Jun 30, 2015 ยท 10 minutes

The robots are taking over.

Taking over Twitter, that is, which over the past few months has been employing algorithms to sort tweeted content.

This content until recently was presented in a raw reverse-chronological feed. If you followed the right people, the result was a uniquely chaotic stream of insight, humor, and information that was unlike anything the world had seen before or sense.

If you didn’t? Chances are, your feed would be full of mundane celebrity observations, "breaking news" about what your cubicle-mate had your breakfast, and everything else that’s given Twitter a bad name among the masses. 

For long-time users, this is less a bug than a feature. The platform’s difficult learning curve, which requires new users to build both their following and followers lists slowly and to prune the dull edges as they go along, tends to scare off new would-be tweeters and prevent them joining the power users’ “cool kids’ club.” The overall growth rates of monthly active users bears this out: Twitter has been adding around 50 million users a year for the past two years with no sign of speeding up. At that dismal rate, Twitter would have the same number of users that Facebook has today by… 2043.

Unsurprisingly, Wall Street is less enthused about this growth rate than power users are, and this was likely a major factor in CEO Dick Costolo stepping down earlier this month. A feed driven by algorithms -- like Facebook’s is -- makes for a much more welcoming experience for new users, which to date is still Twitter’s biggest challenge. As Nate Elliott, vice president and principal analyst at Forrester Research, told BuzzFeed News, “It’s not an easy task, but I think it’s a necessary task. If they implement an algorithm properly, it could be the savior of their platform.”

Already the company’s been experimenting with algorithmically-driven feeds, although they’ve been largely limited to the company’s new homepage which only appears for non-logged-in users. The changes that long-time users see are a bit more subtle – you might see a tweet from somebody you don’t follow that was favorited by somebody you do follow. And when a long conversation appears in your feed, Twitter will now provide the tweet that kicked off the conversation in place of the most recent tweet in the group.

The changes are both too much and not enough. Power users are incensed by any meddling with their raw feeds. “The beginning of the end,” NYU professor Jay Rosen called it. A ‘fuck you’ to users, said another. I was one of many who worried that, if Twitter became too much like Facebook, stories like the Michael Brown shooting which matter intensely to an (initially) small number of users would never have broken through to the mainstream, drowned out by so many Ice Bucket Challenge videos.

Google has encountered a similar problem over the years. When an algorithm is designed to surface the most popular content, smaller – let’s call them “pre-viral” – stories may never hit critical mass. Organic, viral growth is sacrificed and the only things that do go viral are pre-ordained by the biggest players in the content creation game who possess the money or built-in audiences to basically force virality onto the masses. 

Meanwhile, the changes are such half-measures that they haven’t even begun to convert higher numbers of monthly active users. Chris Sacca, a big investor in Twitter, estimates that almost one billion people have tried and abandoned the service. 

Like other long-time users -- particularly in media -- I enjoy Twitter the way it is. Yes, it took work to build the community of newsgatherers, comedians, philosophers, and potheads that populate my feed -- and those relationships and connections are deeper because of the work I put in. But the truth is, Twitter needs to evolve more quickly – just as Facebook and newer upstarts like Snapchat and Instagram have – in order to remain relevant beyond its super-fans. And one way – though not necessarily the only way – is to go further and faster with algorithmic feeds than it has in the past.

Does that mean Twitter’s fated to become just like Facebook? Will the random chaos that made it great in the first place be reined in by robot censors?

While doomsayers are still calling this a tragedy for the service they used to love, I look at it as an opportunity – an opportunity from the start to build algorithmic feeds right and to avoid some of the more negative aspects of Facebook’s algorithms that have made many publishers and users of a certain intellectual stripe so hostile to the service. 

Here are three ways Twitter can use algorithms to make its service easier for new users without saying “fuck you” to everyone else who helped build it into the service it is today.

1.Let users toggle between raw and algorithmic feeds

At any time, users should be able to see a raw stream of reverse-chronological tweets that has not been corrupted by algorithms. I’ve often wished that Facebook would adopt this feature, letting me see what the company's robots don’t think I want to see. I totally understand why Facebook doesn’t offer it – not only would it threaten to reveal the “secret sauce” of the algorithm, it would also inspire no end of conspiracy theories – notions like, “Why did Facebook downplay this story on nationwide gun death statistics but put a speech by the NRA’s Wayne LaPierre at the top of my feed?”

