Pilot Project Offers Insight into Using Social Media Data for Emergency Response

In a recent post, MCP Insights chatted with Dr. Andrea Tapia, associate professor of information sciences and technology at Pennsylvania State University (PSU) in State College, about the impact social media is beginning to have on the 911 community. This post explores a pilot project that concluded in August 2018 at the Charleston County (S.C.) Consolidated 911 Center that explored the use of social media data in emergency management and response. MCP, RapidSOS and RapidDeploy also participated in the pilot project.

Collaborators from PSU’s College of Information Sciences’ 3C Informatics: Crisis, Community and Civic Informatics, led by Dr. Tapia—who is working with MCP for the next year as she takes a sabbatical from her duties at Penn State—explored how access to social media data could impact 911 operations, specifically by improving situational awareness during emergencies.

You can hear from all pilot program participants, including the Director of the Charleston County Consolidated 911 Center, during MCP's panel discussion on social media and 911 on Thursday, December 13, 2018 at 12:00 PM ET. Register here.

Insights: What sparked the idea for the pilot project?

Tapia: Last year I was invited by MCP to participate in an early adopter summit presented by one of its clients, the North Central Texas Council of Governments. I met many interesting people there, but none more interesting than Christy Williams, NCTCOG’s 911 director, and Jim Lake, director of Charleston County’s 911 center. They are the most open-minded, forward-looking people in the 911 community. They both understand that 911 service must evolve. The key to that evolution is finding people who are willing to take risks and Jim is one of those people. He invited a bunch of weirdo researchers into his center to interview and observe so that we might create something that would work in that setting.

Insights: What did the researchers set out to do?

Tapia: The idea was to create software that would gather social media data in a particular way, and then clean it, process it, organize it and display it. The challenge is to produce clear, distinct results that are actionable, especially to the first responder. Social media data is messy and full of noise—even with the best algorithms, it’s difficult to pull something useful out of the mess.

Insights: So, what can be done about that?

Tapia: Someone needs to figure out exactly what is valuable in that data, when it is valuable, what form it needs to take to be valuable and—here’s the big one—how it fits into the workflow of existing 911 centers. It’s not sexy—it’s a lot of hard work.

Insights: Will that be different at every PSAP because they all are dealing with unique operating environments and have wildly disparate resources?

Tapia: There are going to be things that every PSAP is going to want, but yes, this solution needs to be customizable.

Insights: What were some the key takeaways from the pilot project?

Tapia: There were several. One was that 911 centers do not want another new system. What they want instead is to integrate social media data into their existing systems. They also want the data to be filtered by a human being before it is integrated into their workflow. Finally, they want the data filtered based on the “6 Ws”—who, what, when, where, why and weapon. During the pilot project we did an exercise where we asked telecommunicators to construct a perfect tweet, aka a “golden tweet,” and what they came up with is that such a tweet would need to contain information tied to at least one or two of the 6 Ws.

Insights: Why is “weapon” included?

Tapia: When a 911 call is received, the caller always is asked first about their location and the type of emergency they’re experiencing. As it turns out, call-takers also are trained to ask them whether weapons are involved—it’s a question they ask every time they handle a 911 call.

Insights: What was the outcome of the pilot project?

Tapia: We designed a tool that automatically organizes tweets—after they first have been sorted based on type of incident and location. After the initial sorting, the tweets are organized based on the 6 Ws, and then that information is displayed on a map. One thing we learned was that telecommunicators will not trust this data unless a human in the 911 center reviews it first. Consequently, Jim Lake is hiring a communications specialist—we would call this person a situational awareness analyst—who will scrape the social media data using the tool we developed. This person, on a normal day, will use the tool to look for anomalies and/or intelligence to effect a more informed response. The tool also will be queried to look for issues—for instance, after a call has been received concerning an accident on a specific road. Relevant data then will be entered into the call-taker’s system just as if it had been received via a 911 call and not social media. This data will enhance the situational awareness of first responders who are racing to the incident.

Insights: What about the pilot project surprised you?

Tapia: The willingness of Charleston’s leadership and personnel to accept the social media data output. We were under the impression that they were rejecting social media altogether. We assumed that they believed social media is untrustworthy and didn’t want any part of it. But that’s not true—they want the social media information, they really do. What they don’t want is another system, another map, another layer.