Apr 7, 2022 1:11 PM

This Is How Twitter’s Edit Button Can Actually Work

The ability to edit tweets is coming—but it’s full of risks. This is how to do it right.

In June 2021, Twitter told the world “you don’t need an edit button, you just need to forgive yourself.” Twitter founder Jack Dorsey even held firm against Kim Kardashian’s pleas when she cornered him at Kanye West’s birthday party in 2018. For years, the platform has held out against editing tweets. Until now. An edit button is imminent—but tricky questions remain about how it can be implemented without causing chaos.

“Everybody thinks it’s very easy to just put in an edit button,” says Christina Wodtke, a lecturer in computer science at Stanford University. Wodtke, who has worked on product design projects at LinkedIn, MySpace, Zynga, and Yahoo, argues that such a seemingly simple change will require a lot of careful thought. She puts forward a hypothetical situation: Donald Trump—whose return to Twitter some fear may be more likely since Elon Musk’s ascension to Twitter’s board—tweets something shocking or offensive. He subsequently edits his post to blunt its rough edges. But people have already responded to the content of the initial tweet, making their reactions nonsensical.

The obvious solution to that is a Slack- or Facebook-like change log of edits, where people can view the history of changes on a post. Facebook has let people edit posts since June 2012—but it's a feature that has regularly been abused by scammers since its rollout. Alex Stamos, a former chief security officer at Facebook and now an adjunct professor at Stanford, noted that Facebook’s post-editing tools helped legitimize a cryptocurrency scam page to con users. Editing pages is a core feature of Wikipedia, but that leads to “edit wars” where individuals argue about the wording of an entry, including an 11-year battle over the origins of the Caesar salad. Similar third-party tools exist for Twitter bios, such as Spoonbill, which can track how a person’s profile has been amended over time.

Yet such tracking comes with its own problems, says Wodtke. For one, the user who edited the post probably doesn’t want the original text to be accessible. “There’s all this complexity going on around how all the players in the system are going to react to this change,” she says. “You have to think through all these norms you’re now violating and changing.” Simply put, you need to design any new feature of this nature with the worst-case scenario in mind. Even if the majority of people use an edit button to nix typos, if a small minority use it for nefarious purposes it could cause chaos. “A prominent fear is that it just leads to more confusion and exhaustion on Twitter,” she says.

In an attempt to work through the problem, Twitter will start by testing an edit feature among users of Twitter Blue, its paid subscription service, in the coming months. Editing tweets has been Twitter’s most requested feature “for many years,” claimed Jay Sullivan, the platform’s head of consumer product. Twitter has also said that development of the feature has been underway since 2021—debunking any claims that a poll by Musk, asking users whether they wanted an edit button, was behind the decision.

The edit button announcement was welcomed by many—but raised concerns amongst others. Sullivan admits that ensuring the edit feature is used honestly may require “time limits, controls, and transparency about what has been edited.” So how do you code for honesty? Simply put, the way Twitter designs, tests, and implements the edit feature will determine its success—and could make or break the platform. “Are there risks?” asks Christopher Bouzy, founder of Bot Sentinel, a service that tracks inauthentic behavior on Twitter. “Absolutely. It could change the context of a tweet.” Disinformation and misinformation—the former deliberately sharing incorrect information, the latter accidentally doing so—are not exactly in short supply on Twitter, and the platform’s viral dynamics mean that some posters are loath to amend incorrect information. One 2018 academic paper found that fake news travels six times faster than the truth on Twitter, in large part because falsehoods are 70 percent more likely to be retweeted than fact-based posts.

Not all users are driven by the dopamine hit of engagement, however—and some, like Yevgeniy Golovchenko, who studies disinformation at the University of Copenhagen, believe an edit button has the potential for good. “The big question is whether it will reduce misinformation,” he says. He’s uncertain about the answer but suggests that editing could give people a low-stakes way to save face when confronted with the fact that they’re wrong. Bouzy believes that Twitter should implement the edit feature in a way that discourages or limits a high volume of amended tweets. Changing the spelling of “live” to “love” when you tweet “I love Taylor Swift” would be acceptable; completely altering the context of a post wouldn’t.

“It really boils down to how they’re going to implement it,” says Golovchenko, suggesting that contextual AI systems would be able to understand the context of a tweet and intervene if it changes too much. Golovchenko adds that if it were up to him, Twitter would never get an edit button. He equates Twitter to emails, which can’t be edited after they’re sent—but which can be “undone” using opt-in features in clients like Gmail. A similar undo feature is already available for paid-up members of Twitter Blue. So rather than launching a full-on edit button, Twitter might consider limiting the amount of time a user has to edit a tweet after hitting send. That would help get rid of typos while also not encouraging nefarious behavior.

Many researchers specializing in disinformation share the concern that an edit feature may be used by bad-faith actors. “I’m falling on the ‘seems like a bad idea right now’ side of things, at least on surface value,” says Sarita Schoenebeck, who is director of the Living Online Lab at the University of Michigan, which studies human-computer interaction, and is currently on sabbatical as a paid consultant on Twitter’s Birdwatch community fact-checking team.

But it’s not just the potential for abuse that makes the edit function questionable. Schoenebeck is also concerned that it will cause confusion among good-faith users. “I don’t think, from an interface design perspective, it would make it clear what the trajectory of a tweet was,” she says. Essentially, if it is poorly implemented it would be hard to tell whether a tweet was edited before or after going viral. Schoenebeck suggests that implementing the feature could be a nightmare for archivists, a critical issue as Twitter has become a place where the first draft of history is written.

And rewriting that history gives Schoenebeck pause. “Instead of just an edit button for a tweet, they might think more about which tweets should be public and which ones should have more of an archive label to them,” she says. Rather than editing content, she suggests, Twitter could offer people a compromise between leaving a tweet public and deleting it. “Maybe it’s not amplified or not pushed into the news feed,” she says. “It’s still there, and still findable, but someone is indicating: ‘I no longer support this, or I retract this, or I’m sorry for this.’” That seems a more usable, practicable way of avoiding some of the issues that a proposed edit button would cause.

Such concerns are hopefully what’s driving Twitter’s limited trial of the edit button feature—and may mean that any final version has very restricted functionality. “Twitter is basically a sociotechnical system,” says Wodtke, meaning that any technological change to Twitter’s software or functionality will have knock-on effects on the way people interact with it, and with each other. It also probably explains in large part why Twitter has held off on experimenting with the feature for so long and will, hopefully, tread carefully in the coming weeks and months. “It’s hard to get the will up to step into that amount of complexity,” says Wodtke.