Design Exploration

Prototyping Solutions

Focusing on Screening for SNAP benefits

Through our research, we strove to understand the benefits ecosystem and gain perspectives from applicants on the entire process. As part of design, we narrowed our scope to target screening due to privacy and data storage reasons.

For prototyping, it is simpler and more secure not to store private information. Benefits screening is the first step which can introduce potential applicants to how much they qualify for in benefits and highlight policy. A goal is to incentivize people to apply and ensure they get the benefits they deserve.
‍
SNAP was also targeted as part of the benefits space. The screening questions for SNAP are the most comprehensive in terms of collecting general, household and income specific information. By adding on a few questions, screening for SNAP can be extended to screen for other benefits such as health coverage (SSI), specific programs for women and children (WIC), and heating assistance (LIHEAP).

Using Policy to Guide
Design Prototyping

To explore policy and apply it through design, we chose to hone-in on policy around exemption for college students qualifying for SNAP. Students do not typically qualify for SNAP but this is assuming a typical student who has college paid for by parents. According to CLASP, a "non-profit focused on policy for low-income people," 1 out of 4 students fall outside of this.

We chose to tackle building in questions that could help college students qualify in the process. To do this, we leveraged the SNAP Handbook and policy detailing exemption from supplemental sites such as CLASP.

“If students are unable to finish their academic program, they have few opportunities for careers that provide long-term financial stability. Likewise, states are less able to produce a racially diverse workforce that strengthens the state economy and helps to improve the quality of life for all.”
- CLASP brief

Voice Technology

We learned from early prototyping feedback and have iterated and grown our designs. While exploring voice tools, we did a deep dive into Dialogflow offered by Google and into Alexa because we recognized these platforms as having robust technical documentation, support for custom language models, and expandable frameworks. Taking a voice-first approach, we have pursued prototyping using Amazon Alexa as the console provides an effective way to build, code, and test. Alexa's utterance detection and custom voice models allow for modification and documentation for screen interfaces offers an opportunity to provide support to augment a voice experience.

Design Prototyping

We pulled from policy to guide how we thinking about screening for SNAP. We implemented policy for student exemption as part of a voice prototype built with Amazon Alexa which allows us to highlight policy while fleshing out an entire screening to include questions on household and income to conduct a calculation to estimate eligibility and benefits someone could receive.

Our prototype is evolving and we see this as an initial step. Beyond screening for SNAP we see potential to help applicants connect to resources, expand screening for other services, explore voice with screen interactions, and envision how screening can link more directly to applying to improve the overall process.

"The term MVP: Minimum Viable Product is that version of a new product which allows a team to collect the maximum amount of validated learning." - Eric Ries, The Lean Startup

‍

Bucket Structure

We used something called a bucket structure to organize our screening tool.

Guiding Responses

Through design, we have included verbal cues which are provided by Alexa when she asks a question. For example, Alexa might ask “What is your living situation?" and cues "You might say, I live with my parents, I live with roommates, with others (it doesn’t matter if relatives or otherwise), or I live alone.” Through our design of this type of question, we are guiding someone as to what they might say.

Understanding Intent

Through design, we can use the Alexa Developer console to anonymously track the types of things that someone would ask to the screening tool and collect this anonymously along with utterance failures (where the language model fails to recognize what users say). This can help shape our design to be more accurate in language detection and incorporate feedback to develop things we did not anticipate.

Dynamic Calculation

Based on answers to screening questions, we are able to do a real-time calculation to estimate how much someone is eligible to receive in SNAP benefits. This calculation takes into account household, income, and finer details. The advantage of technology guiding the screening is that questions are dynamic based on individual applicant responses. The process can be repeated reliably.

Our Vision

We are considering how voice can touch areas beyond purely screening such as helping applicants navigate answering questions which allow them to proceed further in the application to understand if they qualify for SNAP. We also have the idea to build out local resources which we can help connect demographics (initially students) to support in their communities.
‍
Initially thinking about how to deploy our screening tool, we wanted to place Alexa devices in public places. However, due to the current environment with Covid-19, we are planning to request the downloads of the Alexa app as well as purchase and enable Alexa devices with our skill. We see this as two pathways, the first as a mobile option for a voice assistant available on the go and the other as an at home situated option.

Development and Testing Goals

For development, we plan to explore how screen interfaces can supplement voice-first experiences.

As part of our research initiative, we are identifying participants for a multi-week study and plan to deliver them devices as part of a testing plan for deployment. Creating simple instructions to go with test kits will be an important prerequisite to the study along with devising a plan of how we can instruct participants to use and record their interactions with the device. 
‍
Beyond use with our benefits screening skill, we hope to also gain insight into how research participants choose to use Alexa for other purposes. By getting self-reporting on how participants might use Alexa for other tasks we can get a better idea of our participants’ views toward voice assistant technology.

"The term MVP: Minimum Viable Product is that version of a new product which allows a team to collect the maximum amount of validated learning." - Eric Ries, The Lean Startup

AS PART OF DEVELOPMENT AND TESTING

Things on our mind regarding design through the summer

User Privacy

We noticed that there are legal considerations when deploying a skill on Alexa and cannot collect personal information such as phone numbers, user addresses, or identifiable data. Our goal is not to collect this type of information and any utterances will be anonymous and used for the sole purpose of informing better design.

Non-technical Avenues

As a note to policy, through our research we were touched by the experiences of people we spoke to through our focus group as well as through guerilla research. Limited access to technology as well as disabilities impact populations seeking assistance. Through exploring non-technical avenues, we may address how the current state can improve through changes to policy at a government or NPO level.

Flexible Accommodation

For those who are applying for benefits through paper-based applications, we want to meet where they are and not introduce a higher learning curve in getting adjusted to new technology that they cannot afford. We are considering new ways on how to embed this workflow into other devices and form factors for greater accessibility - such as smartphones, tablets.

Link to Medium Blog