Profile picture
Sarah Eaglesfield @zenxv
, 81 tweets, 25 min read Read on Twitter
I'll live tweet through #TwitterTestifies - introductory statements from @repgregwalden. Talking about the power of Twitter, and how revolutionary it has been. Now moving into censorship concerns and discussing shadow banning to de-emphasise certain viewpoints.
Now @FrankPallone talking about the President's use of Twitter, already mentioned @RealAlexJones, addressing bullying and spread of fake news via Twitter. Swift and equal imposing of regulations, Twitter needs to stop being a platform of discord.
Jack's reading a prepared statement. Says he's going to Tweet it out right now. Indeed, he tweets it out right now over about 40 tweets in a thread... ahem.
Jack points out: "We do not have a responsibility, nor you a right, to amplify your tweets to audiences that don’t follow you." #TwitterTestifies
Interesting aside: @Jack was raised with parents who were Republican / Democrat and feels he was able to express his opinions safely and form a fair opinion.
The auto-fill bug is now being addressed @repGregWalden. Jack explains it was follower behaviour that reduced the visibility of these users (around 600K), not the behaviour of the accounts themselves.
Back to @FrankPallone - he's talking about @realDonaldTrump's use of Twitter to spread discord and misinformation. He's asked how many human content moderators Twitter have and how much they get paid... how much training do they get to instruct their decisions.
Jack to report in writing within a month. @FrankPallone asks:
1. Will you alow a Civil Rights Audit of Twitter by a third party?
Jack says there's Twitter Trust & Safety (which is 3rd party)
2. Will you make the findings public?
3. Will you change policies after finding?
Jack talks about conversational health measures discussed by Twitter earlier in the year:

- Amount of shared attention?
- Amount of shared facts (true or false)?
- How receptive are people to conversation?
- is there a variety of perspective, or is it an echo chamber?
Now @repdianadegette mentions Toxic Twitter report from Amnesty International, @Jack says he's aware of it. Report discusses how women are abused on Twitter. Does Twitter keep records of gender / racial harrassment (as a separate metric)?
Jack says that the data on abusive tweets is being aggregated into a report. Working on technology to recognise abuse and remove it before it is reported, at the moment relies on human reporting.
Nice point from @repshimkus who asks how Twitter are working to prevent "the loudest mob" & group banning from affecting results. Now talking about the verification system in Twitter.
Jack says the verification system needs an overhaul. Sometimes verified accounts get prioritised in conversations, and Twitter are asking themselves if it is still revelant. So far, Jack says that it seems to be the case.
Green from Texas mentions GDPR and user privacy. Jack says Twitter has different needs as when someone tweets they know they world will see it, but they are very focused on protecting private data. Green mentions Iranian and Russian bots.
Jack says that they look for bots that come in through the API (hi bots) and perform human challenge checks. There are also bots that 'pretend' to human and use automated software to act that way.
Now @michaelcburgess is talking about the speed at which abusive tweets get taken down, such as John McCain tweet this wknd. Jack says they need to do a better job at prioritisng (for violent and threatening tweets), and using it as a lesson to help improve the responsiveness.
Jack is read statements from Trump, Nunes - denies shadow banning, and censoring people with a flat "no".
Now @repjoebarton is asking for for breakdown of the 600K accounts by Democrats and Republicans. Jack says that's hard to identify. Says algorithms aren't biased against a particular philosophy and viewpoint.
Now @DorisMatsui is talking about anonymous users, saying that use of geospatial data can work out who users are. Jack explains that all data that is collected can be turned off by the users, and the data that is collected is transparet to users.
Matsui is talking about blockchain. And asks Jack what the potential applications of blockchain are.... 😵 Jack explains that they need to remain focused on the issues they are trying to solve today, but some interesting applications for Blockchain in the future.
Jack says users own their data, regardless of modifications, and should have ultimate control over it.
Rep @stevescalise talking about the Arab Spring (which @GCHQ and @NRA used sockpuppets on Twitter to influence users, although he doesn't mention that) - talks about how Marsha Blackburn got banned for talking about sale of body parts from aborted babies.
Jack says it was a mistake, and Twitter was at fault. Steve asks if anyone has been held accountible, and says that anyone who writes algorithms that are discriminatory against conservatives should be identified. Jack says the algorithm wasn't biased & people got caught up.
Some woman is screaming shit. And the chairman asks her leave. No one can understand her. One of the reps pretends to be a auctioneer whilst the commotion is resolved! Round of applause. Who was that? (@usreplong?)
Now, @repeliotengel asks why it took an external security firm to identify the Russian and Iranian bots. Jack says that they are getting more and more confident in identifying bots, and they need to stay ten steps ahead of the attacks.
In response to @boblatta, Jack says that no third parties are used to collect data about Twitter users. When tweets are embedded in external websites, they get a hit & it may infer a particular interest, to help them better target ads.
With regards to removal of verification mark, Jack says the off-platform behaviour they look for is participation in violent hate groups, etc. They have increased their attack on bots trying to manipulate conversations.
Now @USRepKCastor is asking about automated accounts overwhelming Twitter. Jack says 1/2 million users asked to user check every single day. Jack explains that bots are identified when they sign up, but there are more sophisticated ways by scripting the site to create accounts.
Jack says that users who are not logged in are not tracked, in response to a question from @USRepKCastor.

