Product Design, Voice Design
Read in JAMA Pediatrics that "more screen time is linked to poorer progress on key developmental measures such as communication skills, problem solving and social interactions among young kids over time.” The addictive qualities of screen time and links to obesity and sleep disruption with overuse have also been documented.
What can we do? How can we reduce harm and better leverage the potential of technology to make learning delightful? Our only parameter was to include a voice interaction. So, we went off to reseach...
Voice UI, Research, UX, Testing, Identity & Illustration
Judilee Haider: Research, UX, Storyboarding, Animation Dillon Sturtevant: Voice UI, Testing, Filming
1 Both children and adults learn better when facts are connected to a broader context or narrative.
2 People retain what they learn near bedtime more than at other times of day.
3 Women still only account for 29% of the science and engineering workforce, but exposure to STEM subjects and role models makes a difference.
Voice has the potential to be more passive and less addictive than screens, so we decided to make every interaction conversational/voice based.
Space is a perfect focus: It allows us to connect science, history, mythology and cultural awareness, increase STEM learning, and to utilize the learning potential of evening hours.
Who is our target audience and what are their goals and pain points? For Mae, we have two sets of users to design for: the adults who will buy it and the children who will use it.
Tami is very interested in science and is quick with whatever tech she gets her hands on. She has two older siblings that she is always trying to keep up with.
I really like when there are peaceful times I can share with my Mom. Our house is really busy most of the time.
“I want to be just as good as my brother and sister, and when I grow up I want to be a scientist.”
Leo’s familiar with voice tech. His grandma uses Alexa and he thinks it is really fun to talk to her but she often misunderstands his words. Leo is a bright child but is not that motivated toward science learning.
I just want to play video games and on my mom’s phone whenever I can!
“When my friend Miles had a sleepover with me it was really fun to talk to Mae together. He really liked learning about Orion the hunter and I was proud to share my cool toy.”
Kim is an outdoorsy kid and hasn’t been exposed to voice assistants much. Her house is fairly low-tech. She’s an avid reader for her age.
I’m excited but also worried that it will be hard to learn how to use the robot. Toys that are too complicated are discouraging and make me feel frustrated.
“I like the idea that I could feel like I’m under the starry sky in my own room at home.”
Susan’s intelligence and drive rubs off on her kids, and with her partner they are a very busy household of five. They all use voice assistants often at home.
I want something special for my youngest’s birthday.
“My daughter is interested in science and I love leaning in to that. I like the idea of a toy that could create a fun experience that we could enjoy together, even as a whole family.”
Rand’s family has a lot of tech at home and they often buy trendy new digital toys. He enjoys exploring what’s new as well as his two sons, and a love of gadgets is something they all can share.
I like the idea of something that’s not on the computer or ipad and I’m excited about how voice technology is growing.
“My son wants to play video and phone games all of the time. While I use my screens a lot too, I worry about the effects of too much screentime.”
Lexi is willing to spend for quality toys for her only child, but choosy. She is open to new kinds of tech but is careful to limit Kim’s exposure to screens. She is trying to create a healthy, balanced life for her family.
As I venture into digital toys,I want something that my child will continue to play with after the initial excitement of unboxing.
“One of the reasons I haven’t exposed my child to a lot of tech is worry over her privacy and what she’s exposed to. I want something that doesnt put her on the web.”
In order to successfully market a toy, we need to find the overlap between parents’ and childrens’ wants and needs. Parents want their kids to be learning, want the toys they buy to be able to hold their children’s interest after the initial unboxing excitement, and want the toys to bring their kids joy. They also worry about their kids using tech too much, and the dangers of their kids being online unsupervised. Children are fascinated by space and by tech, and love when they can do things “all by themselves”.
With our research and our users in mind, we designed Mae to teach kids science, history, and myths about the night sky for a rich and memorable experience. We named our toy Mae after the first African American woman in space, Mae Carol Jemison. Instead of a screen, she projects the night sky, and all interaction is by voice.
In order to introduce Mae and how she interacts with users, I illustrated and Judilee animated a short video.
—Common constellations, like the Dippers, Ursa Major & Minor, Orion, etc. —Star facts, like make-up, distance, age, and how they twinkle. —The moon: its phases, make-up, and myths —The planets of our solar system —Space travel
Mae has an uncomplicated, planetarium-inspired design to be both friendly and unobstrusive. In addition to her voice, she engages through simple facial expressions and her ability to project what she is talking about. She can also project the sky without conversation as a “night light.” She uses a bluetooth connection to an adult’s device for updates.
