In October 2018, Hurricane Michael devastated the Florida Panhandle and wreaked havoc in other states. At least 59 deaths in the U.S. were attributed to the hurricane, with damages estimated at $25 billion.
In the chaotic aftermath of the storm, people were contacting United Way’s 2-1-1 service in Florida to ask about food, shelter, evacuation zones, road closures, and other needs.
To help out, IBM and United Way Worldwide deployed CARLA, which stands for Cognitive Automated Response Learning Agent, the day after the hurricane made landfall. The chatbot can take the pressure off call centers by responding to simple queries, such as, “Where’s the nearest shelter?” For more complex questions, it would pass them along to human operators who now had more time to deal with them.
The chatbot was designed in 2018 as part of IBM’s support for Florida’s recovery from Hurricane Irma in 2017 so the state could be prepared for future disasters.
In just days, CARLA answered hundreds of questions related to Hurricane Michael, in both Spanish and English. CARLA uses artificial intelligence to better understand what people are asking. It can provide the right answers even if people are using incorrect grammar, misspelled words, or referring to the same thing in several ways. It also analyzes its own interactions with users to better improve its responses.
The technology can be quickly used during future disasters and crises in Florida on 211.org and local 2-1-1 websites.