Research on Social Media Use in Disasters

by Eric Holdeman / September 23, 2014

I attended the Washington Emergency Public Information Network event in Spokane on Monday. There was some great content that I'll share in the next two blog posts.

First off there was the following information from Kate Starbird, assistant professor at the University of Washington, who has been doing research on the social aspects of social media. My notes from her presentation are below: 

Post-Sandy, 20 million Sandy-related tweets

10 photos per second

All disasters are social because they impact people.

Every disaster now has social media interactions. Social computing & Crisis Events

Citizen reporting

New ability to communicate with people and “they are listening.”

Followers balloon during a disaster. They used to go to your website, now they are going to their social media sites.

Volunteerism is also being enhanced. The crowd is participating in new ways.

Challenges of social media data:

·       Volume of traffic is very high

·       Background Noise — only a tiny bit of information is useful

·       Lost Content

o   Information loses connection to the person who sent it

o   The “truth changes” as the facts and circumstances change

o   Misinformation & Disinformation

o   Unstructured Data

How can we make social media data useful?

The online crowd is helping solve this problem. There are spontaneous volunteers both physical and digital. For example, people tried to help after Sandy #njgas sprang up to help people find fuel. Then a digital volunteer set up an aggregation tool. Franklin High School set up a map to display where gas was available.

Recurring behaviors that are being seen:

·       Voluntweeters after the 2010 Haiti Earthquake

·       Identifying and amplifying actionable info

·       Routing information to response organizations

·       Verifying information

·       Acting as remote operators

·       Emergent Organization

o   Only about 1 percent were connected before the event

o   Ushahidi as a great example

Digital volunteers bring about emergent collaborations 

Catskills Liveblog started with Hurricane Irene. They had a small online presence.

Then they created a live blog:

·       Moderated comments from guests

·       Tweets from selected accounts

·       They created a list of “trusted sources” whose tweets

·       Successful emergent collaboration comes from local expertise journalistic expertise and crowd capacity.

·       They recruited moderators from the crowd, training moderators and transferring journalistic expertise


Crowd capacity

·       Moderating

·       Dealing with misinformation:


When a rumor shows up:

They used “visible skepticism.” They didn’t disallow things like a dam failure. 

You can’t ignore information. 

Bridging gaps in infrastructure and access — Crowd-powered mesh network

 ·      People without Internet connections

·       Media read the feed

·       Volunteers will map the information

·       Kids and others spread the information.

·       Neighbors share information


·       Misinformation is often a product of sensemaking (rumoring)

·       Always been part of the process

·       Sensemaking not something specific to social media

·       Rumoring is going to happen whether you/re there or not


Misinformation can be a problem because the crowd self corrects. In the Boston bombing there was a tweet about a young girl running in the race who was killed. This was bogus information.

The crowd does correct, but not as long or as efficient as the rumor. 

Boston, collected 10 million tweets from Twitter using the Twitter Streaming API.


One type of misinformation is speculation.

How do we make social media data useful?  Listening

How can you use social media effectively?  Engaging

Platforms & Programs