Twitter’s discovery algorithm (which surfaces heavily discussed or shared messages on people’s timelines) “prioritises a very particular type of content”, says Renée DiResta, who studies social networks and misinformation at the Stanford Internet Observatory in California.

“People who maybe don’t necessarily have an institutional credential, but are adept at commenting on a particular topic, can capture public attention,” she says.

The idea of Twitter as a great democratiser also doesn’t always match reality, DiResta adds. Accounts with a large, established following have much greater reach than “your average science experts on the platform”, she says.

And although Twitter’s algorithms elevate humour, delight and entertainment, they can also encourage performative tweets, dismissive arguments and snide comments that veer into abuse. Real-time criticism can swiftly turn ugly, and users can easily butt in on others’ conversations, with hordes of people sometimes exhorted to insult and mock a specific target.

Twitter has always struggled to cope with how to moderate such rapid shifts in online conversation. It’s a problem that seems likely to worsen now that Musk has made cuts to the company’s staff and its safety systems.

This double-edged nature of Twitter has never been clearer than during the Covid-19 pandemic. Many academics built up large public followings through their expert analyses about SARS-CoV-2, and made fruitful connections as scientists rushed to understand the pandemic.

“Twitter was a really powerful way to do rapid science in some of the areas that we were working,” says Carl Bergstrom, an evolutionary biologist at the University of Washington in Seattle. For instance, one of his most important early collaborators in trying to understand and model Covid-19 through Twitter was a hockey statistician, Bergstrom says.

At the same time, prominent Covid-19 researchers experienced insults, abuse and sometimes, as a 2021 Nature survey showed, death threats – often through Twitter.

Meanwhile, some researchers on the site oversimplified information, posted alarmist analyses or shared outright disinformation, Bergstrom adds. And despite Twitter’s self-styled reputation as a public town square – where everyone gathers to see the same messages – in practice, the pandemic showed how users segregate to follow mostly those with similar views, argues information scientist Oliver Johnson at the University of Bristol, UK. For instance, those who believed that Covid-19 was a fiction would tend to follow others who agreed, he says, whereas others who argued that the way to deal with the pandemic was to lock down for a ‘zero Covid’ approach were in their own bubble.

Bergstrom thinks the positives of Twitter outweighed the negatives. During the pandemic, it gave the public more transparency about the uncertain process of science progressing in real time, he says. And if some audiences wanted to leap on to messages of scientific certainty where there was none, that wasn’t Twitter’s fault, he adds.

“I don’t think we’ve done a good job of talking in school science classes about the process of doing science, and explaining to people how the social process of science operates,” he says. “When you actually see science in the making, it looks very, very different.”

Days after Bergstrom spoke to Nature, however, he locked his own account after Musk’s mocking tweet about Fauci.

“You can’t have meaningful and productive scientific collaboration on a platform run by [a] right-wing troll who denies science when its results are inconvenient to him and just simply to hear his audience cheer,” he wrote on Mastodon.

Besides Musk’s personal views, his changes to Twitter have worried plenty of scientists – particularly because he fired many people who work on content moderation. Scientists have noted, in particular, a Twitter announcement on November 23 that it would stop enforcing its Covid-19 misinformation policy. And there have since been reports that hate speech on the platform is increasing, including in areas such as climate science.

“We’ve been having conversations about if Twitter is now a safe place for our organization to exist, because of the way it’s changing,” says Bonner. “At this moment in time, I don’t know.”

Information scientist Stefanie Haustein at the University of Ottawa in Canada, who has studied the impact of Twitter on scientific communication, says the changes show why it’s concerning that scientists embraced a private, for-profit firm’s platform to communicate on. “We’re in the hands of actors whose main interest is not the greater good for scholarly communication,” she says.

Researchers leaving the platform will probably try to find a similar social-media replacement, says Rechavi. “I imagine that if Twitter stops being the place for scientists to be, then it’ll be replaced by something else,” he says. “I just can’t imagine going back to being disconnected from the rest of the science world.”

But Bonner says she doesn’t think there’s yet a space similar to Twitter. Dynamics on Instagram, where #BlackinCancer has a foothold, are drastically different, with less conversation and less reading of posts. And on Mastodon, the open-source alternative to Twitter that Bergstrom and Knight joined, users can post longer messages, but the dynamics of the platform deliberately make it harder to discover or encounter messages from users one doesn’t directly follow, making communities more siloed and fragmented.

User numbers are still tiny compared with Twitter, estimated at some 2.5 million in early December.

“A social network is always only successful if it’s got enough people, and if it’s got the right people,” says Haustein. “It requires millions of people to move from one place to the other.” Even if that happens, she says, you need to rebuild the same networks and structures that existed on Twitter, which is proving hard because of the way that control of Mastodon is distributed across servers, making it difficult for those who were on Twitter to reconnect.

Still, Quintana is hopeful, “Despite the fact that I’ve probably got 10 times more followers on Twitter, the stuff that I posted is getting about the same amount of engagement on Mastodon,” he says.

For many, the tweet about Fauci was a final straw. Afterwards, a fresh wave of scientists decided to leave Twitter. But some are encouraging their colleagues to stick around. Rechavi emphasises that Twitter has had a crucial role in research: “I hope it survives,” he says.

And, although the platform’s worst qualities are becoming more common, say researchers who spoke to Nature for this article, there is still a need for trained scientists to provide their expertise and point people to the best sources of evidence-based information.

In reply to Bergstrom’s farewell, Trish Greenhalgh, a health scientist at the University of Oxford, UK, argued that people like him are still needed, and that she feels duty-bound to carry on: “We can and must stick around and post sensible scientific tweets. I’m staying.”

• A Nature report