$O$ - Social entrepreneurship
Impact
Got more than a dozen people off the street and into permanent housing
Fed 100+ people in partnership with our local Taco Bell franchise
Developed a volunteer program to talk to people experiencing homelessness via SMS
Made partnerships with other local and national organizations to help as many people as possible
Met with congressional representatives and talked with them about our mission as it relates to policy
My Role
Designer/Researcher - Responsible for all the research design, training the volunteers on how to interview clients, communicating research results to stakeholders, and designing the automated survey engine.
Project Summary
$O$ was an information-driven technology startup by Ice 9 Design aimed at helping reduce or end homelessness by leveraging AI and UX techniques to gather information from the unhoused population and translate that into narratives that we could take to policymakers to create long-term change. $O$ was a chance to try some new technology ideas while also doing some good in the world.
We achieved our goals of getting people fed and back into housing, created a model that we could use for later projects, and proved that our AI framework was viable. We were looking at getting a real budget before we shut down due to COVID.
The Challenges
The unhoused have many pain points so we were spoiled for choice, but we didn’t have the resources to solve most of them. We decided to focus on what we could solve. Since internet access can be difficult for the unhoused, we zeroed in on providing information. Information, specifically information about services that they can use right now in their local area or services they qualify for and how to apply for them, can be very useful for the unhoused.
The other thing that we focused on was paperwork. If a homeless person has lost their ID or other critical paperwork, it can make it very difficult to get a job or otherwise participate in society in a way that would help them get off the street. Helping the unhoused get paperwork filed was a low-effort, high-impact way to help them that lined up with our focus on information.
This project also had an internal user because it was a technology demonstrator. We had a concept for an AI framework that we wanted to prove. We needed to create an automated system for effectively interviewing our users. The challenge two-fold: none of us were engineers, and we wanted to do something new. Our main developer on the project was a designer who happened to code, and the idea we had was something that was mostly just theory. We had to convert theory to practice with very little coding skill.
Solutions
There were two users that we were building solutions for in this project. The first was our external users because we wanted to help people. The second was an internal user because this was a technology demonstrator.
For the external client, we needed to create a system that would allow us to efficiently gather and transmit information to the unhoused. With our focus on information, we need two things: a way to communicate with the unhoused, and a way to communicate with each other. A way to automate the system would be a bonus. There were many chat services that would allow us to communicate with each other, but we could only communicate with the unhoused through text messages.
We looked at several free chat services but decided on Discord for several reasons. The primary one was that it was built to allow bots to function easily and many bots already existed, including one that would act as an SMS bridge. There was also a free dev kit that would allow us to create our own. This gave us everything we needed for our external clients.
With many of the unhoused having cell phones, we decided to use SMS as our communication method. We set up a bot on Discord, a free chat server that would receive text messages through Twilio. Users would be sent some automated messages, then asked for their name and one of our volunteers or staff would be paged. After making sure there was no immediate emergency, we would order them some food online (usually from a nearby Taco Bell) and conduct a qualitative interview while they went to pick it up.
This part of the interview was intended to get stories to build narratives we could take to policymakers and was about how they became homeless and their experiences while being homeless. Once they had eaten, we would start the automated survey, which was run by our AI framework. We would introduce “Anna”, who was a bot but was better at filling out paperwork than we were. She would ask them some questions to help them figure out what services they qualify for. A volunteer would monitor the process and could interrupt it at any time. Once they were finished, we would process the survey and send them the results.
The automation involved was, in some ways, a much more interesting challenge. As already stated, we had very little technical knowledge, so any solution we designed would have to be simple enough for our designer cum dev to execute. We also had to come up with something relatively unique. The idea we wanted to demonstrate was ‘intelligence in aggregate”. Instead of one giant program doing everything, we would have multiple programs working together to achieve a shared goal. Each of them individually would be dumb, but all of them together would be smart.
We used BotKit to create multiple bots to successfully achieve this effect. Each one did one thing, but they could all talk to each other. By carefully programming each bot to respond to the others, the overall effect was that of a capable chatbot that could pick the right question to ask and respond to users with variable timing.
Storyboard
 sos-frame1.png |  sos-frame2.png |  sos-frame3.png |
|---|---|---|
 sos-frame4.png |  sos-frame5.png |  sos-frame6.png |
 sos-frame7.png |  sos-frame8.png |