We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 27
WHISTLEBLOWER
ZO AID
Report government and corporate lawbreaking.
Without breoking the law.
- REDACTED FOR CONGRESS -
—ANONYMOUS WHISTLEBLOWER DISCLOSURE—
|
SEC Office of the Whistleblower
Via Online Portal & Fax
Re: Supplemental Disclosure of Securities Law Violations by Facebook, Inc. (NASDAQ:
FB), SEC TCR
Facebook misled investors and the public about its role
perpetuating misinformation and violent extremism relating
to the 2020 election and January 6th insurrection.
To the SEC Office of the Whistleblower:
1. The instant letter is one of multiple disclosures related to the above-captioned
matter. Our anonymous client is disclosing original evidence showing that
Facebook, Inc. (NASDAQ: FB) has, for years past and ongoing, violated U.S,
securities laws by making material misrepresentations and omissions in
statements to investors and prospective investors, including, inter alia, through
filings with the SEC, testimony to Congress, online statements and media stories.
2. On August 26, 2021, the U.S. House of Representatives Select Committee to
Investigate the January 6 Attack on the United States Capitol requested from
Facebook:*
All internal or external reviews, studies, reports, data, analyses, and
related communications regarding your platform(s) and:
i. Misinformation, disinformation, and malinformation relating to the
2020 election;
Whistleblower Aid is a U.S. tax-exempt, 501(c)(3) organization, EIN 26-4716045.
hittpsi/WhistleblowerAic.org — I Hl
Contact via SecureDrop over Tor: http://whistlebloweraid.s
= 2) of a link or a photo,
we find they are 4 times more likely to be seeing misinfo compared to when
they see links or photos on News Feed in general. This increases to 5-10X at
higher reshare depths. . . 30-70% of misinfo viewership occurs on reshares of
reshares . . . In the US, that’s 65% for photo misinfo, and 35% for link misinfo
. The reshare depth of a post is defined by how many reshares away it is
from the original post. . . political operatives and publishers tell us that
they rely more on negativity and sensationalism for distribution due to
recent algorithmic changes that favor reshares."”
» [Comparing the effects of misinfo rom polticians vs ordinary user sources, p 1
» 2020-08-18 Poca Influence on Content Poly. 9.4.
» Effects of Poltcian Shared Misinformation. p. 34
» eshare ranking exp in nda Indonisia,p.1.&;soe also NN Reshare depth inca an indonesia.
» [Deep Reshares end Misinformation 3.
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
13WHISTLEBLOWER
- REDACTED FOR CONGRESS - ZOAID
Report government and corporate lawbreaking.
‘Without breaking the law.
“A ranking change which reduces ranking based on max reshare depth
produces significant wins on a variety of integrity measures .. . Stemming
from the initial finding that reshare depth is correlated with misinformation,
subsequent research has found that other integrity harms also correlate
with reshare depth... °°
“70% of VPVs [View Port View, Facebook's term of art for a user impression]
‘on known link misinfo come from reshares . . . 50% to 80% of VPVs on
known link misinfo come from the top 1K most seen links per day .
Reshare depth appears to be an especially good signal for targeting link
misinfo in India and the Philippines. . . misinfo prevalence is 15-25X higher
among VPVs of depth 2+ reshares. ..”
“An effective, content-agnostic approach to mitigate the harms posted
by high-harm misinfo (e.g. civic or health) would be to dampen virality within
these topics by hard demoting all deep reshares where the viewer is not
a 0
“Political parties . . . claim that Facebook’s algorithm change in 2018 (MSI)
has changed the nature of politics. For the worse. They argue that the
isis on ‘reshareability’ systematically rewards provocative,
low-quality content.”"’
“fin Myanmar] Misinformation, misrepresentation and account issues are a
few known high-risk abuse areas that will be exacerbated by a second COVID
outbreak ... The Reshare Depth demotion reduces the distribution of
highly-viral content to give more distribution to content produced by friends
or connections of friends . .
[N.B.: Measures enacted to demote reshare depth to reduce the distribution
of highly-viral content in fact reduced Viral Inflammatory Prevalence by
25.1% and Photo formation by 48.5% in Myanmar, Facebook sai
We plan to roll back this intervention after the Myanmar election in
November."
