Tag Archives: Kate Starbird

SMEM chat: Monitoring Social Networks–how do we listen?

Post by: Kim Stephens

This week the SMEM chat topic was: “How we use social media in an emergency and how do we listen?”  This is an important topic–as someone pointed out, one of the first emergency responders to adopt social media, Brian Humphrey of LAFD, once stated that  70% of Social Media is active listening. Click here for transcript on “What the Hashtag”.

Q1: Do different emergencies require different kind of monitoring?

Yes and no, the tools are probably the same, especially since there aren’t that many monitoring tools available. But different types of events probably require different tactics–for example, a natural disaster vs. a man-made or terrorist-type event; or a fluid vs static crisis. Chris Hall who monitors events daily, explains that  during a fluid event such as a wildfire, keeping up with situational awareness via mapping is very important. In contrast, after the Japan quake, the tactics included identification of victims in the affected area, rescue needs and first aid needs. This was followed by identification of needs of the survivors, such as shelter, food, water, etc.

Kate Starbird of “Tweak the Tweet” who also monitors SM daily, suggested that differences that do matter seem to be notice vs. no-notice crisis events, as well as the number of people affected, geographic location, culture and language.

Of course, in a terrorist type event, Chris points out that you will need to be listening to see if anyone is trying to intentionally propagate misinformation. Additionally, information coming from response organizations probably will be much more guarded.

Follow on Q: Does your strategy change given the scale of an emergency if so, how?

Large events require more of everything, including the need to listen more. This might require more listeners, which of course led to the question of who will be available to help with that task? See thorough discussion of this point below. Also, Chris pointed out that during a large-scale event people from all over the world will be listening.

Q3: Do different channels get different info?

This is an interesting question which really points to why you, as a response organization, can’t just be wedded to one type of social media platform. The CDC, for example, uses 17 different social media tools–and I’m probably understating the number. In Japan, facebook and twitter aren’t the most popular social media platforms.

This discussion, however, quickly went into a facebook vs. twitter convo. Wendi Pickford suggested that you can explain information more in-depth on FB than on twitter, and therefore you can squash rumors a little easier there. Others, including Wendy Harman of the American Red Cross, seemed to think FB was more important for relationship building.

Twitter Monitoring:

On twitter, the tools for listening are fairly straightforward, including following the hashtags people are using for the crisis; following key actors such as community leaders, local media, other response organizations; and by using matrix tools (such as tweetdeck) to follow multiple streams of info. Other tools, such as google realtime search don’t even require that you have a twitter account to follow what’s happening. These are all mostly free tools, but there are some vendors that are now selling applications that incorporate SM monitoring and data into their overall situational awareness platforms. But if cost is a concern, organizations can start monitoring with the free tools first.

Facebook Monitoring:

Facebook is much different mainly because it is often presumed that you cannot monitor people’s pages unless you are personal “friends”– even if they are one of your fans.  Kate Starbird mentioned how FB is difficult to monitor due to stricter privacy policies, as well as the fact that there’s no real ability to aggregate data from FB sites “Facebook doesn’t allow collection/monitoring, except in-house.”

But @EmergencyTraffic pointed to some tools you can use to monitor facebook–linked above.  As I’ve noted before, if your response organization attracts people to your page as the “go to” source for information, then people will post situational awareness information as comments–especially if you asked specific questions.

But, I have found that some FB pages are not necessarily even monitored very well on a daily basis. This example on the right is from a state emergency management organization’s page. They have allowed a young woman to post questionable content to their wall. This has been up for seven days and is still one of the first posts you see when going to their site. Some would use this example as an excuse why they shouldn’t engage at all, so I’d like to make three quick points:

  1. Your policy should state that people canNOT advertise on your page.
  2. Monitor often enough so that you can remove  irrelevant postings.
  3. Don’t allow people to post to the wall, just in the comment section.

Q4: Resources—staffing and volunteers–how do we get the people to make this work? Many EOCs don’t have enough people to do their planned tasks, so who listens to the SM channels?

