2017 Debug Politics Hackathon
EMBOT
Can't get through to 911?
Ask Embot for fire, police or medical assistance by texting through facebook messenger or the web
Introduction
In December of 2017 I entered into San Francisco’s 2nd Annual Debug Politics Hackathon. The Hackathon is a 3-day event where people in tech come together to transform the political landscape of America. Created in response to the November 2016 election results, it was designed to channel people’s political passion into actual action.
Crisis Management Track
Groups were split into 5 separate tracks centered around politics and government. I was assigned to was “Tech + Crisis” track. The focus being: How can tech be better utilized to save lives and improve crisis management during a city wide disaster, such as an earthquake, hurricanes, or fires.
TEAM & ROLE
UX/UI Designer
Graphic Designer
Developer
Electrical Engineer
Our team was lucky to have a great balance of artistic and technical skill. My role morphed into a Product Manager where I led a majority of the research, as well as triaged our teams resources into various tasks.
Research
Understanding the Problems of Crisis Management Through the Eyes of State Services
Debug brought in emergency service experts to help teams understand the problem. Neil Bragman had just gone through the terrible 2017 California fires so we heard many harrowing stories of evacuation and rescue. David Cruise of the Governor’s Office showed us the Saas technology he used to monitor the Southern California fires, which at the time of the Hackathon had still not been fully contained. We were so thankful to have such great reservoir of knowledge to tap into.
Choosing a Problem
Although there were many problem areas expressed in our expert interviews, the theme of communication was always constant. Emergency services are prepared for all sorts of crises, but the common citizen is not. Some of the information problems are:
- How do we better notify the public of what to do when a disaster strikes
- How can different government departments talk to one another and stay on the same page
- How can we get information faster and in higher fidelity from the person in distress to the services who can offer assistance
People in Disaster use Twitter for Help
When we heard the story of how Hurricane Irma victims used Twitter for their pleas for help, a particular gap in emergency protocol caught our attention. When a natural disaster happens 911 often gets flooded with calls, but there are only so many 911 operators able to pick up the phone at any given time.
This was the problem we decided to tackle
•How do we allow users to contact emergency services when 911 phone operators get overwhelmed?
•How can 911 use social media and new forms of communication to interact with people in distress?
Ideation
Initial Idea - Twitter Scrapper
One of the requests from our advisors was to build something that would allow the EOC (Emergency Operations Center) some way of utilizing social media. Because of the hurricane Irma story, our first idea was to create a Twitter scrapper. This program would sift tweets through natural language processing and other algorithmic parameters. Then it would cull the tweets of interest into a dashboard for EOC to manage.
Roadblock - Emergency Services are Local but Acquiring Location Data Isn't Automatic
However, further research revealed that even though Twitter allows users to demarcate their location, less than 15% of twitter users actually geotag their tweets. This is a problem because 911 services are local, they tend to stay in their district unless they get a call from another EOC request for assistance. Therefore, each program we create for one particular EOC is only interested in seeing tweets in their particular area. So if users don’t know to geotag their tweets, our scrapper won’t be able to know which tweets are in their zone.
Possible Work Around to This Problem
One possible solution to this problem is to have particular hastags for each EOC.
However this method involves training users to tweet in a very particular manner. Reeducating all users to tweet in such a specific manner would take a huge marketing endeavor.
Alternate Idea - A 911 Chatbot to Take in Requests
Because of the needed education requirements to our Twitter idea we decided that if people want to communicate with their particular 911 center, a chatbot could provide much of the same utility as a 911 dispatcher. Obviously a chatbot cannot provide the calming vocal presence of a trained professional, or specific contextual advice, however it could gather much of the same information required to send out a proper response.
The Merits to Building a Chatbot
- Facebook Messenger and other chat apps
- On the 911.gov site which currently has no interactions with users in distress
- In the future texting 911 directly will be common place, however this would require buy in not only from the national and regional government agencies but from telecommunication companies
How Does This Help 911 in a Disaster? - Use Case Scenario
Imagine an 8.4 earthquake hits San Francisco. People are panicked and 911 Emergency is being flooded with calls. Your neighbors’ house has caught on fire and you have an immobile elderly staying in your home. You’ve been on call with 911 for 10 minutes and keep getting a busy signal. How are you to notify 911 to help you move your guests and also put out the fire? With a chatbot, 911 will be able to receive your request.
DESIGN
What Information Does 911 Need to Respond?
Location
This is the most important piece of information to gather because if in the instance the user is cut off from the dispatcher, the dispatcher will still send a police officer or ambulance to the location.
Nature of the Emergency
Dispatchers need to assess the situation properly in order to send the right kind of resources to the incident. The more information they have, the more accurate of a response the User will receive.
Description of the Person or Place of Interest
When police or medical personnel arrive they need to be able to identify the problem area accurately and swiftly.
- Person: Age, Sex, Body type, Hair Color, Clothing, Name
- Place: Condition, Orientation to Other Objects, Potential Hazardous Materials
Additional Resources: Calling 911 Guide
Flow Charts
Mockups By Marie
Facebook Messenger
Chatbot Prototype
Results
At the end of the hackathon 10 teams presented their work and our project earned a 2nd Place Award. Below are some of the testimonials from Emergency experts and Hackathon hosts.
SF Department of EM
SF Department of EM
NExt Steps
More Research about Emergency Operations
Although we had experts to verify protocols and typical User behaviors we still need a more thorough baseline of how emergency operations work, both on a macro level and the day to day minutiae. Our advisors graciously offered to let us shadow them at their EOC headquarters in San Francisco.
User Testing
Although we tested one flow with a few users successfully, we haven’t built enough contextual paths into the bot to cover all situations. The more testing we do, the more we can understand how a person in peril might interact with the bot.
Increased Mastery of Conversational Design
Since the Hackathon I’ve continued my education of chatbots and attended a Meetup focused on teaching people how to build their own. Using Dialogflow has given me greater insight into the mechanics, potential and limitations of chatbots.
Additional resources for creating great chatbot experiences:
19 Best UX Practices for Building Chatbots
Partnering with EOCs
This Hackathon was a great first step but actual partnerships with City resources are needed to continue the project. If you know of someone who might be interested in this kind of solution, you can reach me in the contacts page.