October 29th, 2013 by Rob

It’s safe to say that I know more about social media in disasters than pretty much anyone. I’ve helped run disaster response efforts utilizing social media in English, French, Haitian Kreyol, Urdu, Pashto, Spanish, Portuguese, German, Hindi, Sindhi, Arabic, Sudanese, Indonesian, Japanese, Korean, Italian, and more than a few other languages, across six continents and more countries than I can count. My PhD from Stanford was the first anywhere that looked at applying cutting edge Natural Language Processing to short social media communications in crisis response contexts globally. My company, Idibon, is currently working in social media for disaster analytics in a number of locations.

So I say the following with conviction and expertise:

Do not use Twitter for reporting risks during a disaster.

It does more harm than good to share open information about at-risk populations. In every context that I have seen, the misinformation is more viral than the truth; people using social media rarely self-correct misinformation with the same virality; the information is often condensed to the point of being misleading through omission; and most of all, the world’s most at-risk people are being put in danger.

In my professional work, I have worked closely with crises-affected populations and as a result I have long been opposed to sharing information online about at-risks populations, and a vocal opponent to media organizations that do. But I am also hesitant to criticize the people who *do* share information when they are directly connected with the disaster. It’s not reasonable to expect people to act rationally.

Last week, I provided an example myself. So I’ll use my own error to point out the argument.

When it’s close to home


Last week, fires swept through my childhood home of the Blue Mountains in Australia. Family friends lost their homes, and from San Francisco I saw a stream of photos of familiar streets with an apocalypse of smoke hanging over them. It is the small things that make it so familiar: the cars on the left, the shape of the power poles, the style of houses and lawns. On Tuesday, I heard from my mother that she was evacuating her home to stay with family friends outside of the region.

It felt terrible to view the photos and videos as many people I know were leaving the town I grew up in as an inferno approached. It felt just as bad to be so far away at the time. I received the news of the evacuation on a day that was more than typically busy: I had meetings starting at breakfast and going past dinner; I was moderating a panel at the annual crowdsourcing conference; I was trying to remotely manage staff in-between. For the whole day, the safety of many of the most important people to me was in my mind, and I tweeted without thinking:

@WWRob: my thoughts are with my family and friends evacuating the Blue Mountains today.

Despite my expertise, experience, and my vocal opposition to publishing risks about at-risk populations, when it was my hometown that was threatened, I made the mistakes that I warned against. I quickly deleted the tweet and the Facebook post that it had automated generated. Here’s why:

1. I was (potentially) contributing to misinformation. Nothing I said was untrue, but without the full context my tweet could have been interpreted as a full evacuation because the fire was expected to go through all inhabited areas, which wasn’t the case (see below). This is just the kind of ambiguity that is hardest for me to evaluate when running disaster response services, and I am annoyed that I repeated it.

2. I was misleading through omission. The risk to my parents’ home was small. The fire was only a few kilometers away for several days, but the prevailing wind and availability of fuel reduced their specific risk. People had been advised to evacuate by a certain time or not leave, primarily so that the roads remained free for emergency crews during the expected worst period (there are many places with only one road in an out, often narrow and winding). But it was also positive that many people stayed: many places were only at risk of spot-fires from embers, and it can help disaster response efforts when people are able to identify small spot-fires as quickly as possible. The subtleties of why people were being evacuated couldn’t be expressed in 140 characters.

3. It put my family’s home in danger. It wouldn’t be that difficult to figure out who my relatives in the area are and find their address. And I just announced to the world that their houses were empty, in a neighborhood where the emergency responders were sure to be busy elsewhere. There was no widespread looting, fortunately, but there were thefts of donations for fire victims already reported – I should have known better.

Cheap journalism and cheaper lives

Was I wrong to empathize, worry, and express my concern? No, or at least, if I was wrong, everyone else who reports their concerns on Twitter is less wrong, as they do not work with in the area. So I don’t blame individuals when they do express their fears under more duress and with less knowledge of the unintended outcomes.