5 [Max Reshare Depth experiment, p. 1-2.
2 Further reading from Reshare Depth Article, p. 2.
a Fighting high harm misinfo with deep reshare damping, p. 1; see also MEE Resinare depth by
Pottical Party response to the 18 Algorithm change, p. 1.
JReshare depth exp in Myanmar, p. 1
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
14WHISTLEBLOWER
- REDACTED FOR CONGRESS - ZOAID
Report government and corporate lawbreaking.
‘Without breaking the law.
25. Facebook's failure to limit reshares is incredibly risky because under one percent of
users drive haff of all the user impressions [called “VPVs"] on reshare posts:
“Considering the 28 days from February 22 - March 21, 2021
2.82 B users were active on Facebook.
978 M users created at least one reshare post (34.6%).
37 M users (1.3% created half of all reshare posts.
21 Musers (0.7%) created reshapes accounting for half of all VPVs on
reshare posts (within three days)."®
26.To avoid criticism resulting from inevitable false positives when removing harmful
content, Facebook chooses to “demote” it instead, which it knows to be an
ineffective response:
“Due to the unboundedness of ranking scores, neither global all demotions
nor viewed term demotions can effectively re-rank some content...[If we have
some signal that a piece of content is low-quality or will cause a bad
experience and we want to counter the engagement score, we have to apply
a pretty big demotion . . . How big would the integrity demotion on global all
have to be for a piece of content to be ranked X places lower? ... some
4 _
90% ... .Let me make this last point clearer. If you demoted the
top-ranked VPV by 10%, 74% of these VPVs would maintain their top spot.
If you demoted the top-ranked VPV by 50%, 27% would be unmoved. 4.5%
of VPs are impervious to even a 90% demotion.”
27. Facebook has avoided or rolled back interventions for “groups” and “narrow
subpopulations” that it knew reduced misinformation, violence and incitement,
because those interventions reduced the platform's growth:
“Harmful communities have been seen to grow via using our platform
affordances . .. the @Anon community relied on minimally-connecte
less than 5 months! . . . The civic integrity team explored a variety of
changes to the way civic content was ranked during US 2020 and identified
several features which promoted a more healthy ecosystem . ..™°
"Quantifying the Concentration of Reshares and their VPVs Among Users" p. 1.
[Demotions in practice, p. 5, 9-10.
Harmful Non- Violating Narratives, p.8, 10, 15, 19, 20.
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
5WHISTLEBLOWER
- REDACTED FOR CONGRESS - x AID
Report government and corporate lawbreaking.
‘Without breaking the law.
[However] “We have rollk k election rate limit of 1
i ‘lhe ” i Gi =
[which was implemented in October 2020 to reduce] joins to sensitive groups
or groups that became sensitive.”
“We've seen that integrity harms tend to be concentrated in a small
number of User Interest Subpopulations, with 50% or more of a given harm
often in 1% or less of Subpopulations . . . More than 50% of civic
misinformation on Facebook is associated with a handful of highly-civic
communities. . . Misinformation on Instagram is also disproportionally
concentrated among a few Subpopulations. . . Over 50% of QAnon & MSM
(Militarized Social Movements) Groups are associated with a single
Subpopulation . ."”
[March 2020] “{There are likely meaningful opportunities for impact via
integrating Segmentation with either Voter Suppression or Hate Speech []
The majority of US Groups posts containing Voter Suppression over the past
2 months occurred in Groups associated with just 2 sociographic segments [
|] The Groups in segments with the highest prevalence experience 100x as
many borderline hate speech posts as the overall average.”®
[April 2020] “Disbelief comments are produced at disproportionate rates on
civic content posted by extreme partisans. This suggests content posted by
extreme partisans might be more likely to be false . . "
Facebook's own tools are used to direct narrowcast misinformation:
“Up to 64% of IRA [Internet Research Agency, a Russian troll farm] ads
nay have
en targeted using Facebook's automated targeting
[From August 2020] “Many Users Are Repeat or Serial Offenders For Both
Misinformation and Hate . . . surfaced a connection between misinformation
and hate within the context of highly polarized communities along with
patterns of serial offending and potential recidivism . . . There may be
substantial intersections between hate and misinformation, which could be
lswitch Plan for all Group Recommendation Surfaces, p. 3
ubpopulations: Segmentations Wiki, p. 1, 19-21
‘ociagraphic Segments may be impactful for hate speech and voter suppression, p. 1
identifying Narrowcast Misinfo Hard because missing or miscalibrated doubt, p. 3
[Measuring Human Perception to Defend Democracy, p. 24
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
16WHISTLEBLOWER
- REDACTED FOR CONGRESS - x AID
Report government and corporate lawbreaking.
