Countering False Information on Social Media in Disasters and Emergencies, March 2018

Countering False Information on Social

Media in Disasters and Emergencies

Social Media Working Group for Emergency Services and

Disaster Management

March 2018

Contents

Executive Summary ...................................................................................................................... 2

Intr

oduction

................................................................................................................................... 2

otivations

.................................................................................................................................... 4

roblem

......................................................................................................................................... 5

auses and Spread ................................................................................................................... 6

Incorrect Information .............................................................................................................. 6

Ins

ufficient Information

........................................................................................................... 7

pportunistic Disinformation

.................................................................................................. 8

utdated Information

............................................................................................................. 8

ase Studies

............................................................................................................................... 10

Incident Name: 2014 South Napa Earthquake ........................................................................ 10

Inc

ident Name: 2016 Louisiana Floods

................................................................................... 11

Inc

ident Name: 2017 Oroville Dam Evacuation

....................................................................... 13

uggested Best Practices

........................................................................................................... 14

st Practices - People

........................................................................................................... 15

Best Practices - Processes ..................................................................................................... 16

est Practices – Technology

................................................................................................... 17

dditional Considerations

........................................................................................................ 18

hallenges

.................................................................................................................................. 18

onclusion

.................................................................................................................................. 19

Executive Summary

Rumors, misinformation and false information on social media proliferate before, during and after

disasters and emergencies. While this information cannot be completely eliminated, first responder

agencies can use various tactics and strategies to offset bad information. This white paper examines

motivations people may have for sharing false information, discusses underlying issues that cause false

information and offers case studies from recent disasters to illustrate the problem. Multiple motives

lead people to post false information on social media: some posters seek a particular result, such as

closing schools for the day; some desire to get attention with a dramatic post; some are pushing a

money-making scam or political agenda; and some innocently repeat bad or outdated information.

Best practices for agencies to counter misinformation, rumors and false information are detailed and

categorized in this white paper, and challenges and additional considerations are presented for review.

This report illustrates methods of countering false information on social media with case studies:

• The 2014 South Napa earthquake: Tweets were filtered by geolocation to eliminate posts from

trolls.

• The 2016 Louisiana floods: The Red Cross published and shared a blog to counter rumors and

misinformation about food distribution and shelter policies.

• The 2017 Oroville Dam evacuation: An accidentally misleading tweet suggested the evacuation

area included all of Sacramento County. Local agencies used traditional and social media to

provide correct information.

Examples of best practices include:

• Establishing partnerships with local traditional media outlets before disasters, so means exist to

disseminate accurate information;

• Using the Joint Information System to coordinate public information efforts of multiple

jurisdictions and agencies; and

• Setting up a central website to debunk bad information.

Introduction

Social media and collaborative technologies have become critical components of emergency

preparedness, response and recovery.

From international response efforts after large-scale disasters to

domestic response and recovery after events affecting the United States, many government officials

now turn to social media technologies to share information and connect with citizens during all phases

Social media includes any online or digital medium provided and/or collected through a channel that enables the

two-way sharing of information, involving multiple parties. This includes social networking sites, texting, blogs, etc.

of a crisis. Implementing these new technologies, however, requires responding agencies adopt new

communication strategies, policies and engagement methods.

Recognizing the need to address these challenges, the U.S. Department of Homeland Security's (DHS)

Science and Technology Directorate (S&T) established a Virtual Social Media Working Group (VSMWG)

in 2010. After Public Law 114-80 was passed, the VSMWG was re-named as the Social Media Working

Group for Emergency Services and Disaster Management (SMWGESDM). The mission of the

SMWGESDM is to provide recommendations to the emergency preparedness and response community

on the safe and sustainable use of social media technologies before, during and after emergencies. The

SMWGESDM is a subcommittee of the Homeland Security Science and Technology Advisory Committee

(HSSTAC). The HSSTAC approved the recommendations contained in this white paper by consensus vote

on February 22, 2018.

Drawn from a cross-section of subject matter experts from federal, tribal, territorial, state and local

responders from across the United States, SMWGESDM members establish and collect best practices

and solutions that can be leveraged by public safety officials and responders throughout the nation’s

emergency response community. Below is a list of agencies and organizations to which the SMWGESDM

members belong.

