AI helps responders mine social media for public safety insights
AlenaPaulus/Getty Images
An artificial intelligence platform analyzes social media photos and texts to give responders more situational awareness.
What started in one county as an effort to monitor social media discourse related to the COVID-19 pandemic has now expanded into other areas of emergency response, led by volunteers and supplemented by an artificial intelligence-driven system.
Montgomery County, Maryland, first deployed its Virtual Emergency Response Team (VERT) to identify keywords on social media platforms related to the COVID-19 vaccines, hesitancy and conspiracy theories to get a sense of positive and negative public sentiment.
After seeing success with VERT’s use of Citizen Helper, an AI-driven tool that learns to identify positive or negative sentiment from relevant keywords, Montgomery County is looking at how that tool can be useful in other emergency response settings.
That could involve analyzing photos and text posts on platforms during extreme weather events to show the location and severity of issues that are occurring. By mapping those insights, emergency managers can assess where and what kind of response is most appropriate.
“If there is a picture from your local area that has come up on Instagram during a disaster, it should be able to tell you that this road is flooded and you may not want to send your most expensive fire trucks through that area,” said Katelyn Keegan, VERT team lead in the county’s Community Emergency Response Team (CERT), during GCN’s Public Safety Workshop this week. “Or you may say you need to deploy Bobcats and [rescue] skids to the northern part of the county so that people can get help.”
Researchers from Brigham Young University, George Mason University, the University of Texas at Austin and Virginia Tech initially developed Citizen Helper through a multiyear federal grant from the National Science Foundation to study how humans and AI can team up to aid disaster response in Montgomery County.
In the initial stages of the program almost a decade ago, VERT volunteers were encouraged to submit information or photos to the county during floods or other emergencies through a Google form, to help emergency managers determine where the response needs were greatest. Now, VERT members also scour social media during extreme weather to assess public sentiment and see where the biggest impacts are that require a physical county response. Citizen Helper will also assist in those efforts.
Keegan said anyone in the county can be a VERT volunteer, with the time commitment flexible depending on each person’s schedule. Around 1,000 people are registered volunteers, she said, with 45 of those regular contributors to the program. Volunteers are encouraged to take an online course from the Federal Emergency Management Agency about social media in emergency management, but otherwise there are no other requirements, she added.
That FEMA training helps volunteers understand the different social media platforms, how they work and the kinds of information people post on them. It also warns volunteers of potential misinformation: Keegan said an oft-shared photo on social media during Hurricane Sandy was of sharks washing up on the streets of New York City, but that was taken from the movie The Day After Tomorrow.
“Not only do you know those are out there, we train our team to be aware of them and to not record them as actual emergency situations, unless they can verify that something has gone terribly [wrong],” Keegan said.
VERT has already been deployed abroad to analyze social media photos from Hurricane Dorian when it hit the Caribbean and during earthquakes in Nepal. Keegan said virtual emergency response proved itself to be “very valuable” during the pandemic and can continue to help in future disasters.
“When you're going through the emergency yourself, sometimes it's a little hard to say, ‘I'm going to go and check Facebook to make sure that everybody else is okay,’” she said. “This would allow somebody else who's in an area unaffected by that particular disaster or emergency to be able to be able to contribute in a positive way and to assist in the process.”
NEXT STORY: How co-ops can deliver faster, less expensive procurement