Twitter CEO Jack Dorsey posted a series of tweets on Thursday to concur with growing resentment about his platform—and to admit that Twitter is somehow not in a position to fix the problems he listed.

Dorsey’s tweet thread began with an all-too-familiar promise of “public accountability” (which is a clever thesaurus-twist of his usual “more transparent” promises), before then delivering what might be his most frank admission of Twitter’s woes:

We love instant, public, global messaging and conversation. It’s what Twitter is, and it’s why we’re here. But we didn’t fully predict or understand the real-world negative consequences. We acknowledge that now and are determined to find holistic and fair solutions. We have witnessed abuse, harassment, troll armies, manipulation through bots and human-coordination, misinformation campaigns, and increasingly divisive echo chambers. We aren’t proud of how people have taken advantage of our service or our inability to address it fast enough.

That last part, about Twitter’s apparent inabilities, resonates through the rest of the Thursday posts. Dorsey claimed that Twitter has been “accused of… optimizing for our business and share price instead of the concerns of society” and that the company has fallen behind in part by focusing on the removal of TOS-violating posts instead of “building a systemic framework to help encourage more healthy debate, conversations, and critical thinking.”

A more holistic approach apparently began for Twitter by seeking outside counsel about “the health of public conversations.” Dorsey pointed to the private research firm Cortico, who created a series of conversation “health” metrics based on its studies of Twitter data: shared attention; shared reality; variety; and receptivity. The metrics that Dorsey linked to describe behaviors, however, as opposed to a platform’s design—with phrases you’d expect from an elementary school etiquette chart like, “Are we open, civil, and listening to different opinions?”

Thus, Dorsey admitted that Cortico’s metrics might not fit what he seeks in “a rigorous and independently vetted set of metrics” regarding how Twitter users interact. He announced a grant of “meaningful funding” and Twitter data access for any research team that proposes what Twitter deems a worthy conversation-measurement plan.

But is it a cancer?

Dorsey’s tweetstorm juggles two seemingly disparate ideas: the company’s “inability” to contend with disruptions, abuse, and hatred, and its desperate call for outside help to measure the current state of events, as opposed to shoring up resources or building entirely new systems.

As more data and research pile up to point fingers at Twitter’s biggest issues, there’s also the matter of anecdotal evidence of the service’s inherent issues. I offer the following stories as an answer to Dorsey’s questions about Twitter’s health: its cancer may be incurable.

To review: because it’s a free service, Twitter’s money-making moves are currently limited to either selling ads or monetizing the data it accumulates. To harvest either of these possibilities, the service relies on aggregated, non-chronological content feeds (like most other free social-networking services). One byproduct of this approach is that, in addition to advertised content, a typical Twitter user will see everything from “you may have missed this” summaries to “your friend liked or re-tweeted” content emerge, as a sort of friend-of-a-friend content nudge.

Twitter has also spoken publicly about its investment in automated tools used to identify inauthentic and robotically created accounts. More avid and followed Twitter users who turn off the service’s “quality” filters can sometimes see this in action, where random gibberish replies might appear for roughly an hour before being auto-deleted.

Dorsey’s calls for conversation health metrics do not in any way appreciate the apparent next-level disruption tactic already being rolled out on Twitter this year: subtler, seemingly real accounts popping up with the express intent of passing those four metrics on their face. I have chronicled an apparent rise in this account type for the past few weeks at my own Twitter account, often finding accounts that have existed for as briefly as a few months or as long as nine years.

These accounts have a lot in common: a series of everyday-person thoughts from a unified perspective; a lack of a consistent social circle, in which they use Twitter to exchange pleasantries, jokes, or other group-of-friends interactions; a sporadic posting history that isn’t immediately evident when flipping through an account on a smartphone screen; and a reliance on RTs, likes, and replies to verified and well-followed Twitter accounts.

Let’s review what we learned about political disruptor accounts on social media services throughout the 2016 election cycle. From my report in October 2017, following the reveal that Dorsey inadvertently RTed an apparent Russian troll in 2016:

[Disingenuous actors’] accounts often debut with an apparent political and social identity, along with a stress on link shares with unique, viral-styled descriptions and exclamations. This is perhaps done with hopes of gathering followers, “likes,” and shares. Then, when it politically suits an operator, an account may start sharing politically divisive messages.

The difference with that description and the activity I’ve anecdotally chronicled in the past few months is a reduction in obvious virality. Instead, these accounts are at their most intense when interrupting other major users’ Twitter conversations or threads with rebuttals and “well, actually”s. So long as these interrupters abide by Twitter’s TOS, their disagreeable posts and arguments break no apparent site rules.

And when these accounts’ replies and RTs are in any way acknowledged by verified or bigger-ticket accounts, by being RTed, replied to, or quoted, there’s no current way for Twitter to measure the health or harm of that interaction. Instead, the current algorithm is designed solely to suggest and auto-forward content that is simply the busiest—the most liked, most seen stuff.

As of press time, Twitter continues to ignore calls for a new kind of report: one that alleges inauthentic or troll-farm activity. One that lets a user say, for example, “this account has waded into a local-politics conversation in spite of having few authentic interactions with people or issues in my city.” Worse, Twitter was designed to amplify these exact kinds of voices—the ones that appear out of thin air to condemn brutal and even fascist practices across the world and that are amplified when a groundswell of similar voices appear to call for action. That Egyptian revolution we all crowed about in 2011 might have looked very different on a more “secured” Twitter, and in the present day, an authentic, incredibly shy account that has a lot to say about regional transit budgets can look as suspicious as the apparent inauthentic actors I’ve seen on Twitter for months.

Again, to be totally clear, this anecdotal take is still subject to less dramatic results, like genuine accounts or harmful-yet-simple trolling efforts from bored teens. But if it turns out that accounts like these continue to be run and operated by paid troll-farm services—bounced around via VPNs to avoid IP address scrutiny, operated by minimum-wage workers in third-world countries, possibly fostered by ad purchases using American bank accounts—then the big question is, what’s their end game? American political disruption certainly seems possible, especially in the cases where they pop up specifically to offer thoughts on political issues as big as assault rifles and as small as city council meetings.

But it could be something even scarier: an effort to test and tease Twitter’s systems and to harvest innocent bystanders’ reactions, thereby dumping fuel into an artificial intelligence-powered botnet. 2018’s Twitter is already confusing, in terms of verifying whether a drive-by poster is in any way legitimate. What happens if 2020’s Twitter is inundated in tens of thousands of real-sounding, TOS-abiding robots—assuming Twitter still exists by then?

https://arstechnica.com/?p=1268209