Capstone Care+

How might we design a low-cost smart speaker solution that supports and motivates individuals living with MCI, so they can live independently for longer?

This was a 2-semester-long capstone project in collaboration with IQ Solutions to design high-level product concepts to support individuals living with MCI (Mild Cognitive Impairment), early-stage ADRD (Alzheimers Disease and Related Dementias), and their caregivers. We designed a system using smart speakers and a smartphone that allows individuals with MCI to complete and confirm activities based on time, task and proximity. The design and prototype to be developed during this project will be used to help secure funding for future development.

We used a modified version of the Remote Design Sprint Methodology to test several different product concepts across 5 sprints. We extended the sprints to 4-5 weeks each. Each sprint had a combination of research, design, and testing prototypes with users.

Sprint 1: Understanding
Detecting Proximity

The Problem Space

We started our journey with some baseline research of the problem. We were surprised to learn that MCI is more common than we thought. According to the World Health Organization, a predicted 15-20% of people over 65 have MCI. Dementia is as well: 50 million people worldwide have dementia, with 10 million new cases per year.
And it's costing us a lot. The total monetary cost of dementia in 2010 was $157-215 billion. This means that the average yearly cost per person worldwide was $56,290.

The Scope

We first designed for the baseline technology. IQ Solutions provided that the system must build upon currently existing smart speakers to keep costs low-- in order for the system to work best, users would ideally have multiple smart speakers throughout their homes.

We determined that proximity detection was the most pressing issue to tackle first. Smart speakers remind the user at a designated time, but the user may not be close to the speaker, or might not be in the correct room or situation to complete the task. This is especially an issue for individuals with MCI and early-stage ADRD; they might forget the reminder before they go to the correct location for a task. Being unable to successfully complete IADLs (instrumental activities for daily living) at the right time can prevent this group from aging in place independently in their homes.

How might we detect when someone is close to a smart speaker to provide notifications at the right time, in the right place?

The Initial Sketch: How does it work?

From various presentations we attended and our talks with experts, we learned that medication adherence is one of the biggest issues for our user group, so our scenario involved the user setting up the system and the system reminding them to take their medication in the correct location and time. Once the user confirmed that they took their medication, the system would send a notification to their designated caregiver. This is the sketch I did, which we ended up using for our final design.

What's in the system?

From our problem space, we designed several possible solutions for detecting user proximity, and decided to go with a solution that could utilize several devices: beacons, smart speakers, the user’s phone, and a wearable.
In future testing sessions we hid the beacons unless asked, because we wanted more feedback on other portions of the prototype beyond "how it works."

The First Mockup

We created a clickable mockup using Figma that guided the user through a narrated scenario, which set up the speaker, beacons, watch, and phone app. I laid out the onboarding phone screens.

We didn’t have IRB approval for our project at this point, so we were not able to test with our target user group; individuals with MCI and early stage ADRD. We tried to reconcile this by just testing the prototype with older adults.

We started the test with an onboarding process that took them through the system's components.

Users can add their caregiver info, so they can be invited to use the app as well.

Devices are connected step-by-step. We later switched this to much faster toggles.

Onboarding would go over the capabilities of each device and their role.

Connecting multiple devices quickly via bluetooth

Users would be able to customize the system more to their preferences and needs.

In the prototype, participants would make a task for medication reminders.

Tasks are associated with a rough time and a place.

Testing Challenges

Because we were unable to test with our target users, interviewees had some misconceptions about individuals with MCI and their capabilities. When asked to imagine themselves as an individual with early stage ADRD, interviewees thought that they would not be able to do any of the tasks presented in the prototype.

We later presented the prototype to Emma Dixon, an early-stage ADRD expert, and she said that the tasks would not be too complicated for our user group. Despite this, we feel that the design could definitely be simplified for a better user experience.

There's no way someone with dementia could use this. There are too many components.
-Interviewee without MCI

This definitely isn’t too complicated-- people with early stage ADRD are now more and more tech savvy.
 -ADRD expert

Initial Learnings

Simplify, simplify, simplify. 
Users like the calendar apps and devices they currently use, so we should integrate with them rather than replace them. Onboarding and setup for this system needs to be smoother.
”Privacy is a concern of mine, but I’m ok with some sacrifice.”
 Users are willing to give up some privacy if it means they can age in place for longer.
More compelling use cases. 
People already have methods of remembering to take medication that work.
“How it works” isn’t the the focus. 
We shouldn’t try to fully explain the inner workings of the system during interviews-- it becomes too much of a focus instead of the conceptual features we really want to test.

Sprint 2 + 3: Research and Exploration
Motivation and our Target Users

Going forward we...

Did lots of research.

Shifted to joy and motivation.

Found our target users.

When we ended Sprint 1, we were still lacking necessary understanding of our target user group. To fill this knowledge gap, we conducted a lot more supplemental research before tackling our next design-- we watched video interviews from individuals with MCI, read research papers, spoke with dementia and aging experts such as Emma Dixon, and watched expert presentations.

