MACHINE LEARNINGGAME DESIGNVC FUNDING

LittleMoochi

An AI-based virtual pet that uses embedded design to teach children about healthy eating.

Date December 2018 - March 2019
Context
McGinnis Venture Competition, Swartz Center for Entrepreneurship at Carnegie Mellon University
Team
Summer Xia, CEO & Founder; Yi Xu, Backend Engineer; Zhuyun Dai, Backend Engineer
My Role
Lead Product Designer
INTERACT WITH INVISION PROTOTYPE

PROBLEM SPACE

Obesity has reached epidemic proportions in the United States.

71% of adults and 32% of children in the U.S. are overweight or obese. Eating habits formed in childhood are critical predictors of health outcomes later in life. However, many children are picky eaters, and prefer junk food over healthier fare.

How might we gamify the adoption of healthy eating habits?

SOLUTION

A virtual pet that leverages persuasive design to subtly encourage behavior change.

LittleMoochi is an AI-based virtual pet that encourages children to adopt healthy eating habits. Children care for a Moochi pet by feeding it their own food via AI-powered image recognition software. LittleMoochi earns "superpowers" from nutrient-dense foods, encouraging repetition and habit formation.

Persuasive design techniques such as obfuscation and intermixing distance children from Moochi's underlying concept and conceal its underlying healthy eating message, enhancing playfulness. A machine learning algorithm learns childrens' preferences over time to encourage them to expand their palette and try new foods.

Play with the Invision prototype:

KEY FEATURE 1

Metaphor & fictionalization enhance playfulness.

LittleMoochi is a virtual representation of the player him/herself. Fictionalization and metaphor provide a safe level of psychological distance to explore the concept of healthy eating, which can be a difficult topic for kids to broach.

Fictional narratives create distance from the self, allowing players to process uncomfortable or threatening topics more easily.

KEY FEATURE 2

Obfuscation masks Moochi's intent, making the game more engaging.

LittleMoochi's purpose is encouraging healthy eating. Explicitly broadcasting this goal could backfire, leading to less engagement and enjoyment. The game's educational intent is concealed by its superhero motif and the premise of caring for a virtual pet.

Obfuscation entails concealing or stealthily introducing the true purpose of a game to create a more engaging narrative.

KEY FEATURE 3

Intermixed content makes healthy eating less conspicuous.

Some superpowers (night vision, strength) are related to particular foods, whereas others (flying) are not. Interweaving "on-topic" content related to healthy eating with "off-topic" content makes the former less off-putting, and more enjoyable for players.

Intermixing is an embedded design technique in which relevant and irrelevant content is displayed to users to disguise purpose and make a topic more approachable.

GAMEPLAY

A system of loops and arcs creates short-and long-term goals for players.

The game's core loop is "leveling up" by feeding Moochi healthy food that attributes to his growth over time. The broader game arc is unlocking all of Moochi's superpowers by trying superfoods.

MOOCHI FOR PARENTS

Getting the whole family involved makes LittleMoochi more than just a game.

We created a parent portal to get parents involved in helping their children foster healthy eating habits.

KEY FEATURE 1

A high-level dashboard keeps parents in the loop.

Parents can see high-level reports of their children's eating habits and monitor them over time. The dashboard is centered around eating a well-balanced diet, rather than counting calories, to encourage flexibility and exploration.

Rationale: Parents expressed concern about a dashboard with too much granularity, so we purposefully kept it vague. A skeuomorphic plate with food group goals is easy to grasp for busy parents.

KEY FEATURE 2

An image archive is easily accessible for doctors.

Expert interviews taught us that parents have misconceptions about what is healthy. An image archive feature helps parents easily show healthcare professionals exactly what their children are consuming at a checkup.

Rationale: Parents are too busy to manually monitor everything their child eats. An image archive is an easy reference point when interfacing with doctors.

KEY FEATURE 3

Articles, quizzes, and videos help parents learn, too.

Parents are too busy to scour the internet for resources on healthy eating. A media tab aggregates relevant information for parents to consume.

Rationale: Parents have misconceptions about what is healthy, and stand to learn from LittleMoochi, too!

WHAT I DID

What I contributed to the project

As the Lead Product Designer, I worked closely with the PM and Engineers towards a shared vision. Here are some of my most notable contributions:

  • Defining research goals and drafting interview protocols
  • Developing a survey and collecting responses from 150+ parents of young children
  • Encouraging the team to pivot after analyzing research data and synthesizing takeaways
  • Working with engineers to understand technical constraints pertaining to the AI image recognition backend
  • Defining the look and feel of the app; creating vector illustrations implemented throughout
  • Facilitating user tests with 4 children and 2 with parents after an MVP Unity version of LittleMoochi had been built
  • Teaching the team user-centric design and research fundamentals that they could use moving forward

GENERATIVE RESEARCH

Data from 150+ individuals taught us that 72% of parents experience conflict at the dinner table.