@cathymcmorris is focused on Twitter's procedures on taking down offensive conetnt. "America This One's For You" McCain tweet took 16 hours to take down.
Jack repeats that the burden of reporting abusive harrassment by the victim needs to be resolved, and they are writing algorithms to remove such abusive tweets. Any violent threat or image is at top of priority list. [personal note: a concern is misidentification of abuse]
Now @RepSarbanes mocks the idea of a Liberal Bias on Twitter (despite all the video evidence to the contrary). Doesn't have confidence that the meeting was convened for legitimate purposes.
There is a 'secret spreadsheet' circulating amongst Republicans about things they don't want the public to hear about. Such as Trump's tax returns, Russian collusion, how poorly they responded to hurricane Maria... 60 days away from midterms, same actors are still sewing discord.
Doesn't ask Jack a thing, and uses his platform to bash Republican policy.
Okay then.
Onto @GreggHarper who asks about bias in algorithms.
Jack says that unexpected things were found in the signals they were using, and when they recognise impartiality, they try to fix the signals that are used. Are the algorithms making objective decisions, and are engineers using engineering vigour.
This is the most enlightening questioning so far. @GreggHarper says how does your definition of fairness align with my idea of fairness? Jack talks about conversational health again and trying to avoid the echo chamber & filter bubble that Twitter can become.
Looking at how ads are targeted to people, @RepMcNerney asks if it could be discriminatory. [kind of the whole point of targetting ads?] Has Twitter ever taken down an advert because of it being discriminatory? Jack doesn't know.
Is data deleted when user turns off tracking or logs out, asks @RepMcNerney.

Jack believes so, but would need to get back to him.
Turns out 'the woman screaming shit' was Laura Loomer, surprise, surprise. As trashy as ever.

Olson asks about Trust & Safety Council - made up of 40 groups - and who else they take input from. Jack says it's all advisory and they don't take direction from anyone.
Mr Lujan, @repbenraylujan asks about President Trump losing followers from bot accounts. States that Obama lost 2.3mil, President Trump lost 320K followers folloiwng bot purge.

[Yep, sounds good to me.]
Here, @repmckinley raises the issue of drugs being sold and promoted through Twitter. Why isn't it being filtered out?

"Contact us for any medicine you want", adverts for coke, ecstasy, just in the last hour. Wants to see them removed.
Iowa's @daveloebsack asks about protections in place to protect their younger users, and asks about safety checks for those users. Jack says there are protections in place, and looks at how bullying can be stopped through Trust & Safety council.
What data points are collected and who are they shared with, asks @repguthrie (RIP victims of the Bowling Green massacre).