Ideas:

  1. 911 operators? I’ve heard some organization hint that maybe 911 operators would be the right resource for monitoring SM platforms. Most people on the chat, however, thought that was not the way to go since they are under-resourced to begin with and the skills necessary for monitoring and analyzing the data are not part of their normal functions. So would the answer lie in virtual volunteers instead? (See this article tweet 911, tweet 911 by @chiefb2, for a thorough discussion of the challenges associated with of this approach.)
  2. City County Employees: Chris suggested starting with city/county employees, who are already trusted–e.g. public works employees. Heather Blanchard calls this concept “sourcing your own crowd”. My concern, would be that their contracts would precluded them from this type of additional duty, particularly when incorporating the necessary training. It might work, however, if they volunteered and understood that they wouldn’t necessarily be compensated for the time.  But I can see the can of worms this might open.
  3. Local citizens who use social media. Cheryl Bledsoe suggested that EMs should be collaborating before a crisis with local heavy social media users. Jim Garrow indicated that Ozarks Red Cross and @MRCPhilly are planning to use volunteers for the monitoring function. I love what Kate Starbird said, however: “Real solution lies in combination [of] human computation, plus tools (crowd).”

4. CERT: This comment from Administrator Fugate led to a robust conversation about the role of CERT for social media monitoring. Some suggested it was not only a great idea, but was already happening (e.g. http://twitter.com/ecert).  Some suggested this concept could be broadened to include CERT members reporting observations through SM platforms such as preliminary damage assessments. But in order to make CERT SM monitoring a reality for most locations, standard training protocols would probably need to be established.  This new role would also have to be integrated into plans and exercises.

Cheryl Bledsoe, EM from Washington, stated that they don’t use CERT for SM monitoring “…because CERT, by theory, is self-deploying and not tasked out directly by the EOC.” She also noted that being able to use CERT or not would directly relate to their proficiency in the medium. To be honest with ourselves, most CERT members are not people who enjoy using these platforms in their daily lives. However, would this new function attract a different kind of volunteer? Maybe someone who might find this type of work more interesting than the normal CERT roles. Or, as Kate Starbird asked, could there be a special class of CERT just for social media monitoring?  But Cheryl asked, “Is this role, already being filled by organizations such as CrisisCommons?”

5. HAM radio operators: Others suggested using HAM radio operators for SM monitoring, and this is a discussion we have about every other week. Some people think it’s a great idea, others, not so much. It probably depends on the local HAMs these folks know personally.

6. Pre-trained EOC volunteers: Marcus Deyerin went in an entirely different direction, he stated that some OEMs use pre-trainined EOC volunteer support teams. So it might be “[e]asy to add SM monitoring positions to these groups.”

Alicia asked as a finally question: Why? Chris Hall summed it up: it’s expected, it’s important to the mission, and it improves situational awareness.

Follow up discussions recommended:

  • @densaer stated: “I think we need to reevaluate the role of PIOs re EM. More like intel functions.”
  • Via Patrice Cloutier: NIMS and other docs will have to be reviewed re; SM and roles in JIC for example.
  • eCERT training to monitor SM platforms (a toolkit: policy, best practices, all in one location).
  • Strategies for overcoming liability concerns with using volunteers to monitor SM for your response organization. (An already suggested strategy: “[having] a good plan and meaningful training.” via Kris Hoffman
  • Communicating through social media channels during the recovery phase.
  • Which elements of the response “own” the inbound and outbound messaging? via @dshawnfenn

A big thanks goes to the host Alicia Johnson, Executive Director of the Salt Lake City Urban Area, as well as everyone who took time out of their busy days to engage. As always, I participated, but most of the thoughts expressed in this summary are not my own.

Links and resources for monitoring mentioned during the chat:

Examples:

Using Social Media to Gain Situational Awareness — It’s Time To Question Assumptions

Post by: Kim Stephens

Many have assumed that after a large-scale disaster event all communications would be silenced in the impacted areas.  Recent experience, however, has proved this assumption incorrect — first with Haiti (Jan 2010) and then in recent months in  Australia. In both cases, the cell towers proved to be more resilient than assumed.  Because social media platforms can be accessed on hand-held communication devices, survivors and public safety organizations have turned to these platforms as a way to keep the information flowing during and after a disaster.

