Renee DiResta Profile picture
Stanford Internet Observatory, Mozilla Fellow alum. Interested in how narratives spread online. Prev: quant finance, VC, startups. Many opinions about SF.
Joanna ( ze/ zir/ zey ) Profile picture eDo Profile picture Trump has made normal his DEMOCIDE against us. Profile picture Hugh Manatee ... Who was that masked man? Profile picture Troy Rubin Profile picture 8 added to My Authors
8 Jul 20
4 takedowns tied to prominent political figures (in their respective countries) from FB just now. Activity is interesting bc
1) uses mercenary PR firms
2) bright line for takedown is hard to establish.
3) no real consequences

so we'll likely see more.…
On point 1: mercenary firms afford plausible deniability. State actors have known and leveraged this for years. The IRA is one example; a ton of info ops out of Middle East (Saudi, Egypt) have used marketing agencies to run govt/political operations.
Point 2: domestic activity is really tough b/c it bumps up against free speech/expression issues, and b/c takedowns are politicized (as the Roger Stone one no doubt will be, imminently). There is no "bad guy" from another country pretending to be something they're not.
Read 9 tweets
19 Jun 20
🚨🇷🇺 Last week, Twitter announced a takedown of 3 networks: coordinated inauthentic behavior, out of China, Russia, and Turkey. SIO published on the China and Turkey operations last Thursday (links in thread). Today our report on the Russia set is out.…
These were attributed to a publication called Current Policy, a "media website engaging in state-backed political propaganda within Russia" and contained content related to politics, commercial activity, and an interesting cluster that has links to Secondary Infektion network.
The politically engaged accounts in the network focused on amplifying pro-government activity and cheerleading for President Vladimir Putin and his party, United Russia. Anti-opposition activity was additionally present: attacks on Navalny and the Anti-Corruption Foundation
Read 9 tweets
11 Jun 20
🚨🧵 Today Twitter announced a takedown of a network of 23,750 accounts attributed to the PRC 🇨🇳, with technical indicators linking the operation to the network suspended in August 2019. SIO participated in the analysis. Report:…
Most of the 23,750 accounts in this disclosure were caught relatively quickly; this wasn't a high-engagement op. Topics fell into 4 buckets: Hong Kong protest, attacks on Guo Wengui, COVID-19 spin, & Taiwan. They shifted in prominence over time.
Our report includes TTP analysis of the accounts. Personas are as underdeveloped as they were in Aug 19 takedown (interesting). Accounts have botlike usernames, do a minimal amount of audience-building. Other state actors (​🇷🇺​) remain far more sophisticated in that regard.
Read 11 tweets
2 Jun 20
The IRA and GRU have both been documented doing this. (Thread 👇)
The IRA account was an Instagram account, @_anonymous_news_ that renamed itself after initially pretending to be a local Jackson, Miss. news account for a while; it also spent some time as a gun-related account.
There have also been bizarre fly-by-night Anonymous accounts that have popped up targeting Latin American audiences during times of political unrest. It's unfortunately an easy strategy to execute and hard to refute.
Read 3 tweets
28 May 20
"The draft order also states that the White House Office of Digital Strategy will re-establish a tool to help citizens report cases of online censorship." Re-establish? Does this mean that crazy form for ppl to upload photos documenting "censorship"...…
Almost exactly one year go. And then a month later we got the "social media summit" - at the time, the "censorship" was behaviorally-based takedowns recast as ideological takedowns. The other concern was shadow-banning.

Now "censorship" and "bias" is...a fact-check link.
That form was incredible. Anyone could upload, you can probably imagine what the internet decided to upload.

It was also basically a giant e-mail list generator. You got added to a newsletter so the Pres could 'reach people'.…
Read 3 tweets
27 May 20
First post is up for SIO’s Virality Project, a study of info dynamics surrounding COVID. 1st topic: the recasting of Judy Mikovits -a 👩‍🔬 best known for a retracted paper- as whistleblowing influencer. A case study in how marketing & misinfo can intersect.…
Plandemic was one video in a string of cross-platform promotion efforts. This post explains what happened. We classified 16k pages & groups that generated 41,662 posts about Mikovits, looking at post & interaction dynamics in aggregated communities. (There are interactives!)
We cover the pre-Plandemic period, when most of the activity was in a handful of echo chambers: shared videos by blue-check Facebook and Insta accounts, and YouTube creators with large audiences. Then we look at the explosion of communities and activity following Plandemic.
Read 10 tweets
26 May 20
Hm. IIRC some of us were writing about 'radicalization via the recommendation' engine at the time...which was largely dismissed because there was little in the way of quantitative evidence.