There are a lot of solutions out there to solve medication adherence. However, there aren’t a lot of technological solutions to preventative care for the condition’s progression, and solutions that bring joy into these people’s lives. According to our research, staying socio-cognitively active is key to staying in the MCI stage of ADRD for longer, so this became our Sprint 2 and 3 focus.

New Focus: How might we use motivational factors ("the why's") and contextual information to suggest the right things at the right time in a conversational way?

Our Key Design Concepts

Integrated multi-device system of the user’s phone, smart speaker[s], and smart watch.
Proximity-based reminders
- The smart speaker speaks to the user only when they are nearby, especially when they are close to a specific task: for example, taking out the kitchen trash when in the kitchen.
AI-based Activity suggestions
- When the user has a span of empty time in their calendar or is sitting down for a long period, the system gives activity suggestions based on their interests and needs.
- It uses contextual information to recommend the right things at the right time: for example, using weather, noticing a large gap in calendar plans, smart watch activity / movement, social media, user location, upcoming holidays and birthdays, etc.
- The system has brief onboarding questions to weigh suggestions, then intermittently asks more detailed questions to get to know the user better.
Companion device / "motivational nuggets"
- We researched different methods of motivational techniques, which are integrated into the prompts Alexa gives to motivate the user.
- The system recognizes progress and streaks to further motivate the user.
- Progressive intensity for some tasks.
- Conversational and friendly tone.

Testing Key Concepts

To more smoothly test our ideas, we decided to go with a video showing contexts of use with 2 different use cases. This enabled us to focus on user reception for our larger concepts rather than high-fidelity interactions.

We got IRB approval, so we were able to do 12 interviews to test our 2 prototypes! Six of those interviews were with individuals with MCI, and the other 6 were with caregivers of those with ADRD. We recruited through snowballing and Facebook support groups.

Feedback during interviews

This would be immensely helpful for me. I have depression too, so it's really difficult to get off the couch sometimes.
 -Participant with MCI

I like that it's gentle and wouldn't make me feel stupid or judged like when my family tells me to do things. It seems like it would be easy to use.
 -Participant with MCI

I like that it's more proactive than reactive. It encourages and stimulates you like a coach.
 -Participant with MCI

What did people think about our concepts?

1. Is this more motivating than what people are currently doing?

Yes: Both caregivers and users with MCI found all features to be motivating, and reduced burden on themselves and loved ones.

2. Will people want to do activities suggested by the system?

Yes: Most participants appreciated activity suggestions due to positive reinforcement, and usefulness in habit development and maintenance.

3. Do people find the system intrusive or invasive?

No: As long as they could turn off some data sharing features, users had minimal reservations.

4. Do people like the system having a more conversational tone?

Yes: The conversational aspect was viewed as a key competitive advantage compared to other devices on the market.

What we learned about our users...

Our target users are typically younger, tech savvy, and social.
The MCI population we’ve spoken to so far tends to skew younger (30s-60s) and more tech savvy. We are building for a future audience, so this population will be increasingly tech savvy.

Caregivers aren't part of the equation yet!
Though family members may help out with some tasks, individuals with MCI still complete most tasks independently and usually do not have formal caregivers. This is different than we first thought.

"I'll do what I want, when I want to."
Joyful and motivational aspects are helpful for those with MCI, but likely less helpful for later stages of ADRD when people go more with the flow. Caregivers for those in later and moderate stages felt that their care recipients would not want to do activities recommended by a speaker.

"I would be fine with that feature, as long as I can turn it off."
Users want a very high level of customization to fit their personal lifestyles, routines, and preferences.

Sprint 4+5: Implementation
Making the Final Prototype [WIP]

Since the concepts we tested were received well in video form, we moved forward with creating a higher-fidelity prototype for both the Alexa Skill and the phone application.

Since participants wanted a high level of customization, we decided that onboarding would be the focus in Sprint 4. In Sprint 5, we built out the rest of the system as it would function for 2 different use cases: gardening and guided mindful stretching.

Product research: What onboarding questions should we ask?

How might we ask for just enough information upfront, to maximize immediate functionality without being too cumbersome to set up?

I was the team facilitator for this Sprint, and I had to figure out how to reformat the Design Sprint methodology for our newly product-oriented goals.

We started out by doing a lot of product research of apps for habit-forming and goal-setting. We researched over 20 different apps for this process.

Onboarding Prototype

We started with a Wizard of Oz testing method for the Alexa skill, by creating a detailed map of possible inputs and corresponding outputs, and pre-recording all outputs using Amazon Polly. This worked especially well for gathering potential user inputs, since we haven't built out the full breadth of user responses yet. During Sprint 5, we are using VoiceFlow to fully build out the voice prototype.

We used a Figma prototype to build out the phone onboarding. We split into two working teams; I was on the phone app team and helped with layouts and visual design.

OTHER PROJECTS

Care+

Sprint 1: Understanding​Detecting Proximity