On algorithms
Over the weekend, a rumour posted to Buzzfeed prompted an avalanche of angst amongst users of the world’s least understood social networking site. Twitter, so the rumour went, was intending to roll out a change to how users see tweets from the accounts they follow. No longer would the timeline of posts be in chronological order with the newest at the top; instead, an algorithm will decide the most relevant posts to show first, much like how Facebook and LinkedIn handle the mass of information a user might see when they access the service.
Would Twitter ever become algorithmic? CEO and creator of the platform Jack Dorsey posted later in the weekend that they had no intention of getting rid of the timeline but highlighted that, in places, Twitter is already using algorithms to help show the right information to users at the right time. The trends list analyses what key words and phrases are being mentioned more often than usual, an algorithm looking for the unusual and bringing it to wider attention. Another algorithm looks at who the people a user follows might be engaging with and suggests new and interesting accounts that a user might want to follow. And over the past few months a prompt welcomes users who haven't visited for a period of time suggesting a handful of tweets that they might have missed while they were away.
Signal to noise
Arguably, there's a need for more algorithmic functionality in Twitter. The platform's pushed users to follow hundreds, even thousands of other accounts. If a user follows a thousand accounts and each one of those accounts tweets twice a day, that's an average of 83 tweets per hour that a user might see without even considering that there will be times when users will be more active. For many users, it's therefore become almost impossible to use Twitter meaningfully, to engage or be engaged by those people they've chosen to follow.
A well designed algorithm would undoubtedly help get over this problem, dispelling the noise and only bringing the best posts to light. Better posts would appear in front of users more often, therefore users would have a better user experience and be much more satisfied with the platform.
But this glosses over a massive point - what is the 'best' content on Twitter?
Finding meaning for everyone
Consider how Facebook uses algorithms to display its 'best' content first. On that platform, relevance is fairly easy to understand: a user will be connected with friends and relations who'll post things that will be more relevant to that user than someone distantly connected. Content can be relatively easily judged to be controversial or engaging or even satisfying by the numbers of views, comments, likes, or clicks it has had. The algorithm isn't simple - as a recent article by Slate makes clear, it consists of hundreds of variables - but it's easy to conceptualise. New and interesting stuff from my friends and relations, please.
Twitter is a different platform. It doesn't largely consist of people using their full name with comprehensive personal information. Users may have more than one account for different purposes. It's more anonymous, but it can be personal if desired. It means that different users approach it with different wants, needs, and actions. Some will follow thousands of accounts, others will follow a few dozen. Some will tweet once or twice a week about things they are doing and nothing more, whereas others will tweet hundreds of times a day in conversation with other users.
What would an algorithm do for these very different types of users? How could one all-encompassing routine sort through thousands of tweets to show all these different groups the 'best' content when they all use Twitter so differently?
The problem with algorithms
Algorithms are not bad per se. They're essential for controlling data flow and insight, and they're excellent at finding new information and correlations. Most digital platforms are algorithmic at heart, either on the user side like on Facebook's timeline, or in the background, like Uber's ability to push drivers to a certain area where more users will be expected to request pickups.
On the user side, algorithms are a filter, a barrier between the user and the data which is often difficult to understand and impossible to adjust. It pushes what is considered to be relevant data to the top, but that means that data considered to be irrelevant is suppressed or omitted. The user cannot see what isn't shown to them. It is a subtle system of control.
Worse than omitted data is data which is forced through the system, where some users exploit the algorithms to make their content more visible. Though such companies and individuals who game the system are often caught and the algorithms changed to prevent such problems happening again, they are always susceptible to attack. Is what's at the top of a Google search or a Facebook feed really the best piece of content, or is it just something that's been designed to appeal to that platform's equation for 'best'?
The blessing and the curse of the timeline
The true joy of Twitter has always been that it offers a real-time, unfiltered view of the world the user wishes to see. What used effectively, it's a valuable and engaging tool that gives someone the ability to better understand their world.
But, undoubtedly, that unfiltered view does need filtering sometimes.
The platform has pushed users to follow as many accounts as possible, resulting in a crappy experience where users see an endlessly noisy and messy timeline that's overwhelming to read and frustrating to navigate. The company knows this, but it can't tell users to follow fewer accounts.
The obvious solution is to develop an algorithm to omit the noise, and there's undoubtedly a team working to expand the 'while you were away' feature to better score every tweet for relevance, importance, engagement, and so forth. Who knows, they may be able to create an algorithm that works for every type of user.
But it'd be the wrong fix entirely, because algorithmic sorting would lose that real-time flow of information.
What should Twitter do?
There's no easy solution here.
Twitter's struggled with this for some time. Five years ago they introduced the ability for users to create lists of profiles, so that an individual could see the tweets from a select group of Twitter users. It never really took off, and though the feature still remains it's so hidden as to have been removed. The way that conversations now appear with blue-lines to show a thread was an innovation that was meant to help, but it's just made the timeline difficult to follow.
A better solution would be filter-based but at the user's discretion. Allowing someone to load up their timeline and then restrict the timeline just to show the tweets from their friends, or from those they often converse with, or the comedians they follow, would be very beneficial. But that requires users to define their relationships with everyone they follow, no easy task when they follow thousands. It'd be easy to miss important information too - what if something important happened amongst my friends while my timeline only showed the celebrities I follow? Arguably, the user experience would be worse too, as that simple and engaging single timeline would be lost.
An algorithmic timeline would be terrible. A filtered timeline would be problematic. The way that users interact with Twitter needs to change, but any alternative seems worse than what's currently available.
Maybe the solution isn't to improve Twitter, but to consider a new platform entirely.