UCD + Designing for Voice: Chance — the healthy eating assistant.
- User-centered design class at University of Washington (HCDE 518).
- Wanted to help people overcome procrastination but after initial research, aggressively refined our goal to: help young professionals achieve healthy eating goals.
- Ideated a robot assistant that users would set up, and it would guide them through cooking meals that aligned to their individual goals, and built a paper prototype for it.
- Created and tested two basic usage scenarios with three potential users and improved the interactions with their feedback.
- Put together a presentation for our professors, class, and guests that included the product, research, and long term vision and high level business strategy (plus LASERS).
- Had a blast when we realized Amazon’s Alexa acquired a very similar skill as we wrapped up our project (albeit not as health-focused as ours).
- Familiarize ourselves with the user-centered design process.
- Do basic research to identify an area of opportunity and design question that our project will address.
- Create a basic solution to address the design question.
- Validate the solution with users through user testing.
- Incorporate user testing feedback in an iteration of the solution.
- Present the solution before our classmates, professors, and industry guests.
Initial Problem Space
Our team originally converged with a mutual interest in procrastination. We initially aimed to help people use the misspent time towards working in their tangible goals. Midway through the process we got feedback from our professors that given the scope and time frame of the class, we should narrow our scope considerably.
Original design question: How might we help young, urban professionals disrupt procrastinatory actions and achieve their self-development goals?
Updated design question: How might we help young, urban professionals achieve their goals of eating healthy?
Surveys + Initial User Research
We created a survey questionnaire using Google Forms to capture participants’:
- Goals (type, timeframe, etc.)
- Workday duration
- Activities carried out during workdays, week-ends
- Amount of “free” time during workdays, week-ends
- Preferred method to work towards goals and track progress
Five semi-structured interviews took place with a variety of participant age ranges. These interviews centered on motivations, challenges, strategies that have helped participants achieve their goals, and details on the process — how they felt during the process and when accomplishing their goals to understand motivations and incentives that people have to work towards their healthy goals.
We carried out two contextual inquiries with three participants. In typical fashion, a brief semi-structured interview took place, and then one participant was observed during a reported typical weekday, while the other two observed during a reported typical week-end day. By triangulating the insights from the surveys and interviews with the contextual enquiry, we were able to evaluate how the individual routine of participants aligns to being able to eat healthy and identify specific challenges that might not be described during an interview or captured in a survey response.
• 75 participants completed the on-line survey.
• 5 individuals were part of the semi-structured interviews.
• 3 individuals in two households took part in the two contextual inquiries.
Initial Research Findings
Through our survey findings, we discovered that the majority of our target demographic (94.3%), have goals they currently aren’t achieving. The main categories of goals we found people had were physical health goals, career goals, financial goals, and mental health goals. A structure or system that fits into participants’ daily routines/activities was identified as extremely helpful.
More than half of the interview participants reported that fatigue after a day at work and having to do chores stopped them from working on goals; they prefer to do easily rewarding tasks like doing very lightweight cleaning, and/or more leisure activities (watching TV/streaming media/gaming).
Originally, target users were young, urban professionals that are between ages 25–35. After the initial user research, we refined our target users to be young urban professionals, ages 21–35, who work between 8–12 hours per day, 5 days a week. They have short-to-medium goals all related towards eating healthy.
Based on the goals (and the variables that help participants to accomplish them) we identified the design requirements for the solution:
- Portability — Allows for user to access information on the move. We will likely design something that allows for a mobile application or wearable.
- Adaptability — Design should be flexible and adaptable enough to help users build or re-work a system to help them reach their goals.
- Personalized — Information should be flexible to cater users needs and offer suggestions based on user’s input. For example, someone puts in what types of food they don’t like so the design only offers suggestions that exclude those food types.
- Timeframe Sensitive — Our design should also help users achieve goals within 6 months to 1 year.
Originally, three different personas were developed. After some recommendations from our professors, we centered in one:
Andrew: Recent graduate with little cooking skills and time, whose work and hobbies take most of his time, and someone that, using the right time management and cooking help can achieve his healthy eating goals.
The second persona was Theresa — we only created her persona, but did not create any further usage scenarios for her due to time constraints.
We came up with 15 ideas on how to solve for the scenario + persona. We also validated the top three with some of our potential customers from our target audience. The top three that we selected were:
- A printed stick-on tattoo that contained a circuit that could be programmed to interpret nutritional state based on skin contact.
- Good: Relatively discreet and you can always see the current nutritional state.
