Jack Dorsey is stepping down as CEO of Twitter, to be replaced by Parag Agrawal. It’s not clear whether this will cause much of a change in the company’s direction — Tim Cook and Dara Khosrowshahi don’t seem to have radically altered the courses charted by Steve Jobs and Travis Kalanick, for instance. And it’s an open question whether the company’s famously chaotic corporate culture is even capable of making substantive changes. But on the off chance that change could be in the offing for the platform I spend most of my time on, I thought I’d offer my thoughts.
What’s wrong with Twitter
First, the background. I’ve done a lot of complaining about Twitter over the years. This post will probably be the last in the series — at least for a very long while — so if you’re a reader who doesn’t care about Twitter, you can breathe easy. But as a prelude to saying what I think can be done to fix Twitter, it’s worth it to review what I think are the platform’s major problems. So here’s a list of what I’ve written:
1. The Shouting Class: Twitter’s openness, free entry, and virality combine to allow anyone to gain prominence over anyone else; there’s much less gatekeeping and moderation than other platforms. But this means the discourse gets dominated by people with both the time and motivation to spend all day shouting on Twitter. That includes a fair number of good people with reasonable concerns, but it also includes a huge number of clout-seeking self-aggrandizers, histrionic wailers, ideological extremists, trolls, and contentious people who just like to argue. And because journalists and politicians all have to be on Twitter for their jobs, we’ve effectively locked our nation’s thought-leaders in a room with the Shouting Class. This encourages extremism, exhausts our empathy, and makes our society more divided and contentious.
2. The Shouting Class 2: Last Refuge of Scoundrels: In this sequel post, I gather some empirical evidence that Twitter tends to attract contentious people, and how the platform tends to amplify expressions of outrage.
3. It’s not Cancel Culture, it’s Cancel Technology: Amid all the brouhaha over “cancel culture”, there has been relatively little focus on how social media changes the nature of social ostracism. The way Twitter is set up makes ostracism both more unavoidable — because you can’t choose the group of people you express your ideas to — and more long-lasting, because your tweets, or screenshots of them, get saved forever. This has made social ostracism more likely and more pernicious in some ways than in, say, the 1990s or the 2000s.
4. Twitter and gekokujo: In 1930s Japan, nationalistic young army officers would sometimes force their more moderate superior officers to give them a free hand by appealing to nationalistic sentiments that prevailed among the general population. This was called “gekokujo”, a word indicating subversion of a hierarchy of authority. In the same way, Twitter allows dissenters within any organization to take their complaints to the general public and harness free-floating outrage and unrest to cow their superiors. If you don’t think the organizers of your science fiction convention have given you sufficient recognition, for example, you can take to Twitter and denounce them, leading to a wave of outrage that terrifies them into acceding to your demands. This power is, to some extent, based on an illusion — most people aren’t used to being yelled at online, and even a dozen angry replies can generate the impression of a vast wave of popular anger. But until companies and other organizations learn that they can ride out Twitter outrages with little consequence, the platform will continue to be disruptive to organizations at every level of society.
5. Status Anxiety as a Service: The directness with which Twitter allows “reply-guys” to address “bluechecks” creates the illusion of equality on the platform — an illusion that is then constantly shattered, as the reply-guys realize that the bluechecks have a much bigger platform than they do. This breeds a peculiarly toxic social dynamic, contributing to the platform’s general air of resentment and vindictiveness.
These are not the only problems with Twitter, obviously; there’s also the fact that communication by short text-only messages is inherently attenuated, removing much of the nuance and context that makes video and audio communication so natural. But that’s just a general problem of the internet, really, and Twitter Spaces represent one attempt to address the issue.
Really, the problem with Twitter is just that it’s ruled by shouty jerks. All of the other problems — ubiquitous fear of “cancellation”, disruption of organizations, toxic status anxiety — ultimately come back to the Shouting Class.
How to make Twitter better
So how can the platform be fixed? The product team has experimented with a number of tweaks over the years — hiding some replies, tweaking the algorithm, adding features like “mute conversation”, and so on. And centralized content moderation has increased, leading to a slow squeezing out of much of the alt-right and Qanon. That has improved things incrementally. But if management really wants to improve the platform, it’s going to have to tinker with its basic nature.
The first key change should be to give people more power to moderate their own threads. Three decades of the internet have proven the value of content moderation, but as both Twitter and Facebook have discovered, this task is too enormous for centralized organizations to effectively handle; moderation has to be farmed out to users themselves. Facebook does this with groups, and Reddit does this with subreddits. For Twitter, this is a much harder task, given that the platform is not divided into sub-forums. But content can still be moderated at the level of the thread.
When I create a tweet, the replies to that tweet become a thread. That thread is a discussion forum. But as things stand, I — the person who created that discussion forum — have little ability to moderate the thread. I can hide a reply (a relatively recent feature), but that merely advertises to other readers that the reply has been hidden, inviting them to click a button and take a look at it. Other than that, there’s just not much I can do. I can block the person who made the reply, but others can still see the reply.
Twitter’s product people need to consider how frustrating this is. Suppose I tweet a popular tweet about how good vaccines are, and it draws responses by antivaxers. By making my own tweet popular, I have also given a platform to antivaxers, because many of the people who look at my tweet will scroll down and see the antivax replies. I am forced to platform bad actors as the price for offering my own ideas.