SMWGESDM Member Agencies and Organizations as of March 2018

● A

merican Red Cross

● Argonne National Labs, Public Affairs

Science and Technology Fusion Center

● A

rlington County [VA] Fire Department

● C

alifornia Governor’s Office of

Emergency Services

● C

olorado Division of Homeland Security

and Emergency Management

● Ci

ty of Evanston [IL]

● City of Nashua [NH] Office of

Emergency Management

● Ev

acuteer

● Fe

deral Emergency Management

Agency (FEMA)

● G

eorge Mason University

● H

umanity Road

● In

dianapolis [IN] Fire Department

● Na

tional Institutes of Health

● N

ew York City [NY] Department of

Health and Mental Hygiene

● N

ew York City [NY] Emergency

Management Department

● S

acramento County [CA] Office of

Emergency Services

● Sa

cramento [CA] Fire Department

● C

ity of St. Louis [MO] Emergency

Management Agency

● U

nited States Geological Survey

● U

niversity of Washington Emergency

Management

● V

irginia Department of Emergency

Management

● W

ashington County [AR] Regional

Ambulance Authority

Motivations

Social media platforms have allowed individuals and organizations to share information with their peers

and specific audiences for more than twenty years.

Information typically is shared with good intent;

however, some people post on social media to further an ulterior agenda. Their posts may include

rumors, false information and misinformation (e.g., deception, propaganda and malicious spamming).

Researchers have identified different characteristics of social media posts that lead consumers of the

posts to believe in an alternative, fake reality and suspicious behavior.

3,4

Characteristics of false

information may include uncertainty in the “facts,” emotional exploitation of a situation, trending topic

discussions for hijacking conversations and financial scams, among others.

5,6,7

An e

xample of false information with these characteristics is deceptive content with a malicious agenda,

such as diverting a user towards purchasing a particular product.

Such campaigns are also used to lead

a user to believe in a fake negative opinion to damage an object’s reputation; for example, fake reviews

on online e-commerce websites, such as Amazon or Yelp.

Likewise, deceptive false information has

been posted in large-scale disasters for financial gain.

False information with a malicious agenda has

long existed in the form of propaganda, which has been used by terror and other extremist/criminal

organizations as a tactic to recruit.

Weblogs, or blogs, have existed since 1997, and an early example of social media being used to share information

is the website Friendster.com, which was launched in 2002. <https://en.wikipedia.org/wiki/Blog

> and

<https://en.wikipedia.org/wiki/Friendster>.

Pendleton, Susan Coppess. “Rumor Research Revisited and Expanded.” Language & Communication. 1998. 18,1:

69-86.

Jiang, M., Cui, P., & Faloutsos, C. “Suspicious Behavior Detection: Current Trends and Future Directions.” IEEE

Intelligent Systems. 2016. 31(1), 31-39.

Starbird, K., Spiro, E., Edwards, I., Zhou, K., Maddock, J., & Narasimhan, S. “Could This Be True?: I Think So!

Expressed Uncertainty in Online Rumoring.” In Proceedings of the 2016 CHI Conference on Human Factors in

Computing Systems. May 2016. (pp. 360-371). ACM.

Bessi, A., & Ferrara, E. “Social Bots Distort the 2016 U.S. Presidential Election Online Discussion.” First Monday.

2016. 21(11).

Huang, Y. L., Starbird, K., Orand, M., Stanek, S. A., & Pedersen, H. T. “Connected Through Crisis: Emotional

Proximity and the Spread of Misinformation Online.” In Proceedings of the 18th ACM Conference on Computer

Supported Cooperative Work & Social Computing. February 2015. (pp. 969-980). ACM.

Gao, H., Hu, J., Wilson, C., Li, Z., Chen, Y., & Zhao, B. Y. “Detecting and Characterizing Social Spam Campaigns.” In

Proceedings of the 10th ACM SIGCOMM conference on Internet Measurement. November 2010. (pp. 35-47). ACM.

Mukherjee, A., Liu, B., & Glance, N. “Spotting Fake Reviewer Groups in Consumer Reviews.” In Proceedings of the

21st International Conference on World Wide Web. April 2012. (pp. 191-200). ACM.

Gupta, A., Lamba, H., & Kumaraguru, P. “$1.00 per rt# bostonmarathon# prayforboston: Analyzing Fake Content

on Twitter.” In eCrime Researchers Summit (eCRS), September 2013. (pp. 1-12). IEEE.

Allendorfer, W. H., & Herring, S. C. “ISIS vs. the U.S. Government: A War of Online Video Propaganda.” First

Monday. 2015. 20(12).

When discussing the online context of false information in today’s information age, the concept of false

information driven by a motive of a deceptive agenda has existed for many decades in military

warfare.

12,13

Therefore, the strategies for countering false information with a malicious agenda in the

online environment by either coordinated efforts of humans or bots could be informed by the offline

environment as well.

14,15

Problem

One of the biggest challenges public safety

agencies and organizations face is how to reduce

or eliminate the spread of false information,

especially as public demands for a response from

these authorities increases. Social media can

distribute news faster and to a wider audience

than traditional news sources. However, that also

means the potential for misinformation, false

information and rumors to spread and go viral is

high.

