hypergrowth spurt. But I hope that, once developers have nailed the basics of this bigger ecosystem, they can turn their attention to taking advantage of the openness of Mastodon to innovate rather than merely copying Twitter clients.
What radically new approaches would you like to see in a Mastodon app?