In early April, Elon Musk broke the internet when he polled his Twitter followers: “Do you want an edit button?” Some 4 million votes later, nearly 75% of respondents said yes. And Twitter, at long last, confirmed its existing plan to ship the feature.

The decision is superbly controversial. Some critics argue an edit button will disrupt democracy, giving malicious users like former President Donald Trump (who is now banned from the platform) the option to rewrite history. On the flip side, many of the same people who are arguing for the “immutable ledger” of cryptocurrency are eagerly backing Musk’s desire to make tweets editable (an irony I am hardly the first to point out). But we shouldn’t allow this mini culture war to distract us from a real design problem: People accidentally screw up tweets every day. We make typos. We misattribute quotes. We simply mess up somehow, publicly, and the only real fix is to delete the tweet.

That is a solvable problem, according to Michael Leggett, a designer at the solar startup Evergreen. In a past life, he was responsible for building “Undo Send” into Gmail and helped design how users could delete their chats in Facebook Messenger. “You’re trying to maximize good and minimize harm,” says Leggett of designing these tools. “And I think there’s got to be a way to minimize harm enough to make [an edit button] possible.”

THE MANY, MANY PRECEDENTS FOR EDITING TWEETS

When Leggett worked at Meta (Facebook at the time), Facebook Messenger allowed users to delete their own messages. But only for themselves. So if I deleted a message in our group chat, you would still see it there. And there was no indication to me that you could still see it. This approach was obviously problematic, and a bit pointless.

Internally, people pointed out that deleting a message completely could be bad. What if it’s an abusive message? But when a Facebook employee accidentally shared their tax forms in a Messenger chat with others at the company—forms full of personal information—it became clear that sometimes you really do need to delete a message. So Facebook tweaked deletion to work as it should (no one sees a post once it’s deleted), a choice that followed from years of debate and studying data.

“It’s never going to be perfect,” Leggett says. “With Facebook Messenger [we asked], What are we not solving?” In that case, Facebook decided that it wasn’t solving what Leggett calls “the drunk message,” explaining, “If you send this at 2 a.m., wake up, and regret it, we’re not solving it.”

In practice, that meant that a Messenger post could be deleted only if you caught the error or changed your mind within minutes, perhaps—an action you took immediately when your stomach dropped. After that, it became part of the conversation’s record forever. (Facebook has since allowed users to delete messages in perpetuity.)

On Gmail, Leggett developed the glorious Undo Send button that lets you cancel an email you’ve already sent. Technically it appears as a “toast”—a skinny banner notification that pops into the interface.

“The whole point of an Undo Send is you see it right after you sent it, and you have that visceral [sense of], ‘Oh no, I made a mistake,’” Leggett says. But the solution the Gmail team created wasn’t obvious from the get-go. They considered, for instance, an option that would let you unsend an email all the way up until someone saw it. So in theory you could unsend it days later if it were just sitting in a recipient’s inbox that entire time. But the design team realized that approach infringed upon the privacy of the recipient. A sender could see if the email had been read or not. So ultimately, the Google team decided on a simple message that explained the email had been sent and offered an undo button for a few seconds after.

Twitter tried something similar with an Undo Tweet feature, which launched last year as part of its premium Twitter Blue subscription service, which adds some UI perks to the platform. But Leggett says that Twitter’s Undo Tweet totally misses the mark. That miss comes down to micro decisions within the user interface.

Instead of saying a message is sent, Twitter says a message is “sending” and you can hit an Undo button to stop it. The problem Leggett outlines is that, at least from the data he has seen at Google, the user will sit there and wait for the message to actually say “sent.” This wastes their time, and so eventually, they hit an accompanying “send now” button to get that tweet out immediately. If you’re always hitting “send now,” you’re going to hit “send now” for the mistaken tweets, too. At that point, the tool has just defeated its own purpose. “If they got that wrong, and I’d contend they did, how well can they do with editing tweets?” Leggett asks. “It makes me nervous.” But he insists editing tweets is possible—and his peers in the design community agree.