16,17

A factor that may impede first

responders’ ability to mitigate and minimize the

spread of misinformation, rumors and false

information is the decreasing public trust in

government, media and nongovernmental

organizations (NGOs). While

2017 was a low point

in terms of credibility of the media, the 2018

Edelman’s Trust Barometer showed trust in

journalism jumped five points and trust in social

media platforms dipped two points. In addition,

the credibility of “a person like yourself” — often

a source of news and information on social media — dipped to an all-time low in the study’s history.

While this paper is focused on social media, responder agencies should be aware that many people still

get their news from television, which serves as an additional resource to counter false information.

Whaley, B. “Toward a General Theory of Deception.” The Journal of Strategic Studies. 1982. 5(1), 178-192.

Holt, T. The Deceivers: Allied Military Deception in the Second World War. Simon and Schuster, 2010.

A computer program that performs automatic repetitive tasks. <https://www.merriam-

webster.com/dictionary/bot>.

For future reading on this whole section, see Manheim, Jerol. Strategy in Information and Influence Campaigns:

How Police Advocates, Social Movements, Insurgent Groups, Corporations, Governments and Others Get What

They Want. Routledge, 2010.

Incorrect or misleading information. <https://www.merriam-webster.com/dictionary/misinformation>.

Madhusree Mukerjee. “How Fake News Goes Viral – Here’s the Math.” Scientific American, July 14, 2017.

<https://www.scientificamerican.com/article/how-fake-news-goes-viral-mdash-heres-the-math/>.

Pew Research Center. “Pathways to News.” July 7, 2016. <http://www.journalism.org/2016/07/07/pathways-to-

news/>.

Virality of Social Media

By Catherine Graham, Humanity Road

After the April 2015 earthquake in Nepal, a

Facebook post described 300 houses in Dhading

needed aid. The post was shared over 1,000 times,

reaching over 350,000 people within 48 hours. The

originator of this message was trying to find help

for Ward #4’s villagers via social media. Facebook

statistics show that the average user has 350

contacts, meaning this one message was viewed

by approximately 350,000 Facebook users. A week

before the viral post,

this need had already been

shared on

quakemap.org, a crisis-mapping

database built by online volunteers and managed

by Kathmandu Living Labs. On May 7, Helping

Hands (a humanitarian group) was notified, and by

May 11, Ward #4 received much-needed food and

shelter. While the late Facebook post was meant

to be helpful, the need had already been taken

care of. This short example demonstrates that

sharing outdated information can waste resources

and risk lives.

Solving the problem of how to reduce or eliminate the spread of false information requires an

understanding of the following questions.

● W

hat are the causes of misinformation, rumors or false information, and what are its

characteristics?

● H

ow does false information spread?

● What are best practices to counter the spread of false information?

Thi

s paper builds on real-world case studies of several incidents to explain and investigate answers to

the aforementioned questions.

Causes and Spread

In social media, misinformation, rumors and false information are most often caused by four underlying

issues, which are detailed more fully below:

1. I

ncorrect information - intentional versus unintentional;

2. In

sufficient information;

3. Op

portunistic disinformation; and

4. Ou

tdated information.

Incorrect Information

Incorrect information can be caused by

situations where the true situation is

difficult to confirm. Radiation in Japan was

a good example. After the meltdown at the

Fukushima Daiichi Nuclear Power Station

in March 2011, many rumors

circulated

regarding appropriate safety precautions,

such as whether people should evacuate,

the possibility of food and water shortages,

and whether there would be additional

radioactive releases (this example also illustrates insufficient information).

correct information and rumors can also be caused by individuals who wish to create confusion. One

example is when fake accounts are created that impersonate an official account. Fairfax County, Virginia

was proactive during a winter storm in January 2014 as its school system was faced with many fake

accounts announcing incorrect closures (Figure 1, above). Government and schools worked together to

actively advise people where to find official information.

Humanity Road Rumor Management Team Training, June 20, 2016.

Figure 1: Official Tweet from Fairfax County Government

addressing fake Twitter account.

Figure 2: Tweet from user @ComfortablySmug sharing a rumor.

Another example comes from Hurricane Sandy in October 2012. Twitter user @ComfortablySmug began

spreading several rumors via social media, including that the New York Stock Exchange Building was

flooded, Con Edison was preemptively shutting off power in New York City, and all bridges going to and

from Manhattan were being sealed off (Figure 2, below). Additionally, digitally altered pictures of sharks

swimming in the streets, screenshots from the movie The Day After Tomorrow and other dramatic

pictures from past storms proliferated on social media.

Incorrect information can also be malicious

(see the

Motivation section above), as with online conspiracy theorists harassing survivors of the Las

Vegas mass shooting in October 2017.

Insufficient Information

When information is slow to emerge on circumstances surrounding an event, rumors can start rapidly.

