Elon Musk’s Idea for ‘Open Source’ at Twitter Sounds Good, but it Might Not Work

Some


Twitter

users who think the site discriminates against their views have latched onto one of the geekier-sounding prescriptions that owner-in-waiting Elon Musk has for improving the platform: making the algorithms that govern the site’s behavior “open source.”

Twitter’s algorithms are embedded in countless lines of abstract code that, by themselves, would hold no insights into its alleged biases, even to a trained eye, said Vipin Chaudhary, who chairs the computer and data sciences department at Case Western Reserve University.

Musk’s “open source” talk “may be more marketing or creating a sense of trust,” Chaudhary said. “I think people are not informed.”

Much of Musk’s commentary surrounding his $44 billion plan to buy Twitter has revolved around his professed desire to see less restrictive policing of what gets posted on the site. Many have questioned what this means in practice, and whether it’s a viable strategy if it alienates users and scares away advertisers.

But other changes that the


Tesla

CEO has said he plans for Twitter, including the “open-source” algorithm, may also be more easily said than done.

“Anybody that’s looked at Twitter very carefully and understands that things need to change,” said Robert Picard, who researches online media and communication policy at University of Oxford and other institutions. But Musk’s ideas for improving the site, he said, are “very opaque.”

A message sent to Tesla’s investor-relations department seeking comment from Musk yielded no response.

Musk, the world’s richest person, announced his deal with Twitter’s board Monday to acquire the company for $54.20 a share. The announcement came three weeks after Musk disclosed that he’d become a major shareholder in Twitter and 11 days after he first said in a regulatory filing that he aimed to buy the company outright.

Throughout that sequence, Musk has sprinkled hints about the direction he hopes to take the company. In a series of tweets after disclosing his 9.2% stake in Twitter, Musk outlined a plan for a cheaper, more feature-rich version of the site’s current $2.99-a-month subscription program, called “Twitter Blue.”

His version would cost a dollar less each month, and—he said in a separate tweet that was later deleted—would earn subscribers the coveted blue check mark that Twitter awards to users it deems to be “notable,” among other criteria.

Musk has also called for Twitter to allow users to edit tweets, something that the company has said may soon be one of the perks available to Twitter Blue users.

The billionaire has cast his desire to rely more on subscribers as a way to become less needy of ad revenue. Some advertisers have concerns about being seen to support a platform thought to have lax moderation policies.

“The power of corporations to dictate policy is greatly enhanced if Twitter depends on advertising money to survive,” he wrote in another tweet that he later deleted.

Jim Friedlich, a longtime media executive who now leads the Lenfest Institute for Journalism, owner of The Philadelphia Inquirer, said that Twitter has a large, highly engaged user base, something that other online brands have leveraged into lucrative subscription businesses. “If and when Twitter creates successful premium subscription offerings in addition to their core free and ad-based business, it could create quite meaningful value for the company and its new owner,” said Friedlich.

But the path to a successful subscriber-based strategy is unclear, said Howard Saltz, a former news executive who recently served a media-innovation residency at Florida International University’s journalism school.

“If he’s talking about a subscription in the typical sense of the word—that is, somebody pays to read content—I’d say it’s dead on arrival,” Saltz said. “The content is exactly the kind of content somebody would not pay for.”

The alternative to that is to give paying customers enhanced features on the platform, which could “bring up ethical questions,” he said.

Just granting blue verification check marks to paying subscribers arbitrarily implies that they have something more valuable to say than nonpaying users, Saltz said.

And it’s a slippery slope from there to a scenario where, say, paying users have their posts more prominently featured, or are given the right to downgrade others’ posts, he said. “It’s a potential minefield when you’re charging for opportunities to do things, rather than just to consume content,” Saltz said.

Still, there could be a synergy between a subscription strategy and another of Musk’s stated objectives: cracking down on the fake user accounts known as “bots.” Most users with paid Twitter accounts would use credit or debit cards to pay for their subscriptions and those payment accounts could be used to verify that they are actual humans, said Picard.

But really weeding out the phony accounts would require huge numbers of people to canvas the site looking for suspicious posting patterns and stilted language, said Vladimir Filkov, a computer science professor at University of California, Davis, who specializes in data-heavy software projects.

Artificial intelligence isn’t yet advanced enough to pick up on telling nuances that human trackers can catch, Filkov said. “It will have to be done by trained individuals,” he said. “And this will over time increase the cost of this operation.”

For some, the most confusing of Musk’s proposed tweaks to Twitter is his “open-source algorithm” vow. The term “open source” generally refers to the practice of making a computer application’s underlying code available for other programmers to review so they can collaborate on development, suggest improvements and catch bugs, or borrow snippets for their own projects.

An algorithm, meanwhile, is a complex set of math-based rules that tell a system how to behave based on the data that it encounters. Software developers write code that tells computers how to execute these theoretical processes.

But “open sourcing an algorithm is a non sequitur,” Filkov said. It’s “a weird term.” Musk could mean that he wants to publish Twitter’s source code for users to review, but that would shed little light on the site’s behavior, Filkov said.

For one, source code is written for computers to read, not humans, so it would take someone with lots of programming knowledge to even understand what those lines of numbers and commands are meant to represent, he said.

Even a skilled programmer might find few insights. The Twitter behavior that some see as biased isn’t overtly embedded in its algorithm, or in the code that executes those processes, said Chaudhary, of Case Western. Rather, it stems from how the algorithm reacts to—and in some cases learns from—the vast amount of data that it encounters, Chaudhary said.

Musk and Twitter would almost certainly never make those data available for public review, since they contain user records and valuable internal business information, he said. But even if they did, few people or institutions have the computing power needed to marry that data with the software code for analysis, Chadhari said.

Write to jacob.adelman@barrons.com

.

Leave a Comment