‘Without breaking the law.
an area for intervention . . . 59M users have posted or shared at least 3
pieces of predicted misinformation in the past 90d. (l)... 99% of these
user accounts remain active, and some of them have passed dozens of
authenticity checks.”*'
2020 Election & Capitol Insurrection
28. In April 2020, removing “Downstream MSI!" from Facebook's Meaningful Social
Interactions algorithm (the News Feed's goaling mechanism) was suggested as a
“soft action" that would significantly decrease the amount of hateful, polarization, or
violence inciting content from the News Feed without doing any content based
censorship. CEO Mark Zuckerberg rejected this intervention that could have
reduced the risk of violence in the 2020 election:
“Downstream model depreciation: Mark doesn’t think we could go broad .
za Mt the in ic
misinformation] if there was a material tradeoff with MSI impact
[‘meaningful social interactions” ."°
Note: The "Meaningful" part of Meaningful Social Interactions is more a
flourish than substance - for example, all comments were considered
"meaningful", even if they were hate speech or bullying (for six months after
this memo). All likes were considered "meaningful" along with all reshares.®
29, Facebook's records state:
[August 2020] “We should avoid making significant changes to the IFR [in
feed recommendations] publisher ecosystem prior to the US 2020 election, if
those changes will have a large impact on entity-level aistribution for political
figures, commentators, or news organizations in the US... . FINAL DECISION
. » we have decided to not make any changes until the election is over.
We will revisit the issue after the election."**
[August 2020] “I've seen promising interventions from integrity product
teams, with strong research and data support, be prematurely stifled or
severely constrained by key decision makers--often based on fears of
ublic and policy stakeholder responses. . . . For example, we've known
a Serial misinfo and hate offenders, p. 2.3.
a Mark Feedback on Soft Action Proposal + Deck presented fo Mark, p. 1.
8 "Filtering out engagement-bait, bullying and excessive comments from MSI Deltold metric 11/1 onwards"
“ (Update On Poitical Publisher Issues, p. 1-2.
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
”WHISTLEBLOWER
- REDACTED FOR CONGRESS - ZOAID
Report government and corporate lawbreaking.
‘Without breaking the law.
for over a year now that our recommendation systems can very quickly lead
users down the path to conspiracy theories and groups. [] While the
Recommendations integrity team has made impressive strides in cleaning up
our recs, FB [Facebook] has been hesitant to outright ban/filter
conspiracy groups like QAnon until just last week. [] In the meantime, this
fringe group/set of beliefs has grown to national prominence with QAnon
congressional candidates and QAnon hashtags and groups trending in the
mainstream. We were willing to act only “after* things had spiraled into a
dire state.*
30. Thus, Facebook allowed misinformation and harmful content to persist.
“Through most of 2020, we saw non-violating content promoting QAnon
spreading through our platforms. Belief in the QAnon conspiracy took hold in
multiple communities, and we saw multiple cases in which such belief
motivated people to kill or conspire to Kill perceived enemies . . . Policies
don't fully cover harms [] We implement policies for many of these areas that
limit our ability to act. . . high-profile entities were able to serially spread... .
claims without crossing our falsifiable misinformation based lines for
enforcement. . . we've often taken minimal action initially due to a
combination [of] policy and product limitations making it extremely
challenging to design, get approval for, and roll out new interventions quickly.