Jack says that data on your inferred interests is viewable from your settings, and can be turned off, and is not shared with outside parties.
Twitter wants to be representative of as many perspectives as possible, and defends sec 230. Moments is a curated view of twitter, they internally view Twitter as a conversational product, and as a place for news.
Drooling Joe Kennedy isn't drooling today. He asks Jack about his own suspension, Jack reiterates that the burden shouldn't be on the victim to report abuse and they need to build proactive systems that recognise behaviours against terms of service.
Interesting point by @RepKinzinger about a law regarding storage of information on Russian Citizens to be kept. [Which I've never heard of, despite him saying it's being in the news for the last two years].
Jack confirmed that users are responsible for their content but the company tries to improve the health of the content, by ensuring users don't silence or bully others.
Bullying online is inescapable, according to @RepCardenas: names young celebrities who have stopped using Twitter or taken a break due to bullying, and how Twitter handles bullying of younger 'everyday' folk.
Once again addressing the issue of opiod abuse, @repmgriffith says this needs to be tackled as a priority, kids are dying, AI exists already - Jack agrees. He also says that the Twitter Trust & Safety Committee need a right wing group on the board. "We're not all crazy".
.@repscottpeters points out the irony of Trump using Twitter to bash Twitter. Quizzes Jack's definition of fairness: Jack says impartial outcomes, nothing based on bias, healthy conversation. Not an arbiter of truth; % of people with 'shared facts' (regardless of truth) measured
Now @RepGusBilirakis talks about threats made on Twitter and asks about local enforcement, particularly regarding school distrcits. Jack says they have lines to local enforcement and they're looking at ways to quickly alert the concerned people.
Moments are customised to show what is popular and relevent to your interests. Now @RepGusBilirakis points out that they tend to reflect a one-sided view or political narrative [I also agree with this]
Jack says that they are looking for opportunities to show a variety of perspectives; @jack - I'd flag this, the alternative perspective, if presented at all, is usually at the bottom of the Moment.
Now @RepDebDingell addresses the black box nature of the algorithms, and asks about the datasets used to train the AI and how often they are retrained. Jack says they use previous experiences and outputs that they've seen, natural use cases & edge cases.
She questions Jack whether they can see WHY an algorithm made a decision. Jack says that the engineers are working on explanability, and it will be the one that bears the greatest fruit in terms of trust. @RepDebDingell now asks about acceptible error rate, Jack to get back.
Is the alogrithm a decision tree, asks @RepBillJohnson? Ten different programmers will program ten different decision trees. How do you go about weening out that inate bias? Peer reviews of algorithms? Jack says "yes". Bill points out advances in AI should allow faster response
He also asks for metrics that make a topic trend (#TwitterTestifies not trending...). Jack says the algorithms look for velocity switch in the use of hashtags. Users can choose whether it is personalised or whether it is geographic.
Legal or market forces could drive Twitter to a place they don't want to go says @RepBillJohnson, and they need to use external parties to address that. Over-runs his time.
On the subject of advertising @reppaultonko talks about the need to go to the Ad Transparancy Center to get the details. Jack says anyone, even without a Twitter account can look at who is behind and advert, and how it has been targeted.
Auctioneer @usreplong congratulates Jack Dorsey on his demeanour and genuine presence. Lists off the Moments Highlights that have been sent to him via Email over the last week, all showing very liberal bias.
Out of 84 "hanpicked tweets" in his emails, Billy says they were all "Trump bashing", except one (potentially two). Asks for it to be taken into consideration.
Now @repschakowsky points out that Twitter in incentivised to keep all voices on the platform. Jack disagress, and says that Twitter will be removing accounts; however, all perspectives should be represented.
On the question of timeline raised by @RepLarryBucshon, Jack says they want to move as fast as possible (and understands that is a frustrating answer). They want to increase the health of the public square they have created.
Jack also says there is a potential that they may get third party users involved in inspecting their algorithms, but don't want to let people game the system by making it open source; so they need to decide how it's going to go.
Asks about Jack's hiring practises. Jack has recognised they need to decentralise their workforce outside of San Fran, and they're considering ways how to hire more broadly globally and being more flexibile, to get a variety of opinion within their employee base.
Now @congressmanruiz asks about bots making up 5% of the Twitter population, looking at the behaviours of those accounts. How are they identified? Jack says spammy like behaviour to amplify conversation beyond its organic reach.
Ruiz's concern is how the amount of users following someone is an important metric, and that the number of followers is not an accurate representative of who is really following someone.
Ruiz asks Jack to get back to him regarding specificity and senstivity - false positives and false negatives. Some people are concerned with too many false positives, some too many false negatives.
Now @RepBillFlores asks Jack about balancing moderating vs free speech. Twitter's definition of behaviour, hateful conduct, low quality tweets, explanation of abuse reports, and what the signals are for ranking and filtering, signals for suspicious activity.
Three biggest changes to expect in a year:
@Jack -
1. Progress on increasing health of public conversation.
2. Reduced burden on victim to report
3. Deeper understanding of real world affects of platform to public & individual.
On co-ordination of violence through Twitter, bought up by @repbobbyrush, Jack says that they are often removed and there are a series of enforcement actions.
Jack talks about the forthcoming transparancy report - already finished for terrorism, still being finished for safety.
Jack asked to provide behavioural signals used for decisions by @repryancostello. Says that there could be 100s or 1000s and they change weekly. There are no conclusions just yet, and they are considering which signals will be static in the future.
He's also quizzed by @repryancostello about Moments human curators, and how biased they may be, as they decide what to use. Moments try to reflect how much things are being talked about on the network.
Now @repmullin talks about Mary Liu's racist tweets against White Men compared to Candace Owens tweets which simply replaced the word 'White' with 'Jewish'. Jack says - again - a mistake was made.
In response to @RepWalberg asking how to be a "productive, positive member of Twitter" if he doesn't know why he's been taken down (yeah, I laughed).... Jack says the timelines are skewed. Says there will be a transparancy report on throttled content.
20-something female staffer who started a Twitter test account for a separate project got a 100% suggested accounts of left-wing politicians, from email address and 202 area code, reports @repjeffduncan. Same issue reported by @WarRoomShow yesterday.
No atheletes, no celebrities, only liberal politicians suggested. That was obvious bias. Jack says that they were probably the most engaged accounts for that area code, however @RepJeffDuncan isn't buying it.
Last questions from @repbuddycarter - says a conservative teenager with 40K followers hasn't been verfieid. It's not to do with followers says Jack (indeed some accounts with 200 followers verified) - Buddy claims that this student's friends got verified.
Now on the subject of IP on live streams and piracy. It's been going on for a very long time, and Periscope hasn't fixed it. Jack said that work was stalled because methods of attack constantly change.
The unverified teen that @repbuddycarter was talking about in the congressional hearing (but possibly gave the worst reasons ever for verification) is @TheNoahRing, host of the Let Freedom Ring podcast, encouraging students to #WalkAway. You should consider him @jack / @verified
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to Sarah Eaglesfield
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member and get exclusive features!

Premium member ($3.00/month or $30.00/year)

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!