May 11, 2016, article written based on my own personal observations, which were clearly anecdotal but also bizarre. Getting Pizzagate group recos after never typing the keyword, but after joining an anti-vaccine group, was a hell of a thing.…
It seems FB began to promote Groups while aware of this dynamic? Outside observers had very little in the way of quantitative evidence but saw really disturbing intersections between fake trending stories and group sharing dynamics.…
Read 3 tweets
25 May 20
Lots of high-profile political & media discussion of story suggesting that 1/2 the tweets ab #reopen are bots... Bots began to captivate politicos/ public debate ~2yrs after their effectiveness heyday. Researchers had moved to looking at multiple factors to gauge manipulation.
There are still lots of fake accounts on Twitter & FB; company transparency reports are a good place to see the #s companies are taking down. But even so, they don’t generally achieve the kind of impact they used to in 2016; things like weights in trending algos have changed
Unfortunately a lot of the bot-detection classifiers overindex on certain factors that particular groups think of as activism - rapid, high volume retweeting, for example. Saying the same thing. Bombarding a targeted hashtag or person.
Read 6 tweets
23 May 20
The policy change that Facebook made in 2019 during the Brooklyn measles outbreak involved splitting HEALTH MISINFO ab vaccines off from POLITICAL DEBATES about vaccination.

"vaccines cause autism" - no.
"school vax requirements are govt tyranny" - ok…
I have an insanely long paper on this in the works but: the gist was that this would allow preservation of political debate about vaccination legislation, which has been incredibly heated (CA, NY, NJ within the last year)...
...while stopping the absolutely macabre and wholly made-up "vaccines cause SIDS" insanity by Larry Cook of the group Stop Mandatory Vaccination, who was running GoFundMes and then ads targeting pregnant women with pictures of deceased children.…
Read 8 tweets
21 May 20
In a post that is simultaneously delusional and a glimpse at things to come, here’s an example of how the antivax mvmt will attack any COVID vaccine in development, then extrapolate out in an attempt to also undermine confidence in demonstrably safe, decades-old school shots
His Instagram post appeared as an article on Ft Russ.
With these totally normal tags.
Read 3 tweets
17 May 20
In 2015 I gave a talk at Berkeley School of Public Health along w/ @DrPanMD & @VaccinateCal ab how credible public health voices were largely absent from Twitter conversation. As usual, antivaxxers showed up to record, heckle, selectively edit, post to YT, share to FB, & tweet.
There are so many layers to this problem. There’s the fact that many, many doctors who DO speak up become targets, so it’s not an easy call. Some of the doctors who made Plandemic debunking videos on YT got threats or had their channels mass-reported.
Others public health orgs post, but see almost no uptake in the content unless they pay to boost posts - and most don’t really have the budget for that. The algorithm curates based on Consensus of the Most Liked, and that still rewards sensationalism, outrage, anger, conspiracy.
Read 5 tweets
14 May 20
Interesting new paper out in Nature today! Paper finding: anti-vaccine groups are more adept at interlinking with "fence-sitter" communities and unrelated communities than pro-vaccine groups. Here's why this matters...…
In 2015, @gilgul and I looked at the networking of these communities on Twitter, observing the early indications that anti-vaxxers were moving away from toxins & autism narratives in favor of "medical freedom" and "choice" narratives (beautiful image his)…
At the time there was no really good way to do that style of work on FB; this new paper did it on Facebook and it's worth a read to understand the dynamics. Good summary here:…
Read 10 tweets
13 May 20
Enjoying this review of Plague of Corruption, which for some reason Amazon continues to classify within the "Science and Medicine", "Pathology / Forensic Medicine" "Biology", "Chronic Fatigue Syndrome," and "AIDS" book categories.…
One of my favorite things to do with Amazon books is look at the cross-promotions in the various recommendation engines. Sponsored products: books about colloidal silver and the New World Order
"Customers who bought this item also bought" is a mishmash of Q-anon conspiracy stuff, antivax classics like Andy Wakefield's book about autism, misc content about the war to destroy Christian America, Dinesh D'Souza, Owens, and a book about how lemons can save your life.
Read 5 tweets
9 May 20
🧵I’ve alluded a bit to how Mikovits was boosted for weeks as part of a book promotion effort before her breakout "success” in the Plandemic clip. The NYT's @daveyalba breaks down the amplification & PR/marketing aspects of the story clearly here:…
Plenty of experts have been working hard to refute the specific pseudoscience in the video clips so I'm not going to discuss that. What I & some other researchers had been looking at for 3 weeks prior were efforts to turn her into a counter-Fauci 'whistleblowing' influencer.
Marketing techniques are used to spread ideas, grow influencers, etc. It's what social media was designed for. There are 3 rough components to a successful social media influence campaign: content creation, dissemination, amplification.