Afterward, we've often been prompted by societal outcry at the resulting
harms to implement entity takedowns. For instance, we've enacted this
pattern with both QAnon and delegitimization, wherein we initially took
limited or no action, and later decided to take down Groups, Pages, and
even Users supporting these movements . . . Bringing problematic content
prevalence [conspiracies about COVID and vaccine discouragement] in the
top 2% of communities in line with average communities could reduce the
size of the overall problem by up to ~80%. Based on our experiences with US
2020 election delegitimization, QAnon, and Dangerous Orgs, there are at
least three conceptual roles we need to address... ."*
31.Only after and in response to the Capitol protest and public pressure did Facebook
take certain steps related to the above items on a temporary basis, such as:
® Badge Post - DS Misinfo, p. 2-3.
“| Harmful Non-Violating Narratives, p.8, 10, 15, 19, 20
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
18WHISTLEBLOWER
- REDACTED FOR CONGRESS - x AID
Report government and corporate lawbreaking.
‘Without breaking the law.
“Immediate response to Val [violence and inciting] risk in DC [] US2020
Levers, Previously Rolled Back . . . Increased [] ability to action more .
Applied demotion to content deemed likely to violate our Community
Standards in the areas of hate speech, graph violence, and violence &
incitement. . Made users less likely to see content we expect them to report
Require Admins review and approve posts in civic groups that
accumulate four strikes . ..”
“Reduced the strength of a LIVE video boost that was contributing to
ultra-rapid virality for low-quality videos . . . removed 10 product boosts for
civic and health content . . . Capped the strength which videos could be
boosted . . . Explicitly include delegitimizing terms (ex. stop the steal) into the
civic classifier .. . Lowered] threshold to demote hateful comments . . .
Dropped the Groups invite rate limits from 100 to 30 to address evidence of
concerning groups building growth through mass invitations . .."°”
32. Internal analysis also noted:
“Early January 2021 had three events on the political calendar that deserved
attention .. . we anticipated new, narrow-topic misinfo stories to appear
and spread rapidly . . ."*
IG iG ineteorar ve Capitol | Hill Violence | Data Oe = ta ees aad
Teump or were videos of Donald Trump. Several of the top (reported) posts
called for [violence], suggested the overthrow of the government would be
[desirable] . . Top Reported User... . President Donald J. Trump [with 121
reports, compared with the second highest at 46 reports}.’?
“[We should have adapted already long ago .
the Steal groups active up until yesterday .
resources we have, we should do better.”
“We've been fueling this fire for a long time and we shouldn't be surprised
it's now out of control.”
There were dozens of Stop.
. With the unprecedented
“[DJo you genuinely think 24 hours is a meaningful ban? You mention the list
of things we've changed in the past few years but how are we expected to
Capitol Rots Breaks the Glass, pp 1-4
sponse to events in early January 2021 - lessons for algorithmic interventions, p. 3-4
“| 1G Jan 6 Capitol Riot Violence Update p. 1-2.
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
w- REDACTED FOR CONGRESS -
WHISTLEBLOWER
OND
Report government and corporate lawbreaking.
‘Without breaking the law.
“A 24-hour ban is nothing for Trump. His proxies on social media can easily
keep the fires going during that time, and he will return...”
“How about refusing to take another dime of Trump's advertising money? | do
acknowledge that a 24-hour ban is a pretty big deal, but that’s only because
up until now, our response has been completely tepid.”
33. Facebook acknowledged that how they resolved tradeoffs was dangerous, which is
why they chose more conservative solutions for the US 2020 election. In contrast to
their claims to their advertisers that they done all they could do to prevent the
insurrection, they reverted back to these safer defaults only after the Insurrection
flared up:
“Action: CAP LVEQ (Live Video EQ) boost linearly for civic videos at max
83.5x instead of 850x - Return to original US2020 threshold
Description: Reduced the strength of a LIVE video boost that was
contributing to ultra-rapid vitality for low-quality videos."*"
'US2020 [election] Levers, Previously Rolled Back.