Insufficient information can be a result of several factors, such as: not having clearance to release the

data, lack of a designated official for that information, or a belief that information must be complete to

release and therefore intentionally withheld. Confusion continues to arise when official channels do not

release information fast enough, provide information updates in the right social media and traditional

media channels, or the population is unaware of or does not trust the official source for that

information. The public will generally follow and amplify official information when they can access

information they believe. This happened after the Nepal earthquake in 2015

. When there is a new

emerging situation that can be confusing, agencies will open their channels of information (such as a

conference bridge for volunteers and partners), which can be critical to avoiding mistakes in information

management.

DHS S&T Virtual Social Media Working Group. “Lessons Learned: Social Media and Hurricane Sandy”, June 2013,

p. 22. <https://www.dhs.gov/publication/lessons-learned-social-media-hurricane-sandy

Lois Becket and Sam Levin. “U.S. Gun Violence Spawns a New Epidemic: Conspiracy Theorists Harassing Victims.”

The Guardian, November 28, 2017. <

https://www.theguardian.com/us-news/2017/nov/28/us-guns-mass-

shootings-hoax-conspiracy-theories>.

Opportunistic Disinformation

Opportunistic disinformation occurs when predatory individuals attempt to capitalize on a particular

event or incident.

Opportunistic misinformation generally falls into one of two categories: revenue-

generating and financially incentivized, or malicious and politically incentivized.

Revenue-generating disinformation attempts to hijack the attention of social media users from a

particularly newsworthy happening, and redirect their attention for commercial purposes. A phishing

scam or spammer may mimic a pre-existing website and redirect the user to a sales pitch or other ad.

This technique is similar to malware that operates by hijacking a browser and redirecting traffic to an

alternate website. Scammers capitalize on a popular hashtag and use click-throughs to boost viewer

statistics on a website, or encourage the purchase of a specific product or service unrelated to the

original hashtag. An example of this is an article that circulated after a

2014 Sicilian earthquake to

supposedly provide news, however, the article was referencing a 1908 earthquake.

alicious disinformation is typically politically motivated, and can be even more challenging to both

identify and counter. Studies that have examined the volume, timing and location (e.g., tracked IP

addresses, associated time zone and geo-tagged posts) of this category of social media posts indicate an

intent to cause harm and disrupt the standard flow of truthful information during a specific event or

incident. During Hurricane Harvey in 2017, a rumor spread on Twitter that officials were asking shelter-

seekers about their immigration status. Also in 2017 after Hurricane Irma, a rumor surfaced that

survivors would receive generators from the federal government. While some posts could be attributed

to innocent mistakes, the scope and velocity of amplification seems to indicate an intentional

rebroadcasting of disinformation with the intent to frighten vulnerable members of the local

communities and weaken their trust with government entities offering essential aid.

Similar behavior

was demonstrated during response to a train derailment in DuPont, Washington, in December 2017

Emergency managers and social media specialists noticed an immediate surge of propaganda articles

assigning blame for the derailment to an anti-fascist group, despite no evidence supporting this claim.

Outdated Information

Today’s media environment relies heavily on being first with information. When crisis rumors start to

surface, novice and experienced users alike will scour the internet, often posting images of the initial

returns from their search without first verifying the date or accuracy of the data they are sharing. This

happens most often with users sharing photos from past disasters in a hurry as evidence of a disaster,

which is often believed as being true as the phrase “pictures or it didn’t happen” have permeated social

False information deliberately and often covertly spread (as by the planting of rumors) in order to influence

public opinion or obscure the truth. <https://www.merriam-webster.com/dictionary/disinformation>.

A scam by which an internet user is duped (as by a deceptive e-mail message) into revealing personal or

confidential information which the scammer can use illicitly. <

https://www.merriam-

webster.com/dictionary/phishing>.

Cory Nealon. “False Tweets During Harvey, Irma Under Scrutiny by UB Researchers.” University of Buffalo News

Center, September 28, 2017. <http://www.buffalo.edu/news/releases/2017/09/044.html

media users’ mindsets. Secondarily, older articles describing a past incident can resurface when

reposted, and publication dates are changed automatically through re-posting. Some examples include:

● Ol

d photos that are tagged and used for a new event. A user tweeted a photo from

2010

Nashville, Tennessee, flooding and incorrectly tagged it as Houston flooding in 2016. Using

www.tineye.com, the photo was verified as from Nashville. The user was notified, deleted the

tweet and replaced it with another photo.

● One of the most famous examples of an old picture (Figure 3) being shared erroneously is that

of the Bitterroot fire complex in Montana in July 2000, which was one of Time magazine’s

photos of the year. This photo titled “Elk Bath” shot by John McColgan has been shared in every

major wildfire since.

Figure 3: “Elk Bath” photo taken in 2000, which has been erroneously shared and attributed to other wildfires.

As recently as 2016, this fire was used to represent a fire burning in Tennessee. The rumor was

corrected by KRTV, correctly assigning the location and date the photo was captured.