Of course, to allow people to delete the replies to their tweets, as a forum moderator can delete posts, would be unfair. But dropping the tweets of blocked users from thread view seems like a very reasonable move. After all, if I block someone, I’m preventing them from replying to my threads in the future; why should I also not be able to prevent them from replying to my threads in the past? If I consider someone to be such a bad actor that I never want to allow them to interact with me in the future, why should my interactions with them in the past be permanently preserved for all to see? They should not. So replies by blocked users should be dropped from thread view.
Next, users should be given the option to customize who is able to reply to their tweets. They can already do this a little bit. When I write a tweet, I can open this menu to choose who can reply:
Why should these be the only options? Why not add a “custom” option, where I can select a few people with whom I want to discuss something? This would allow people to have discussions in public without involving the public, which is a useful mode of discussion. And note that it’s a mode of discussion that is already the default in Twitter Spaces; there are only a few speakers, and the person who creates a Space gets to decide who is a speaker. In fact, Twitter could allow people to designate other “moderators” for their threads, who are able not just to reply but also to open up the thread to replies by specific other people! That would let people create text discussions that are more like Twitter Spaces, instead of automatic free-for-alls.
Twitter should also allow people to opt out of being quote-tweeted. The creators of the quote-tweet feature recognize that it has become the “dunk mechanism” — the means by which outrage mobs are summoned to attack other people’s tweets. The way to stop this is to allow people to opt out of having their tweet be visible in specific quote-tweets. Just add a button the side menu of a quote-tweet that says “Hide quoted tweet”, which only the writer of the original tweet can select. Also, allow people the option of allowing replies but not allowing any quote-tweets on a particular tweet. Currently, they can restrict replies, but quote-tweets are still allowed; simply allow them to do the reverse. (This won’t restrict debate, because people can simply screenshot instead of quote-tweet. It will simply short-circuit pile-ons.)
Those three changes — 1) dropping blocked users’ tweets from thread view, 2) allowing people more control over who can reply to their tweets, and 3) allowing people to opt out of being quote-tweeted — will remove much of the platform’s shoutiness and toxicity, without restricting dissent or speech on the platform overall.
In addition, Twitter should implement the dislike button — something they’ve been toying with for a long while but have never quite pulled the trigger on. The reason is twofold. First, a dislike button will allow people to register disapproval passively. Currently, if you agree with a tweet you just click the like button, while if you disagree you have to write a full reply. Hence, replies tend to be heavily weighted toward the negative, which means that if you scroll through reply-tweets, you’ll see a lot of negativity and not a lot of positivity. A tweet with 1000 likes and 17 angry vicious replies will still look like a dumpster fire, even though almost everyone liked it! A dislike button would allow those 17 angry people to register their disapproval lazily, by clicking a button, instead of actively adding to the platform’s negativity.
A dislike button would also allow Twitter to fine-tune its algorithm for ordering tweets within threads, deciding which ones to hide, and deciding which tweets to recommend to users in the first place. Currently, there’s no good quantitative way for Twitter to measure negative feedback, which robs the algorithm of key data. A dislike button would thus empower Twitter’s product team to make all sorts of other decisions.
These are all small tweaks to the platform that I think would go a long way. But there’s one huge change I’ve been thinking about, which might be worth considering. One reason Twitter was so successful in pushing the Nazi types off the platform over the past few years was the existence of Gab, an independent Twitter clone geared toward Nazi types. Obviously Gab is an evil thing, but the result has been that when Twitter began squeezing the Nazis out with intensified moderation, they had a natural place to go instead. So the net result was that America’s journalists and politicians, who are forced to use Twitter for their jobs, were exposed to fewer Nazis on a day-to-day basis.
This approach could be successfully replicated by Twitter itself. Twitter could allow people to create totally unmoderated, unregulated Twitter clones that don’t converse at all with Twitter itself. Tankies could create a Tankie Gab, Antivaxers could create an Antivax Gab, etc. But just like with Gab, tweets on those unmoderated clones couldn’t be retweeted into the real Twitter. And there could be additional barriers between the services, such as requiring people to open a separate app to view each unmoderated clone. Combined with intensified central moderation, unmoderated Twitter clones could allow Twitter to slowly push all organized groups of bad actors off of its main service. And instead of leaving the creation of these clones up to rival entrepreneurs, Twitter would retain ultimate control, so that the company could delete a clone whenever it decided to.
This last suggestion is a pretty radical one, and would be a huge change for the platform. So I don’t want to put it in the same category with the other, more modest changes suggested above. I’m not totally sure creating a bunch of unmoderated extremist hideouts is worth the price of cleansing Twitter of the worst of its shouters. It’s just an idea that’s worth thinking about.
Anyway, these are my suggestions for how to improve Twitter. Basically, the platform needs to be more about discussion and less about combat. Keeping that ultimate goal in mind will guide the product team in the right direction.
On dislikes: once long ago I worked at a company that makes a reasonably popular piece of software (>10 million users) and they were implementing Likes.
Their internal data showed just how much Dislikes turn off casual users from ever participating again. It led to a filtering of the contributors to a certain kind of brash young male engineer who is convinced that dislikes are just society failing to understand their brilliance.
Normal people, on the other hand, were strongly affected. After all, humans have millions of years of evolution where the only control was social shaming.
Since the Dislike was much easier than writing a reply -- and conveyed no information explaining Why -- they tended to accumulate faster and quicker than negative replies.
You've probably heard the social science finding about how a single negative interaction is weighted the same as seven positive interactions?
That's the kind of thing you're fighting against with Dislikes. They are a surprisingly huge speedbump for new & casual users to actively contribute.
For many of us lurkers, not posters with many followers, Twitter is okay as is. Why? We decide who to follow, and that's who we see. Generally avoid replies (unless it's a thread) and we miss most of the flame wars.