ID Name
[Description |
\Val (Violence and Inciting): Cluster
Increased our ability to action more similar
| English hate speech
|- Spanish hate speech
'- Graphic violence
- Violence & incitement
RE45 [Val ignores and drive specialist [content in bulk, based on SME (subject matter}
(PResc) Review at the cluster tlevel expert) review
Use classifiers to demote content
starting with a 5% demotion for [Applied demotion to content deemed likely to
[content that has a 5% chance of \violate our Community Standards in areas of
pe |veleting lhate speech, graphic violence, and violence &
incitement; the demotion's strength is in
lproportion to how confident we are that
|content is violating,
|Comments on Zuckerberg’s response to capitol riots, p. 4, 5,8,
[Capitol Riots Breaks the Glass, pp 4
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
2REDACTED FOR CONGRESS -
WHISTLEBLOWER
OND
Report government and corporate lawbreaking.
‘Without breokng the law
Increased the max strength of linear probable
esq {V8I: Increase demotion strength cap_|volating Val demotion (PEt), allowing us to
from 50% to 80% further downrank higher confidence (50-80%
lconfidence) Vl content.
'V&l: Increase strength of p(report) to [Made users less likely to see content we
E48 _|reduce visibilty of V&l posts likely to jexpect them to report to Facebook,
be reported laddressing subjectively problematic content.
[Require Admins review and approve posts in
leivie groups that accumulate four strikes
[strikes require admins to have ereated or
‘Turned on mandatory post approval
ee, |p eens lapproved the violating content) as well as 57
eet fast-moving new groups, helping ensure that
ne |Admins are properly policing and being held
laccountable for serious violations during the
election
[Prevents commenting on Groups posts that
teh tial sos wil |show a high rate of hate speech and violence
incitement in associated comments,
‘groups that start to have a high rate u
PE17 laddressing concer that group Admins and
lof hate speech and violence and na
Imembers are "baiting violating material and
incitement comments. . :
coordination of harm via otherwise
lnon-violating posts.
[Updated pre-emptive lever to disable
'V8l: Add reduced criteria to trigger [comments in group threads (PE17) that were
ESS comment auto-disable in group starting to have too much hate speech or
|comment threads (extension of PE17) |violence & incitement to better catch threads
leary on.
[Prevents groups from changing names to
Prevent groups from changing their [8 28s0ciated with recent attempts to
RES eaves a a a Pa a " ldelegitimize the US election results (Stop the
gamer’ |Steal), helping to slow down abusive
laudience building
[Prevent Abusive Audience Building by
ire-naming and re-purposing/arowth existing
Ress [Preventing pages fom changing thelr Jpages associated with recent attempts to
ali \delegitimize the US election results (Stop the
Steal)
a Fier entities with names matching "stop the
REs7 _[Fiter deleattimizing entities from | 2" regex from recommendation surfaces
recommendations
lacross FB inc.
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
aWHISTLEBLOWER
OND
Report government and corporate lawbreaking.
‘Without breaking the law.
- REDACTED FOR CONGRESS -
[Addressed challenges with enforcing existing
|scaled policy and a spike in violations of
lescalations-only subsets of the V&l policy by
Ire-reviewing and deleting more user reports.
'SEt up a special re-review queues for
RE46 [high risk contents from our user
reports with Risk & Response
Increased the volume of user reported content
\V8l: reduce reactive auto-delete lon the platform we were auto-deleting for
{threshold from P90 to P70 \viotating our V8 policy, allowing us to
Imanage the influx of repors.
REAM
TE
34. Invoking immunity under Section 230: Facebook has attempted to claim
“immunity” for content under Section 230 of the Communications Decency Act, as
it did in a recent court case, arguing:
“Section 230 requires dismissal of lawsuits like plaintiffs’ [victims of sex
trafficking on Facebook and Instagram] that seek to impose liability on
interactive computer services for harmful content posted or communicated
on their platforms by third parties .. . Section 230’s plain text makes clear
that it immunizes interactive service provides [i.e., Facebook] from suit, not
just from liabilty.""°
KNOWLEDGE OF MATERIALITY
35. Facebook has admitted investor risks with thes
Facebook's recent 10-K™ acknowledged:
“Our brands mi us negativel; the actions of thai
are deemed to be hostile or inappropriate to other users, by the actions
of users acting under false or inauthentic identities, by the use of our
products or services to disseminate information that is deemed to be
misleading (or intended to manipulate opinions) . . .”
issues. For instance,
36. Likewise, Facebook's 10-Q* stated:
8 EIEN C2p101 Rts Breaks ine Glass, pg. +3
Facebook's Petition for Writ of Mandamus, In re Facebook, Inc. and Facebook, Inc. d/bva Instagram, Cause Nos.