Twitter would encounter the same problems, but because it’s waited so long to implement algorithmic feeds, it doesn’t have the luxury of slowly easing users in. If it’s going to heavily overall our feeds, sorting and resorting the content, surfacing things it think we’ll like, and adding tweets from people we don’t follow, it needs to do it fast – which means it can’t kill off the raw feed immediately. Users would revolt. I would revolt. 

So let users toggle between them. Twitter might even win over some of the pro-raw feed cranks. I think I do a good job of self-curating the people I follow, but there’s still a ton of tiresome, asinine content I see. If Twitter can find a way to purge this content without throwing the baby out with it, I might become a convert.

Again, letting users pick between raw and, shall we say, “processed” feeds would threaten to reveal the innerworkings of the algorithms to readers and competitors. But that’s not necessarily a bad thing, which leads me to the next point…

2. Be transparent

Facebook’s algorithm is among the most powerful forces on the web, dictating what content users do and don’t see every day.'

It’s also astoundingly mysterious. Even Facebook’s own researchers don’t seem to know precisely how it works; how the sum of all those tweaks and tinkerings over the years manifest themselves to users. That’s why these researchers spend so much time studying the algoritm.

Facebook uses this mystery to its advantage, however. Whenever somebody complains, for example, that there are too many puppy videos and not enough substantive journalism in a feed, Facebook blames it on the algorithm – and by extension, on the user who provides the algorithm data through the content she likes, comments on, and shares. Don’t want to see puppy videos? Stop clicking "like" on puppy videos, Facebook tacitly suggests.

This is a deeply disingenuous line of defense. Facebook alters its algorithm constantly to achieve specific outcomes, regardless of whether users demanded a change or not. And while Facebook often communicates these algorithm tweaks to users, I think it’s incredibly unlikely to say the company communicates all of them.

While total transparency is a lot to expect from any company, Facebook could do a lot more to communicate to users how its algorithm works – and this is where Twitter can really differentiate itself.

Sure, Twitter could offer a source code for the algorithms or a bunch of wonky documentation that few will read and even fewer will understand. But here's a better option: Why not put the control in the hands of the user? Yes, the algorithm will learn from and adjust to the content each user retweets, favorites, or whatever. But beyond that, Twitter should let users specifically highlight topics they’d like to see more of.

For example, when I pulled up Facebook last summer and saw a deluge of Ice Bucket Challenge videos and not a single story on the Michael Brown shooting, my only recourse was to “dismiss” every Ice Bucket Challenge video – and hope that I didn’t lose that person from my feed forever – and track down a Michael Brown story to like. And then maybe my feed would be just a little more on the side of social justice and less on the side of “look at me, look at me, I’m giving to charity!”

But there was never a guarantee. Twitter should offer much more granular algorithm controls that operate on a tweet level and a broader user settings level. Just as users self-curates the people they follow, each user could essentially write their own algorithm. Again, this takes work. And you might fuck it up. But you can always revert to what Twitter thinks you’ll like based on your day-to-day actions.

What always made Twitter great was the amount of control each user could exert on the chaos inherent to Twitter. Dynamic algorithms that operate on both explicit and implicit cues from users – not to mention the first feature mentioned here pertaining to keeping the raw feed – would preserve that sense of user control.

And while that means Twitter itself would have less control – and therefore less power to serve up content that meets certain outcomes that are favorable to the company – these features would certainly endear the company to publishers, which are going to continue to be a big source of revenue for these content platforms. And sure, Twitter can still put ads and promoted tweets wherever it wants. We’ve accepted them -- like I imagine we will the algorithms -- as a necessary evil.

3. Don’t be a prude

One of the biggest ways that Facebook’s algorithms affect what we see and what we don’t pertains to risqué content. Zuckerberg's little robots work hard to excise the platform of just about anything that wouldn’t pass muster in a movie rated PG-13. Nipples, either real or drawn by a New Yorker cartoonist, are a definite no-no. Nudity that surrounds something historically or culturally important is another thing you can say goodbye to. Last week, Nathaniel Mott wrote that if the famous Pulitzer-winning image of Kim Phuc running naked to escape a napalm attack was posted to Facebook today, the company’s algorithms would make short work of it and maybe nobody would see it. 

But as I’ve written in the past, if Facebook is bland, safe network television, then Twitter is HBO. Pretty much anything goes there, for better or worse -- there’s still a lot of abuse and harassment that takes place on the service.

But no matter what Twitter does, it shouldn’t look at algorithms as an opportunity to rid the platform of nudity or other elements of the risqué. I doubt that it would, but it bears mentioning, anyway.