Are traditional media who rely on social media for stories about disasters wrong? Maybe. In all my experience, I have never seen an example of social media being useful in disasters for collecting information from the population (‘situational awareness’). It is worth pursuing as a research topic, but not yet for response. From hundreds of millions of social media interactions that have been collected and analyzed during disasters, I have only seen one example of information going to responders via social media that definitively helped someone. Against that one positive case, I have seen dozens of confirmed examples of people being killed or going missing as a result. In the 2009 Mumbai hotel attacks, the terrorists used social media to identify and kill American and Jewish hostages. Mexican cartels have repeatedly used social media to identify and execute their perceived enemies. The majority of people tweeting in Tehran during the ‘Twitter Revolution’ are now unaccounted for, not because Twitter actually played an important role: the media declaring that Twitter was important put the spotlight on people who had no choice but to suffer the consequences.

Against so much death and hundreds of millions of tweets analyzed with little or no positive impact, the media then turns around and produces articles like this:

Twitter served as a lifeline of information during Hurricane Sandy

This article, published yesterday by Bruce Drake of Pew Research, is a dangerous lie. The article is a report on a different article that was also by the Pew Research center, which simply categorized how Twitter was used. Drake side-steps the finding that the most shared information was incorrect and adds this ridiculous title about a ‘lifeline’ which he seems to have invented for publicity. And it work – this lie was happily retweeted by people like Alex Howard who should know better. As is already well-known, NYC sent less than one tweet per day per social media staff member. When my company was helping FEMA run information processing following Sandy, I don’t remember much talk of social media at all.

These false reports about the importance of social media are encouraging more people to share information about at-risk populations online and will ultimately contribute to more difficult disaster response efforts with fewer people helped.

Loose lips


To separate out the functions of social media, there’s a strong argument to be made that emergency response organizations should use social media as a broadcast medium, as people consume news via social media, and use it to point people to the correct reporting channels. I first argued this in 2011: disaster response organizations should use social media to encourage the centralization of information. The recent Twitter Alerts project follows this advice. Twitter have (thankfully) ignored requests by media agencies to remove disaster-related hashtags from their API limits, and instead allowed professional response agencies to use their services to broadcast information.

But disaster response professionals and the media alike should stand against reporting sensitive information about disasters on social media. There’s nothing new in campaigning for safer information sharing, but where are calls for this? It is a natural compulsion to speak out about what worries us, but it is not natural to think of the implications of spreading information to such a wide audience. When I tweeted about the evacuation of my home town, it was in solidarity with the people that I knew there, but I didn’t think beyond these people when caught up in the moment. I share my mistake as a warning to others: we need to be pro-active in curbing the publication of information about at-risk individuals.

Robert Munro
October 29, 2013

12 Responses to “Why you should not use Twitter in a disaster”

  1. Mark Says:

    I understand your arguments but don’t you think that it is futile to ask people to stop using social media when social media is moving everyone away from private data?

  2. Rob Says:

    Hi Mark

    Social media is moving *towards* more private data, not away from it:
    “[W]hen we use digital technologies to communicate, most of us are privately interacting via sms, email and instant messaging, and more likely to communicate in our first languages as a result.”

    It’s our natural linguistic tendency to think about addressing a small number of interlocutors, which is why the global human population is moving away from publicly broadcasted social media communications. But even if we were communicating more openly in our digital communications, it doesn’t mean that we should be doing this for all aspects of life.

  3. Brendan O'Connor Says:

    I was going to tweet this, but in the spirit of propagating verified information… I think I’ve seen reports about the Mexican cartels, but was less familiar with the rest. Do you happen to have references for these cases?

    In the 2009 Mumbai hotel attacks, the terrorists used social media to identify and kill American and Jewish hostages. Mexican cartels have repeatedly used social media to identify and execute their perceived enemies. The majority of people tweeting in Tehran during the ‘Twitter Revolution’ are now unaccounted for, not because Twitter actually played an important role: the media declaring that Twitter was important put the spotlight on people who had no choice but to suffer the consequences.