Read 17 tweets
7 May 20
At this point, this is absurd. More to come, but: a variety of groups have attempted to meme Mikovits into an authoritative influencer for *weeks* now. Many indications it was picking up velocity. Remove is the wrong call; fact-check the thing. Use "Inform", and not 36hrs late.
ALSO: the video has so *much* health misinformation in it that people on Reddit have done literal minute-by-minute refutations at this point. Of a 25 minute video.…
ALSO: the distribution of that particular video has only sped up in the last 36 hours, particularly since ~2pm yesterday; curious if "Reduce" was used to throttle at any point...seem it should be used w/"Inform" to buy fact-checkers time, if the content doesn't justify removal.
Read 4 tweets
7 May 20
New thing in The Atlantic today about the complex challenge of surfacing authoritative info in era of fragmented trust, weak institutions, gameable system. Automated curation isn't going anywhere; how we structure it is 🔑 to healthy info ecosystem. 1/…
Automating curation based on engagement isn't delivering "wisdom of the crowds". It's Consensus of the Most Liked: content surfaced because ppl like the presentation of the content. Early popularity, often orchestrated, begets amplification, which begets more popularity. 2/
Information space is no longer dominated by institutional authorities (good); institutions are struggling (not good). They're not adapting to how people expect to receive info. Our realtime expectations conflict with the pace of science, but also, CDC/WHO have really poor comms.
Read 6 tweets
3 May 20
The CDC and WHO do a generally bad job of producing content for the social media age (article on that to come!) BUT they are also constrained by a need to hew to facts and reality, which Epoch - with its social media booster networks of fake ppl - is not.…
Epoch is one of a range of entities that skate through enforcement gaps and policy inconsistencies between social platforms. FB punishes them for their scam on FB but they keep advertising on YouTube. YT videos & links are shared into sympathetic FB Groups. Etc.
There are interesting questions to debate here, bc while having that policy variability is in some ways good - preserves platform community character - it does mean that attempts to address problematic situations are less effective.
Read 4 tweets
3 May 20
Press coverage of this topic has got to get clearer about what happened to make these groups connect: they distrust government. Distrust is the unifying theme. The CA protests are a mashup of antivax, Q, MAGA, militia, sovereign citizen groups...…
and what they have in common is the belief that some facet of the govt is lying to them/trying to control them. The anti-vaxxers, adopted the language of "freedom" because "vaccines cause autism" wasn't palatable to the mainstream. We saw this in **2015**…
That article documents the direct outreach from CA's antivaxxers to the Tea Party, at the around the idea that the CDC was instituting a vast cover up of vaccine harms; in 2015, they worked to connect w/the subset of 2A activists who believed Obama was going to take their guns.
Read 7 tweets
1 May 20
CA protests in Sacramento today, co-organized by several anti-vax groups, which included planned arrests, have now *had* such arrests (imagine that), and the FB Live spin is happening.…
As with the mom arrested at the playground in Idaho, which was planned, one of the narratives is that 👮 should ignore the law. Some CA anti-vax group leaders began to use wording and themes from sovcit groups last year during protests related to school vaccine requirements.
if you don't like a law or a decision doesn't go your way, you can "research" yourself into deciding that the whole government is illegitimate and people will be there to support you...
Read 5 tweets
29 Apr 20
SFUSD board meeting on adopting an official grading policy is happening now. The board advocated for the "give them all A's" nonsense that went viral. District admins, including the Superintendent, presented a meticulous report on why A's was a bad idea, advocated for P/NC...
Student advocates ran their own focus group. 5000 students responded. They, too, recommended strongly against "#TeamAllAs" b/c of the impact on their college admissions & credit. Here is the district report... So how did the School Board ideologues react?…
The first member to speak, Ali Collins, is quibbling about the Superintendent's research. She is demanding evidence not that grades reflect reality, now, but that grades reflect reality DURING A PANDEMIC. She wants to know why there is no cited data FROM A PANDEMIC. (I'm dying)
Read 5 tweets
27 Apr 20
Back in 2016 I gave a talk at UC Berkeley School of Public Health on the absence of authoritative health/medical/pro-vax voices on Twitter. In Q&A some of the MPH students asked if they should get on Twitter. At the time I felt vaguely silly saying yes...…
Because 1) at the time I was getting tons of coordinated harassment over the CA vax law (trolls posting photos of me, of my baby, following me, etc) and felt like ‘who in their right mind encourages others to get in that game’ and 2) surely they had real “medical” work to do.
This article does a great job explaining why this kind of expertise is so needed - the value of having more authoritative voices on Twitter, and having content to counter data voids in search results and conversation topics related to health.…
Read 5 tweets