Photo courtesy John McColgan, Alaska Fire Service from NASA’s Earth Observatory website:

https://earthobservatory.nasa.gov/IOTD/view.php?id=843

Case Studies

Incident Name: 2014 South Napa Earthquake

by Jennifer Lazo, City of Berkeley [CA] Emergency Services

Rumor Type: Phishing/Spammers Spreading Malicious Information

August 24, 2014, at 3:20 a.m., a 6.0-magnitude earthquake struck the area of American Canyon and

Napa, California. The earthquake shook awake many residents in the Bay Area and provoked a nearly

instant social media response, particularly on Twitter. As information about the earthquake became

available online, the hashtags #NapaQuake and #NapaEQ were broadly used by people in the affected

area and those responding to the earthquake. Popular hashtags often serve as a useful way to find and

sort relevant, actionable information during disasters, and the South Napa Earthquake was no

exception.

Problem:

Soon after the earthquake, the top earthquake related hashtags began trending locally and across the

world. As hashtags become popular on Twitter, “spammers” and “trolls” (i.e., people who sow discord

on the internet, including through social media platforms) target those hashtags in an effort to have a

broader audience for their unrelated message. In the case of #NapaQuake, a particularly disturbing

“hashtag hijacking” took place. For much of the first days of the earthquake response, a significant

portion of the tweets on the most popular response hashtags contained graphic pictures of dead bodies

from unrelated events. The main subject of the hijacking tweets were accusations of U.S. military

misconduct with images of people being tortured or horribly mangled bodies being included as

evidence. This was shocking content for social media monitors who were used to dealing with more

standard disaster response tweets, not inflammatory and graphic material.

It was unclear initially whether Twitter was taking any organizational approach to removing the

inappropriate tweets. However, some techniques used by social media monitors were effective at

blocking the worst of the content. One approach was to filter out geolocated and geographic

information that did not seem relevant to the incident. Most of the graphic tweets originated from

outside of the United States, so adding a simple geographic filter to tweets made it easier to filter out

the hijacking tweets. Over time, it seemed that some of the tweets were also being deleted on the back-

end by Twitter.

Newer algorithms on Twitter may hide these inappropriate tweets, but it is unclear if

those filters may also pick up and hide disturbing but relevant and useful content that social media

monitors need to function effectively, such as images of bodies left outside during Hurricane Katrina or

anger directed at a response organization for alleged dysfunction.

“Hijacked #NapaQuake Hashtag Turns Up Images Of Islamic Militant Slogans & Dead U.S. Soldiers.” August 25,

2014. <

http://sanfrancisco.cbslocal.com/2014/08/25/hijacked-napaquake-hashtag-turns-up-images-of-islamic-

militant-slogans-dead-u-s-soldiers-james-foley-isis-ferguson-world-cup/>.

Best Practices:

The tactic of location-based filtering of unrelated tweets remains a best practice for those attempting to

monitor social media during disasters, but in recent years, the spammers and trolls on social media sites

have used different methods and tactics to overtake hashtags and cause confusion. Agencies should be

aware that filtering by only location-based tweets can suppress local information originating from

devices that are not geolocation enabled. A Georgia Tech study conducted in 2012 indicated that less

than 1.4 percent of all content on Twitter is geolocated.

Another study conducted by Humanity Road

and Arizona State University on 2012 Hurricane Sandy data indicated that there is a potential significant

decline in geolocation data during weather events.

Incident Name: 2016 Louisiana Floods

by Amy Greber, American Red Cross via email on July 5, 2017, and LaVondra Dobbs, ViaLink Louisiana via

phone call on July 12, 2017

Rumor Type:

Incorrect Information

Ryan Gomba. “What Percentage of Tweets are Geotagged?” January 30, 2012. <https://www.quora.com/What-

percentage-of-tweets-are-geotagged-What-percentage-of-geotagged-tweets-are-ascribed-to-a-venue>.

Morstatter et al. “Finding Eyewitness Tweets During Crises.” Arizona State University.

<http://www.public.asu.edu/~fmorstat/paperpdfs/lang_loc.pdf

Problem:

ViaLink Louisiana, a 2-1-1 provider, found itself

overwhelmed with calls following the March 2016

floods in Louisiana. ViaLink noticed multiple

inaccurate Facebook messages and posts (Figure 4)

that went viral and contributed to the number of calls.

After FEMA declared a disaster, the calls kept coming.

In addition to the continuing Facebook messages and

posts, FEMA was also giving out the incorrect

information and referring people to 2-1-1 for claim

assistance (the information was later corrected).

In a similar situation, during the response to

Louisiana’s summer floods in 2016, the American Red

Cross was confronted with multiple rumors and

misinformation on social media related to its shelter

policies and food distribution. False claims spread

especially rapidly through new video tools, such as

Facebook Live, and threatened to erode the public’s

trust and support, as well as eclipse the personalized

care and outreach that the organization was providing

through social media.