  4. Rob Says:

    Hi Brendan, yes, of course, here are some references:

    “60 Dark Hours at Hotel Taj” in 26/11 Mumbai Attacked
    It looks like the terrorists also used social media for situational awareness. They didn’t hack any accounts: it was just one guy on a phone to someone outside the Hotel Taj who was looking at social media. It could have been thwarted by reporting the same information through non-public channels with the weakest security. This was why the government called for people to stop mindlessly tweeting about what they saw.

    See the extended discussion in “A Revolution in Search of Revolutionaries” in The Net Delusion: The Dark Side of Internet Freedom
    It was Al-Jazeera’s director of new media who found there were only about 60 active Twitter accounts in Tehran, which fell to just 6 after the ‘revolution’. The book also covers how the misinformation was spread, leading to the crackdown, which is chilling.

  5. Mark Says:

    Interesting! But if people are moving to private data anyway, then why bother trying to convince people like Alex Howard to give up his role as the spokesperson for open data?

  6. Ih8Sandy Says:

    Robert, I just wanted to say Thank You for being the voice of reason. The Pew story made my blood boil yesterday. How shameless to exploit the anniversary of when we lost our friends and homes to talk about the triumph of technology.

  7. Rob Says:

    I am sorry for your losses and know that the rebuilding keeps going. People who actually work in disaster response know that so few people are helped no matter what technology is used. Be the bigger person and ignore the others.

  8. Rob Says:

    Why? Well, for one, it’s worth the effort. Alex Howard is not the blind spokesperson for open data that he is made out to be in some circles (including by the author of the the ‘Net Delusion’ above). He’s a smart guy who takes it all in – I see him spend all day at conferences, some disaster-related, often helping out through by moderating panels, etc. So I don’t question his motivations.

    Even though we are moving to more private data, we need more discourse and analysis about *when* open data has better outcomes, which is what’s lacking. I can’t do this myself – I wrote this post last night because my evening plans were canceled last minute, and I’m responding to comments over lunch. My path is to create a company that builds the technology to help with response. Plus I only need Alex Howard to report on what he already knows:

    When he published “Data for Public Good” early last year, he allowed my comments correcting the bias that portrayed private crisis response efforts as open (and if I remember right was happy to quote my criticisms via Twitter):
    When I spoke at last year’s strata about how data was becoming more private, he was among the first to tweet these facts.
    When he wrongly portrayed our 99% closed efforts responding to Sandy as open, he allowed the correction there, too:

    Knowing this, it’s irresponsible to retweet a shoddy piece of journalism with an ‘opendata’ tag, to an audience of the size that a professional could never reach. It’s the kind of misinformation spread that I’m trying hard to stop.

  9. Marie Says:

    Thanks for having the courage to admit you made a mistake during a disaster. I hope that your family are safe.

  10. Jake Says:

    Here’s another example for you:
    Somalis were killed and harassed for reporting to Al Jazeera about the war. Sounds to me like Al Jazeera learned the wrong lesson about social media in Tehran; they went after the publicity.

  11. Rob Says:

    Thanks Marie. Yes, my family is fine, and thankfully there were no deaths or even serious injuries when the homes were lost in the Blue Mountains.

    Jake, I remember that, and the Fast Company/Co.Exist Reporter is correct: Al Jazeera and their partners did expose the identities of people in Somalia and try to quietly cover it up. They were text messages in that case, not tweets (which means they should have been *more* private as the Somali citizens had no idea). In light of the recent targeting of ‘soft’ human targets in/near Somalia, it is sickening to read the comments from 2011 that were trying to downplay the risk. This was deplorable and did not go unnoticed in the tech world: it is one of the reasons why the “Volunteer Technical Community” and related media organizations are no longer welcome at mainstream crowdsourcing events. I realize this punishment does not really measure up to the crime.

  12. Let’s leave disaster anniversaries to the victims | jungle light speed Says:

    […] For disaster response professionals, it is natural to focus and reflect on the most important part of their jobs: actually responding to disasters, so it is easy to have the wrong focus and forget that many people will not share the enthusiasm for any part of the response effort, no matter how successful some component might have been. I was reminded of this recently with a comment that was left on a post about social media and disasters: […]

Leave a Reply

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>