Citizens in impacted areas don’t just receive information, but increasingly, they send out bits of data about what they are seeing, hearing, and feeling through these platforms. These data, if aggregated, can contribute to overall situational awareness.  We are really beginning to understand Brian Humphrey of LAFD’s phrase “every citizen is a sensor”, a take on the phrase every soldier is a sensor. But what now? Citizens, obviously, do not pass the information up through chains of command, nor do usually pass that information in any structured way. How do we filter, verify, aggregate and make sense of ALL THAT DATA?  As crisis mapper and PhD candidate Kate Starbird said in a recent interview: “Should it be the only source of information? Absolutely not. But if it’s there, why not use the information?” Also see this video of Craig Fugate, Director FEMA, talk about how important this is.

This is a really big topic so I have just tackled a small part of the issue in this post. I also have tried to include as many links as possible to articles that explore the topic in much greater depth. The point here is to try to aggregate some of these issues and questions for use in our discussion on the SMEMchat hashtag, which is scheduled to take place Friday, Feb. 3 at 1230. If you have other questions you’d like brought up, please post them to twitter on #SMEM or at the bottom of this post.

QUESTION #1: How do we gather  information from social media platforms?

(a.) One way to gather data: ask for it. The US Army Handbook on Social Media suggests that during an emergency “Organizations should encourage people on the scene to send information.” They go on to state that “No matter how information is submitted, the command site should promote this content when appropriate.”

(b.) Be a magnet. What I think we are seeing in Australia, is that the Queensland Police Service social media presence has created an avenue for people to provide information that can be more easily monitored by response personnel. They have done this both with their twitter account, by establishing and using hashtags that were widely adopted during the flood and the current cyclone, and by creating a robust facebook page. Just by reading through the comments on the QPS site, you get a sense of how people can provide situational awareness information directly to you. One person states: “Just gotten in contact with family in Kewarra and they have power, not to sure of damage but it wasn’t as bad as we first thought…”

One concern I’ve heard voiced from response organizations is privacy. However, if people are volunteering their information to your open and public site, they most likely understand it is not a private conversation. Another concern I’ve heard came from the QPS media team themselves, there is a LOT of information to sort through. This brings up the question:

Can/should emergency operation centers use volunteers to help sort through the data pouring in through their own social media sites?

(c.) Have trusted sources: Other emergency managers, Cheryl Bledsoe in particular, have noted the importance of  having a presence before an event, which helps create real trust with people online. During an event you can turn to these “trusted agents” as sources of information. (Hey @greatguy What are you seeing around the lake?).

Again, Jeff Phillips, aka @LosRanchosEM, provides a great example of this. Here in this screen shot of his twitter feed, you can see that he is retweeting information supplied by others. When asked about his practices in RTing Jeff states: “I do my best to verify “trust” before RT – not the same as saying only “official” sources. Sometimes I RT with a question mark.” I asked him if he includes that information in his official situational report, and he indicated that he does include verified information in his county’s sitrep.

QUESTION #2:  Can We Ask for the data, but in a structured format? Even in Australia, however, we have seen that being a magnet for information is really not enough. There is just too much information for response organizations to make sense of it all in a timely manner. Some of the posts on the QPS facebook page received over 1000 comments. There were thousands of tweets during and after the cyclone with the tag #TCYasi. Trying to sort through and make sense of all of that potential data is a real problem. (I say potential because a lot of comments are merely “thanks for the good work!”.)

What smart phone applications and other formats have been developed to help citizens report data in structured format?

(a.) For smaller-scale events, an example of an application that would make it easier for the public to send information in a more structured format is the application “See Click Fix” which is promoted for use in identifying non-emergency issues in neighborhoods. (Thanks to @UrbanAreaAlicia for pointing this out). As stated on their webpage, this application “allows anyone to report and track non-emergency issues anywhere in the world via the internet.” As the “click” implies, people are encouraged to send in photos of the problems.  If you are reminded of Ushahidi, I’ll get to that in a moment. But applications like this one might be worth exploring for use after a disaster, particularly for local government with limited resources.

(b.) The public can also be educated about how to structure information shared through social media platforms so that it can be integrated with other data feeds and placed on visualization platforms. One example of this is the National Weather Service’s new experiment called “Twitter Storm Reports” In their flyer they state: “You can now submit your significant weather observations to the NWS via twitter.” The two page flyer gives very specific information on how-to structure tweets, including a full description of how they should be written with or without geo-tagging.  One of the example tweets demonstrates the importance of including the person’s location if they do not have geo-tagging. “#wxreport WW 378 W. 156th Rd. Anthony, KS WW Wind Gusts estimated at 60 mph”.