To dispel rumors and also raise awareness of critical

resources, the Red Cross published a blog entitled,

“Top Questions About Louisiana Flood Relief” that was

shared 2,175 times. The Red Cross, its digital

volunteers and other online supporters used it to spread information among their personal networks, as

well as to confront critics.

The Red Cross Social Engagement team also created a secret Facebook

group where they could funnel important updates, flag urgent issues and collaborate among a larger

public affairs team; any further outcomes were then updated on the blog. At the local level, the Red

Cross communications team in Louisiana created regular informational videos from the organization’s

Louisiana warehousing site, where they provided situational updates for those seeking help and support

– and also as a way to combat misinformation. The first on-the-ground video update alone received

447,000 views.

Best Practices:

Actively publish frequent updates to help promote transparency and control the message.

Digital volunteers as applied to emergency management and disaster recovery is a group of trusted agents that

can lend support via the internet to those on-site who may otherwise be overwhelmed by the volume of social

media data generated during a disaster.

Figure 4: Inaccurate Facebook post claiming 2-1-1

would help with disaster claim forms.

Using trained digital volunteers can help shepherd affected people to critical resources and spread

reliable information online.

Incident Name: 2017 Oroville Dam Evacuation

by Mary Jo Flynn, Sacramento County Office of Emergency Services (OES) via email on January 3, 2018

Rumor Type: Insufficient Information

ue to rapidly eroding earth at the site of the emergency spillway that was compromising the integrity

of a small portion of the side of Oroville Dam, an evacuation order was sent out to residents in the

nearby communities and downstream of the dam to evacuate immediately. In the chaos of a rapid

mobilization and evacuation of thousands of people, a significant amount of misinformation spread

rapidly.

Problem:

The National Weather Service (NWS) Sacramento was distributing hydrology information impacting the

rivers, dams and levees within the northern portion of the Sacramento Valley. The area was receiving

significant rainfall accumulation from repeated Atmospheric Rivers causing widespread flooding. Prior to

the Oroville Dam evacuation, the NWS Sacramento station distributed a tweet that showed the area of

flooding from expected rainfall. This mapped image encompassed all of Sacramento County and areas

that included Oroville Dam. This image was distributed by @LNU_fire_photo (Figure 5A) just after 6p.m.

local time with information regarding the evacuation of Oroville Dam, leading those who saw the image

to believe that all of Sacramento County was under an evacuation order. Even though @JDLazo was

correcting the errant tweet (Figures 5B and 5C) at 6:30 p.m. local time, calls began flooding Sacramento

County 9-1-1 dispatch in the hours that followed. At 8:50 p.m., staff from Sacramento County EOC

completed a Facebook Live Video

and a Periscope video to correct misinformation. Almost immediately,

news media including television and iHeartRadio began quoting the live feeds in their broadcasts and

the videos were widely circulated on social media.

est Practices:

This rumor was believable due to the easily misconstrued flood map during the evacuation, and people

were generally starved for information. The best practice of having support from a digital volunteer

helped tremendously in minimizing the rumor; however, once it took hold, people’s genuine concern

caused increased stress and anxiety, and resulted in calls to 9-1-1. The use of live streamed video was

another best practice in this situation by getting the right information to the right people at the right

time. Because of previous live broadcasts, television and radio news media followed Sacramento County

OES’s social media feeds to share pertinent information. Pre-established partnerships with 3-1-1 and 2-

1-1, as members of the Joint Information Center, were receiving the same social media feeds and

information directly to provide to their callers. The radio stations broadcasted the audio portion of the

live video feeds, and television stations utilized their anchors to quote information from the Sacramento

County EOC. After the broadcasts, the County’s 9-1-1 dispatch center reported calls regarding Oroville

had stopped and their call volume had returned to normal.

Figure 5: Twitter conversation during Oroville Dam crisis discussing misinterpretation of Sacramento County flood

map.

Suggested Best Practices

Agencies should first determine their comfort level with the various tactics suggested in this section. It

will likely vary from agency to agency and will also depend on technology, training and policy issues.

Agencies should focus on the information sharing challenge or goal, not the tool (there are many

possible tools to use).

Here are some recommended goals for agencies to keep in mind when using any of the suggested best

practices below to counter misinformation, rumors and false information:

● Imp

rove the quality of communication through double-verification of information;

● R

emove ambiguity and uncertainty caused by misinformation, rumors and false information;

● Re

duce alert fatigue and the risk of “cry wolf” scenarios;

● S

eek ground truth as opposed to assumptions;

● B

e swift with releasing accurate information or acknowledging the situation to help the agency

work with a network of truth amplifiers and establish credibility early on; and

● D

etermine relevance of various social media information.

Best Practices – People

● Work with local television and radio news media (traditional media) to disseminate useful

information and correct bad information. Pre-establish partnerships or relationships with

stations or reporters that can be called upon in a disaster or emergency. While this paper is

focused on social media and increasingly more people are turning to this medium for their news,

recent Pew report estimates 50 percent of people

still get their news via television broadcasts,

compared to 43 percent who get their news

online.

● Use Virtual Operation Support Teams (VOSTs)

or other digital volunteers to monitor social media,

identify rumors and report back to officials so they

can work to correct it. Prepare workflows, practices

and activation procedures. The SMWGESDM’s

perationalizing Social Media report

offers

guidance in this area.

● Create partnerships through mutual aid

agreements with FEMA, Red Cross, or other

agencies and organizations that have the necessary

skills, personnel and systems to identify rumors and

misinformation. These partnerships could also be

collaborative efforts among local, state and federal

agencies, as well as jurisdiction-to-jurisdiction (e.g.,

local partners amplifying facts on behalf of other

municipalities if and when requested).

● Identify and leverage trusted crowd sources or

influencers, such as on-ground/on-scene users and

emerging influencers to propagate critical ‘good’

information perceived by the crowd and engage

them to disseminate rumor correction information.

r example, the National Voluntary Organizations Active in Disasters network is on-the-ground

and can be trusted to provide key information because the organizations work closely with the

government.

● Train and exercise first responders and digital volunteers to spot misinformation, rumors and

false information, and when and how to respond to bad information. Develop rules of

engagement or a concept of operations document for when or when not to respond. One way

Information provided by Elizabeth King, University of Washington Emergency Management via email on January

31, 2018.

Pew Research Center, “Americans’ Online News Use is Closing in on TV News Use.” September 7, 2017.

<http://www.pewresearch.org/fact-tank/2017/09/07/americans-online-news-use-vs-tv-news-use/

Hurricane Harvey and the Oregon VOST

By Elizabeth King, University of Washington

Emergency Management

Local Texas emergency operations agencies and

FEMA Region VI experienced a tremendous

amount of social media traffic in the wake of

Hurricane Harvey. A FEMA Digital Reservist

suggested that FEMA reach out and request

support from a VOST in three mission areas. The

first mission involved t

racking and delivery of

large donations,

and the second mission

addressed

tracking the recruitment of

volunteers and their locations. The Oregon

VOST provided a daily listening report to the

FEMA Region VI Mass Care Public Information

Officers. Among other things, they discovered a

olunteer group using a Google Sheet that was

publicly available to collect personally

identifiable information, which was shared in

the listening report. The third mission of the

Oregon VOST was tracking scams for donations,

including spoofing the American Red Cross and

webpages that redirected funds for victims to

other funds unrelated to Hurricane Harvey.

Multiple VOSTs worked together to provide

much-

needed support and not duplicate

efforts.

to research questions from the public is to search social media platforms with the event name

and a question mark (for example “flood?”).

Best Practices – Processes

● Use the Joint Information System (JIS) to coordinate public information efforts among

jurisdictions and agencies, and for standing up a Joint Information Center (in-person or virtual)

to facilitate the operation of the JIS.

● Prepare pre-scripted messages

and choose or create a hashtag specific to place, disaster,

agency, etc.

Be consistent and provide useful and actionable information.

● Prepare draft visuals, including

graphics, photos and videos, for

potential disaster scenarios. Archive

these in such a way that they are

rapidly accessible and can be quickly

modified to meet the needs of a

situation. As an example, a tornado

warning graphic (Figure 6) might

include simple instructions on what to

do.

● Actively tweet or post to correct

misinformation, rumors or false

information. Use hashtags such as

#mythbuster, #RUMOR or

#IncidentNameFact in posts and

redirect back to official sources. Use checkmark emojis or big red Xs on images. “Share back”

some of the misinformation (labeled as such) with the general public while the events are still

relevant to directly address inaccurate tweets or posts.

● Ensure older information is appropriately labeled and not re-circulated as new:

○ For critical information, continue updating a single Facebook post or existing news story,

rather than starting a new one.

○ Use the pinned post features on Facebook and Twitter so critical information remains at

the top of the news feed.

○ List known conspiracy sites and consider creating lists of bad actors in order to have

awareness of their posts.

● Verification tactics:

○ Reference this infographic

to do a quick fact check of the information.

○ Verify the legitimacy of information disseminated on social media platforms using two

or more sources, which could be websites (Factcheck.org or Snopes.com), social media

Document provided by Emergency Operations Coordinator Mary Jo Flynn, Sacramento OES.

Figure 6: Sample tornado safety graphic.

platforms, eyewitnesses or first responders (ground truth). Compare questionable

claims or posts to content being disseminated by official sources.

○ Acknowledge the rumor when appropriate. Statements such as “We are aware of the

rumor about [Topic].” Then use the opportunity to:

■ State the facts; and

■ Direct readers to sources of correct information.