- Bad: Awkward in certain social contexts — battery life and disposal challenges. Hacking risks are a concern.
2. A self-piloted drone that would bring in food/ingredients on-demand.
- Good: No need to worry about having ingredients at home — freshness always guaranteed.
- Bad: Legality of self-piloted drones is a gray area. Additional risks in case the drone fails and causes accidents.
3. 3D Printer capable of printing recipes — either custom owner-created recipes, or recipes from professional chefs. It would also have a screen to show the final product.
- Good: Leverage social networks/big distribution media for recipes (YouTube/Facebook/etc). Preparing meals does not require significant skill.
- Bad: Need to be able to have the raw material/ink to print the food from, plus manage the creation, stock, and distribution of said raw material/ink to print from. Current 3D printing technology does not produce appealing results. Will not help users learn how to prep food.
We decided to do a mix of these, plus use a few additional characteristics that would allow us to reduce the potential negatives that we obtained in our concept validation. Our proposed solution became:
A voice-controlled assistant that can guide you to prepare healthy meals that align to your personal goals. It will be able to identify personal trends and preferences to suggest meals that you like and facilitate the process to obtain additional ingredients when necessary.
We sketched a couple of ideas based on attributes we identified:
- Should have a relatively compact footprint since it will most likely be placed in a kitchen counter.
- Should have hints of a character/personality — it is a combination of assitant/friend/guide, and a personality makes it easier to identify with.
- Should allow integration of other sensors/devices — ability to project an image, or integrate with bluetooth devices.
We came up with the first paper prototype of our healthy eating assistant:
After a few iterations:
We also created screens for a phone companion app to set it up and configure additional settings:
Voice Interaction Design
Since users will speak to their assistant, we created an interaction flow for our core scenario:
User Scenario Validation (with users!)
We used a paper prototype to familiarize participants with the solution concept and to make it easier to build a connection when interacting with a new device.
The Usability Testing focused on two big processes: OOBE set-up, and “cook-off”, where the user cooks one meal with help from Chance.
Testing the voice interactions was challenging — we created pre-recorded dialogs using Google Translate. The audio conversation was downloaded and embedded in a PowerPoint slide so we could set the context, play the voice prompt, and then capture the participant’s response:
From our testing, we refined the OOBE and the voice UX design.
- The OOBE provided more details around permissions, and offered info on additional services that could be connected to Chance (Amazon Fresh, InstaCart, etc).
- The voice UX was changed to offer menu options one by one rather than a barrage of 10 options — participants got confused with too many options without enough time to think them through and decide.
Wrapping It Up
For our presentation, we created the high quality screens incorporating user feedback:
After a visit to Ikea, Chance became real(ish):
We created a presentation with the process and some cool long term vision ideas like:
- Cooking is a very humane activity — it can bring together families and friends. As people relocate farther from home, Chance could enable telepresence to bring those friends and families closer (albeit digitally). We could use its eyes as cameras, and the dot on its forehead as a projector so you can cook with distant family members.
- The projector in Chance’s forehead can also be used to guide you visually through the cooking process — how to efficiently peel garlic, or fillet a fish through video.
- Chance can also help one achieve sustainability goals — I should be able to tell chance that I want to get seasonal foods as to reduce my carbon dioxide footprint, or use ingredients from my local food coop.
- Recommended/popular healthy eating programs can be featured by Chance based on my goals — training for a marathon? Get the Runner’s World diet program. Cleansing or specific diets? There’s a program for that!
User-centered Design is awesome! Our participants really enjoyed the idea, and their input was extremely valuable for our prototypes. Rather than create another gadget that seems half-baked, we defined the right personas and then solved for their needs through the UCD process.
Designing for voice is fun — and hard. Here are some principles we identified after testing the voice flow with participants and redesigning it afterwards.
Voice UX design principles:
- Set yourself in the most specific context possible — you should design for the most granular scenario.
- Present the user with clear and specific options — ambiguity is your worst enemy when designing to expect potential responses as inputs.
- The flow should be as minimal as possible — it is not convenient to have a five minute conversation to do a task that can take seconds.
- Endless lists of options are not a good prompt from the assistant/device. One by one is much easier for the user to respond (each response will be yes/no/maybe) — besides… when you’re figuring out what to do, would your friends/spouse drop an ordered list of 10 options? Not really.
PS — Why the name “Chance”?
We originally were 4 team members. Due to some work reasons, one team member had to drop the class. We wanted to include him in the process, so we reached out to him: