Inside Twitter’s ambitious plan to change the way we tweet

There are few things the internet loves more than a vicious Twitter dunk.

Whether it’s a celebrity ripping on President Donald Trump or a politician roasting her political rivals, Twitter is the perfect forum for the kind of wisecracks, sarcasm, and snappy one-liners that can go viral just like a LeBron James windmill slam. In many ways, Twitter was built for this. The brevity, the ease of virality, and the general snarkiness of the platform have turned dunking into, as Slate called it, a “delicious sport.”


The problem, though, is obvious to spot: Sure, a Twitter dunk, even a nasty one, can be fantastic to watch. Unless, of course, you play for the other team. Publicly trolling or mocking another Twitter user isn’t conducive to good, clean, productive conversation, which is a problem for a company that’s made facilitating conversations a top priority.

So a year ago, Twitter CEO Jack Dorsey announced plans for a somewhat idealistic solution, not just for minimizing Twitter dunks but also for minimizing all kinds of angry, vile, and abusive user behaviors that don’t necessarily violate its rules: He wants to invent a new metric that measures Twitter’s health, and then optimize for it. Twitter even partnered with outside researchers to come up with new metrics for what “healthy” actually looks like.

If Twitter can identify which user interactions are healthy, the thinking goes, then maybe it can change the product to encourage more of those behaviors while discouraging more antisocial conduct.

“If you dunk on somebody and you get a lot of engagement, a lot of ‘Likes,’ a lot of retweets, that is encouraging you to be mean, basically,” said David Gasca, the Twitter product executive in charge of the company’s health efforts, in a recent interview with Recode. “We could imagine ways of changing the product in order to [discourage] that.”

“At the same time,” he continued, “you could imagine changing [the product] such that you provide positive incentives for encouraging more constructive conversation.”

It’s a nice idea, though measuring Twitter’s health is taking much longer than expected, according to exclusive interviews with both Twitter employees and company partners.

The research teams that Twitter announced last July to help them work on this project haven’t even started. One of the two teams has abandoned the project altogether. Internal metrics Twitter is building on its own are still in the “experimentation” phase and aren’t being tested in the wild.

Which means that while Twitter might not think the dunk is a particularly healthy user behavior, it’s probably not going away anytime soon.

Twitter is sick

The idea to measure Twitter’s health was planted in Jack Dorsey’s ear by Deb Roy, an MIT researcher and one-time Twitter employee.

Roy sold his TV analytics startup, Bluefin Labs, to Twitter in 2013 and quickly became Twitter’s chief media scientist, a part-time role that allowed him to teach and conduct research at MIT. It was there that Roy started the Laboratory for Social Machines, a research effort to study public conversations, and Cortico, a nonprofit that partners with that lab to promote its work outside the university. Twitter committed $10 millionto help fund the lab in 2014.

Roy stayed in touch with Dorsey and routinely shared with Twitter executives the lab’s research on topics like viral rumors; he even presented other research at Twitter’s week-long, all-company retreat last summer in San Francisco.

A few months before that retreat, Roy asked Dorsey a thought-provoking question which, according to sources, spurred Dorsey’s tweetstorm last March outlining the health measurement project.

“Recently we were asked a simple question: could we measure the ‘health’ of conversation on Twitter?” Dorsey tweeted at the time. “If you want to improve something, you have to be able to measure it.”

“Health” has been Twitter’s top buzzword for the better part of a year.

Everything the company seems to do — from cracking down on bots to building new conversation features — has been done in the name of a healthier Twitter. When the company’s user base started shrinking noticeably last year, Twitter said that its focus on health was at least partly to blame.

Measuring the health of interactions is just one part of that broader effort, but it’s one of the more challenging and confusing parts. Removing bots and spam are technical problems. Truly understanding the health of a conversation requires things like understanding who is talking, what they’re talking about, or when someone is using sarcasm. Not all arguments, of course, are bad.

“There is great diversity in what people consider ‘healthy,’” Roy explained in an email to Recode. “Since the platforms did not at their onset define norms, it is much more difficult to retrofit norms long after the networks have grown and taken root.”

On the same day as Dorsey’s tweetstorm, Roy’s nonprofit Cortico published a blog posttitled “Measuring the Health of Our Public Conversations.” The post introduced four new metrics that might be used to quantify what a healthy conversation looks like, metrics like “receptivity” and “shared reality.”

If you haven’t heard of those metrics, that’s because they don’t yet exist — at least not in their final form. And that, in a nutshell, is why the idea of measuring the health of conversations promises to be one of the most challenging aspects of Twitter’s recovery plan. There’s no widely adopted way to quantify the health of human interaction, especially at the internet’s scale.

Carolyn Penstein Rose, a professor in the computer science department at Carnegie Mellon, has spent the past decade studying conversations and the technological systems that can be used to improve them. Rose’s area of focus has mostly been studying how human interactions impact learning, but she believes the problems Twitter faces are related.

“Let’s not make this a machine learning problem — it’s a language problem,” Rose said in an interview with Recode. “My advice to Twitter would be, if you want this to be done right, get people who know language, not necessarily [just] machine learning people.”

Inventing a new metric

Dorsey seems to be taking that advice. He was so serious about measuring the health of conversations on Twitter that, a year ago, the company asked researchers to submit proposals for how it could actually do that.

More than 230 proposals were submitted, and last July, Twitter announced that two research groups had been selected as official company partners — one from Leiden University in the Netherlands and one from Oxford University in England. They would receive access to user data and a monetary grant from Twitter; in exchange, these researchers would create new metrics intended to measure the health of interactions on the service……Read more>>

Source:- recode