○ Conduct reverse image searches using Google and TinEye. These may also be done via

this shortcut in Tweetdeck

○ Use a geofence and/or Twitter searches for locations to mine and help separate real and

false information (false information often comes from outside the disaster-affected

area).

● Consider a cross-entity social media campaign as a complement to more official resources

(incorporating something like a "verify2x" hashtag so that it could also be shared by partner

organizations).

Best Practices – Technology

● Set up a central website as a one-

stop shop/portal similar to the

Federal Emergency Management

Agency’s (FEMA) rumor control pages

(Figure 7 is a recent example from

Hurricane Harvey). Have a generic

page ready to go in case of disaster,

so an agency can turn it on and start

publishing. Embed a list of social

media accounts that provide

information in an emergency (i.e.,

local police, fire, city, 311, etc.). This

way, the most up-to-date

information is always available on that page.

○ If a rumor contains information that provokes a strong emotional response, it is

probably a troll. Agencies should be careful not to directly engage the troll, as they will

often go through great lengths to increase what they are doing. Instead, attack the

content of the rumor through acknowledgment or posting to a rumor webpage.

● Consider using Tweet hyperlinks

embedded within webpages and blog posts that encourage

sharing of factual information. A sample tweet might contain part or all of the message and

easily allow a user to share.

● Use live video (Periscope, Facebook Live or YouTube) to correct information with empathy and

accuracy. An example is this video from the Oroville Dam crisis

where rumors were addressed

and corrected.

Figure 7: FEMA’s Hurricane Harvey rumor control page.

● Monitoring software, such as Hootsuite or Tweetdeck, and more powerful tools, such as

Dataminr, Liferaft or Meltwater, can help agencies identify false information and watch for

influencers. This white paper does not endorse a particular product; however, such tools can be

a helpful addition, taking into account an agency’s circumstances (i.e., cost, human resources,

etc.)

Additional Considerations

With the best practices given above, agencies should be aware of the following additional

considerations when choosing to counter or correct misinformation, rumors and false information.

There are some risks to engagement, including:

● Accidentally amplifying the false message;

● Repeating the false message, even with a correction, can lead to more people believing it; and

● Correcting efforts can “backfire,” i.e., challenging a story may cause people who already believe

the false information to believe it even more.

Furthermore, the following factors should be considered:

● Geography: If the rumor is a local rumor, agencies should address it. If the rumor is not

spreading among the locals, agencies and responders may want to keep it on their radar for

their situational awareness but not address it.

● Volume: If the rumor is at a low volume, keep an eye on it. If the rumor spikes and becomes

high-volume, respond with a message highlighting the true facts that does not repeat the false

story.

● Subjects: For stories around collaborations and humanitarian response, highlight the positive

side of the story to indirectly counteract the effect of negative rumors.

Challenges

Below are some challenges first responder agencies may encounter when countering misinformation,

rumors and false information.

Legal Issues:

Agencies do not want to incur legal liability when correcting information and rumors. They

are recommended to consult their legal department or general counsel’s office in their jurisdiction to

create a framework to cover an agency’s action, which can be challenging, as the legal system is not

always keeping up with the pace of rapidly advancing technology. In addition, agencies should consider

the need for continued engagement with the public once an issue has been identified and how to best

handle that so they do not incur additional liability.

Lack of buy-in from executive staff and decision makers:

Agencies need to establish clear rules of

engagement for responding to rumors and false information. As part of establishing these rules of

engagement, agencies should engage with executive staff and decision makers (including the legal

office, see above) to ensure that everyone is on the same page regarding the rules of engagement.

Privacy:

The SMWGESDM’s previous report on

using social media for situational awareness and

decision-making discussed privacy issues around social media, so they will not be repeated here.

Funding:

Some of the best practices recommended in this paper require funding allocations to support

training, personnel and technology. If an agency chooses to implement a particular social media

monitoring technology, for example, they will require funding to support acquiring and using that

technology.

Conclusion

As rumors, misinformation and false information will continue to circulate, they cannot be entirely

eliminated. Agencies can leverage the above proactive and preemptive measures to lessen the risks

during disasters and emergencies as a result of misinformation, rumors and false information. Some of

the measures detailed in this report include mutual aid and partnerships with credentialed digital

volunteers, pre-scripting messages, verification tactics, setting up a centralized webpage and more.

Agencies should consider testing and exercising with rumors, misinformation and false information to

help them determine which best practices will work best for their audience. The SMWGESDM’s previous

report on incorporating social media into exercises

offers how-to guidance.

cial media is a continually changing topic, and while the tactics discussed in this paper are relevant

now, the landscape continues to evolve. In the future, the authors of this paper may add to this paper or

create an external living document of references and resources that may be relevant for first responder

agencies.