(c.) This reminded me of the “Tweak the tweet”, an ongoing effort on the part of aforementioned Kate Starbird, a PhD student at the University of Colorado, to educate the public about how to better format tweets in order to”leverage twitter as a semi-formal communications channel”. The campaign also informs users on how to format tweets so that computers can aid in processing the information.

“This processing includes extracting location information, creating incident reports from tweets, and sorting these reports into different types of categories. The processed tweets can then be displayed on public web-pages in a variety of formats that allows users to see where different types of information has been reported.”

This is a very impressive endeavor that is starting to yield some great results. See this 3 page description here.

Correction: In my attempt to be brief it seems I left out some important information. I received an email from Jeannie A. Stamberger, Ph.D., Adjunct Faculty,Carnegie Mellon Silicon Valley, that sheds some additional light on how and why the tweak-the-tweet was created. Dr. Stamberger states:

I wanted to let you know that I co-created Tweak the Tweet with Kate Starbird at the Random Hacks of Kindness in November 2009; we have published on creating the idea together. At Carnegie Mellon Silicon Valley Disaster Mangement Initiative we continue to work on Twitter use in disasters exploring further questions related to gathering accurate credible information from the crowd including just-in-time credibility building, use of social media in disaster drills which teach the public how to use information resources during a disaster, and we will be testing methods in May using Tweak the Tweet in amateur radio; amateur radio is the backbone of communication in disasters, yet the information is missing from the digital feeds currently being processed by the crowd to aid disaster management.

We are also working with local authorities to develop optimal “canned” alert messages to familiarize them with the in’s and out’s of how to get your message across. Others at CMU are working on identifying location of Tweets from colloquial language in content (see January publications) and comprehensive analysis of characteristics associated with re-Tweeting likelihood.

(d.) And Ushahidi (if you’ve never heard of Ushahidi watch this video) has an iPhone app as well. From the iTunes preview page: “Ushahidi is an open source platform for democratizing information, increasing transparency and lowering the barriers for individuals to share their stories.  The iPhone and iPad app synchronizes with any Ushahidi deployment allowing viewing and creation of incident reports on the go.”

The app supports loading of multiple deployments at one time, quick filtering through incident reports, exploring incident locations on the map, viewing incident photos, news article, media as well as sharing incident reports via email, sms or Twitter. Once the data has been downloaded, the app can function without an internet connection, allowing accurate collection of data utilizing the device’s camera and gps capabilities.

QUESTION #3: Can We Combine social media with geo-spatial mapping?

The description of Ushahidi’s app dovetails perfectly with the question of integrating social media with geo-spatial mapping. Again this example comes from the resent back-to-back crises in Australia. Although geo-spatial mapping with crowdsourced data on the Ushahidi platform  became very well known after its well-publicized use in Haiti, I think it showed even more promise in the application’s deployment in Australia when it was combined with the power of the GIS mapping giant, ESRI.  The application allowed for “trend analysis” and, based on reports from the field, was used by responders “to create releveance and context from social media reporting.”  See this article by Alex Howard, of O’Reilly Radar. Alex continues:

The Australian flooding web app includes the ability to toggle layers from OpenStreetMap, satellite imagery, topography, and filter by time or report type. By adding structured social data, the web app provides geospatial information system (GIS) operators with valuable situational awareness that goes beyond standard reporting, including the locations of property damage, roads affected, hazards, evacuations and power outages.

Here is a screen shot of ESRI’s application during Cyclone Yasi.

QUESTION #4How do we create feedback loops so that responders know when information coming from social media platforms has been acted upon? See the article listed below “From Haiti to-Helmand” for a detailed discussion of this point. Lin Well’s states that feedback is essential to not only know what has been acted upon, but to identify what has not.

There is a lot to discuss. I’m anxious to participate in the online chat with the emergency management community on this topic. I will report back with everyone’s thoughts.

Some Great Sources:

Your comments and suggestions are invited.