HOW COULD WE EDIT TWEETS, RESPONSIBLY?

When you think about how a tweet might be edited, you can start by seeing what other platforms have done. Reddit does not allow post headlines to be edited after they go public—those are treated something like a printed newspaper. However, it allows you to edit comments as much as you’d like. Whether or not you see those comments have been edited varies depending whether you’re browsing on the web, where the post is marked with an asterisk, or the mobile app, where it’s currently not indicated at all. A Reddit spokesperson explains that Reddit culture takes over where design leaves off. Generally speaking, it’s considered good manners to also note your edit in your own message. And indeed, you see this practice in almost any Reddit thread you read. But fostering an online community like that is rare, and it takes years of work.

What about Slack? You can edit messages on Slack, and they appear with a small “edited” note in the UI. But can you see what changed? No. That limitation is problematic to anyone who might be harassed by a coworker, only to have those tracks erased. However, all messages are stored in the cloud, and those edits made in public channels can be downloaded by managers who have access.  (A conspiratorial take would be that Slack’s edit revisions, only accessible by members of management and IT, protect the power structures inside companies that decide to use and fund a Slack subscription rather than lower level employees who simply use it.)

As for Facebook, all of your posts are editable. The catch is that, in your main feed, you can’t tell whether something has been edited. You need to tap the three-dots on a post (a.k.a. meatball menu), to see a step-by-step revision history of the message. That’s how you know it has been edited or not.

None of these approaches to editing work for Twitter, which has become the de facto communications platform for public figures. In high-stakes moments—like when Trump tweeted praise for insurrectionists who stormed the U.S. Capitol—edited tweets need to be understood at a glance, and burying changes deep in the UI doesn’t make a lot of sense for a platform that’s designed to be skimmed.

A smart potential solution comes courtesy of Glen Murphy, a senior director of UX at Google who has worked on Android, Chrome, and Chrome OS. Murphy actually mocked up his approach to what Twitter edits could look like (in Google Docs, of all places).

This is an edited tweet. And it looks like an edited tweet, using all of the strikethrough conventions we’re accustomed to when tracking changes in documents today. Rather than hiding these fixes, Murphy suggests they can simply be aired as part of the message.

On its own, Murphy’s edited tweet format could make Twitter an absolute chore to read, sure. But Leggett points out that we don’t need to think about these UI decisions as on/off binary choices. Leggett suggests that maybe Murphy’s edited tweet appears this way for an hour. Or maybe it appears this way until a certain threshold of viewers reads it and accepts it, rather than flagging for misinformation, harassment, and so on.

To be clear, this is not the solution that Twitter is testing, per se. We reached out to Twitter for a preview of the feature and they declined to share one. In any case, many edge cases need to be tested to know what works most effectively. Paul Stamatiou, who previously spent nearly a decade designing at Twitter before landing at the Bitcoin trading platform Kraken, notes that the challenge at hand is really all about handling the unintended consequences of editing tweets.

He explains that once any tweet is edited, that edit has a downstream effect. What about the news article that embedded it? What about the person who was just verbally abused before that abuse was erased through a revision? What happens if you like a tweet that then turns abusive through an edit—are you unintentionally roped into abusive behavior?

We will see what different approaches Twitter releases later this year, through what we can expect to launch as a public beta (Twitter’s design team now releases its new features following public experimentation). Twitter has been working on this feature in pieces for a long time. Twitter’s former Chief Design Officer voiced his support for an edit button back in 2021. And based on my own reporting about Twitter, I suspect members of its design team have been creating mockups for the last decade.

But however long it’s taken the company, make no mistake: Creating an effective edit button is possible, and for most of us, it could make life better. The only real impediments to editing tweets are all built within Twitter’s own platform. And you know what is ultimately responsible for repairing Twitter? Twitter. It’s about time that we stop pretending that the design problems invented entirely by corporations aren’t also entirely solvable by those same corporations.

“Just fix it,” says Leggett.