2018-69816 & 2018-82214 et seq., (Supreme Court of Texas, Filed May 29, 2020).
StpsilsecreportiDacumenv0004326801-2-0000141
‘*hitpsileec.reporvDocumentv000132680!-21-000049),
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
2WHISTLEBLOWER
- REDACTED FOR CONGRESS - ZOD
Report government and corporate lawbreaking.
‘Without breaking the law.
“In addition, we have been, and may in the future be, subject to negative
publicity in connection with our handling of misinformation and other
illicit or objectionable use of our products or services, including in
connection with the COVID-19 pandemic and elections in the United States
and around the world. Any such negative publicity could have an adverse
effect on the size, engage: and loyalty of our user base and
marketer demand for advertising on our products, which could result in
decreased revenue and adversely affect our business and financial
results, and we have experienced such adverse effects to varying degrees.”
“For example, we have been the subject of significant media coverage
involving concems around our handling of political speech and advertising,
hate speech, and other content, and we continue to receive negative
publicity related to these topics. ... Any such negative publicity could
have an adverse effect on the size, engagement, and loyalty of our user
base and marketer demand for advertising on our products, which could
result in decreased revenue and adversely affect our business and
financial results, and we have experienced such adverse effects to varying
degrees from time to time.”
37.Role for the SEC. The SEC is charged with enforcing the laws that protect
investors in public companies like Facebook. Facebook’s investors care about
misrepresentations and omissions by Mark Zuckerberg and other Facebook
‘executives on the topic of misinformation relating to the insurrection for two
reasons. First, to the extent that users become aware of the dangers that Facebook
platforms present, they are likely to use the platforms less, leading to lower
advertising revenue and lower profits. Second, some investors simply will not want
to invest in a company that facilitates misinformation and violence on and off of the
internet and then engages in misstatements and omissions on the topic.
38.Whistleblower Aid is a non-profit legal organization that helps workers report their
concerns about violations of the law safely, lawfully, and responsibly. We
respectiully request the SEC’s assistance ensuring that our client never faces
retaliation
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —
2WHISTLEBLOWER
- REDACTED FOR CONGRESS - ZOAID
Report government and corporate lawbreakin,
Without breoking the aw
39.On information and belief, none of the documents enclosed here constitute
attorney-client communications, were obtained during a meeting with an attorney,
or otherwise indicate that they are in any way privileged.
40.We plan to continue supplementing this disclosure with additional information and
evidence. Our client would be happy to meet with investigators at your
convenience. Please feel free to contact us using the information below.
41, We are representing an anonymous whistleblower who is making the above
disclosures solely for reporting the suspected violation of laws as outlined.
Sincerely,
yee
John N. Tye, Attorney at Law
Chief Disclosure Officer
eT
Andrew Bakaj, Attorney at Law
Of Counsel
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —WHISTLEBLOWER
- REDACTED FOR CONGRESS - ZO AID
Report government and corporate lawbreakin,
Without breoking the aw
Enclosures: (ERE -acebo0k documents including —
(EB Re share depth exp in Myanmar
HER Capito! Protest Break The Glass
(RR A Risk Countries Planning
(ERE Comments on Zuck's response to capitol riots
TER Capito! Riots Break the Glass
EER 1G. Jan 6 Capitol Riot Violence Update
Ee
ER Sermtul Topic Communities
HER Coordinated Social Harm
EERE Visinfo got worse in 2019
ERR Harmiul Conspiracies
GER | otus Mahal
HER Aciversarial Harmful Networks - India Case study.
GER What is Collateral damage?
ER Anicrew Bosworth Thoughts for 2020
(ER Measuring Human Perception to Defend
Democracy
(ER Diverse Engagement May Identify Valuable Civic
Comments
EERE Visinfo in comments
RR They used to post selfies, now they're trying to
reverse the election
ERE Comparing the effects of misinfo from politicians vs
ordinary user sources
— ANONYMOUS WHISTLEBLOWER DISCLOSURE —