One of the pleasures of a project like Mae is the ability to do brand strategy to communicate our tone and values. We interpreted the colors of space and the stars in rich, bold, warmed-up tones. Comfortaa for type feels both kid-friendly and a touch futuristic, and we used it in our logo as well, modifying the “m” and the “e” and bringing it into space with a simple moon. Using gradients and texture in our illustrations, we pursue a tone that is friendly, youthful, and smart: Bring the night sky in.
Mae has an eportal for parents, allowing them to see what their child has been learning and how much their child has been using Mae. Parents can feel safe knowing their children are not exposed to internet unsupervised. Mae will only teach them night-sky related information. We use a closed system, and Mae connects through bluetooth to the parent’s device for updates or additional learning packs rather than connecting directly to wifi.
Mae has partnered with NASA to create learning packs that parents can add to expand possibilities for engagement, such as: “Our Nearest Neighbors,” “The Brightest Light,” “Journey to Mars,” and “The Sky is falling.”
We designed with principles of conversational design that we learned from Cathy Pearl, consulting voice designers, and from our research on kids and voice. Particularly important were:
—limiting choices and cognitive overload —warning about the length of stories since they are much beyond the “one breath rule” —responding to errors or incomplete requests by using whatever she’s heard to move the interaction forward: kids especially will repeat the same request over and over and over.
We were ready to invest more time in building out how users could interact with Mae, and we decided to do Wizard of Oz style testing. We scripted happy paths, growing our understanding of what Mae could talk about and how she handles scenarios. We discovered a small program called SayWizard that allowed us to create a 36 line script for Mae and then test her in a limited way with users.
We created a mock Mae and a mock projection to help users interact with her. We connected my computer with a bluetooth speaker so that Mae’s voice came from her body rather than my computer. Then we brought users in and filmed our tests.
On the mock projection were shown and labelled the big dipper, the little dipper, and the North Star. We instructed users that for the purpose of the test they could only ask Mae about what was on the “projection” in how to wake Mae, and that their goal was to learn a couple facts about the stars that they could see.
We iterated Mae in several rounds as we learned from the tests. We learned that she needed to be more warm and conversational, and we learned to anticipate users’ questions better so that Mae was ready to reply.
Through testing we quickly discovered that:
1 Mae needed to sound more friendly and conversational.
2 Mae needed to guide the user into an interaction by prompting them with an option immediately when she wakes up.
3 We needed to A/B test for how much she should prompt a user vs. how passive she should be.
4 We needed to A/B test for pause length: how long should she wait for the user?
How passive should Mae be? How does she teach a user how to interact with her? Does Mae prompt a user after she shares a fact or story—does she wait quietly or ask the user if they’d like to hear more right away? We want the experience to feel open-ended and low-pressure, while still teaching users how to make the most of their interaction with Mae.
This is an example of Mae teaching a user what she does upon initial wake.
Is this longer in children than in adults? This proved our most fertile ground for testing, and it is different in a mult-turn dialogue, when she first wakes up, and after she’s shared a fact or story.We’d read that 1.5 seconds feels like a good delay when the user is expected a response from Mae, but what about after Mae had replied, and is waiting passively to see if the user will interact with her further?
We decided that Mae should prompt a user after 4 seconds of being woke, if they hadn’t made any requests, with “Would you like to see the night sky?” or similar, and that she would prompt a new user after each fact with “Do you want to hear more?” or variant, and an experienced user only some of the time.
It’s important to have Mae reiterate whatever slot or word she has recognized in order to move the user forward and not get stuck in a loop—we know children often repeat the exact same wording of their request for long periods when they are not understood.
We determined that she should give the user options if she can hear the intent, following the “one breath” rule of thumb, after waiting 4 seconds, in case they correct themselves.
We have begun working on a prototype of Mae in the Amazon Developer Console, ideally to be published as an Alexa Skill so that Mae can interact with people in the real world! We wanted to show proof-of-concept and also to learn more about the business of voice design, and so after reading Cathy Pearl's Voice UI book, Amazon's Skill-making tutorials, and a whole bunch of internet articles, we worked from our happy paths to break down our intents, slots, and utterances, and to get them into the console.