Through interviews & surveys, we found that the majority of parents experience arguments with their children over food choices. Furthermore, over 68% of parents view their childrens' eating behaviors as "mediocre" or less.

When children exhibit picky eating behaviors, half of parents choose to compromise or give up.

Picky eating is a frustrating behavior that parents create work-arounds to avoid. Whereas 35% of parents choose to compromise, 15% give up entirely, giving in to their childs' demands.

→ INSIGHT #1

Parents are busy & overwhelmed, & don't always have the energy to argue with their children over what they should eat.

The overarching theme of our generative research was that, despite good intentions, parents are overwhelmed, and lack the time, resources, vigilance, and patience to consistently feed their children healthy food.

→ INSIGHT #2

Parents have misconceptions about what is healthy.

We learned that, although all parents acknowledged the importance of good nutrition for their children, many harbor misconceptions about healthy eating. For example, several interviewees remarked that gummy bears contained fruit, and that French fries were a vegetable.the development of healthy eating habits.

COMPETITIVE ANALYSIS

The existing meal tracking apps are boring & arduous.

A competitive analysis taught us that the majority of health apps focus on calorie counting and professional fitness knowledge, and are not designed for children. Two children-oriented weight management apps—Kurbo and Habitz—are simply tracking apps, and are boring and tedious.

ITERATIONS

Inspired by the Tamagotchi craze of the '90's, LittleMoochi was born.

Following user research, we landed on LittleMoochi— a gamified virtual pet that users "feed" their own food. Leveraging the ML backgrounds of our engineers, we envisioned that Moochi could react to food choices to create teaching moments in-app.

The character evolved over rounds of user testing.

My first iteration of the Moochi character didn't have a body. As time went on, and we envisioned monetization via purchasing clothes in-app or selling a physical Moochi doll, the character evolved and became cuter.

Our screens became more tailored to children over time.

The first few iterations of our screens were difficult to navigate for children. We incorporated larger, more prominent buttons in later iterations to make Moochi's capabilities clear.

USER TESTING

We tested med-fi wireframes with 4 kids and 6 parents.

User testing helped us iterate towards a more refined product and understand pain points in the existing user flow.

→ INSIGHT #1

Vitamins & minerals are too heavy of concepts for kids— make it more visual & less technical.

Our initial design focused on the vitamin & mineral benefits of consuming certain foods. After testing, we restructured this to how foods benefit one's body, rather than what vitamins can be derived from a food.

→ INSIGHT #2

Incorporate an in-app reward to keep children engaged.

Parents expressed that a simple reward, like stickers or badges, would motivate children to return to the app, and teach them concepts like delayed gratification & saving for later.

→ INSIGHT #3

Children are more likely to have iPads than iPhones.

Most importantly, we heard that children between the ages of 5-10 are much more likely to have tablets than phones. Parents prefer tablets as they are larger surfaces for games & movies. Thus, we switched our final design over to a tablet form.

→ INSIGHT #4

Parents don't want a dashboard to see what their children are eating...but it is a source of monetization.

Interestingly, parents expressed not wanting a dashboard to monitor their child's eating habits, as it could make them obsessive and anxious. Our team ultimately decided to keep a parent dashboard component, however, as law prohibits advertisements from being on children's apps.

DESIGN DECISIONS

1. Focus on body benefits rather than calories.

Action: We remained aware of LittleMoochi's potential to jump start hazardous or harmful eating behaviors. To avoid triggering an unhealthy relationship with food, we framed eating around bodily benefits rather than calories.

Rationale: Clinical research indicates that the foundational elements of eating disorders may trace back to childhood. We took all possible steps to avoid maladaptive eating behaviors.

2. Positive reinforcement to foster a healthy relationship with food.

Action: We chose to exclude any negative messages or criticism from LittleMoochi— only giving positive reinforcement to users. If children eat something unhealthy, Moochi simply won't gain as much energy or health as he would otherwise.

Rationale: We want our users to foster a healthy relationship with food. Instead of labeling food as "good" or "bad", we frame it in terms of energy and growth.

3. Superhero reactions after milestones.

Action: Superfoods grant Moochi superpowers. For example, after eating foods high in Vitamin A such as carrots or tomatos, Moochi gains night vision. Similarly, Moochi will appear strong & lift weights if the user eats something with lots of protein.

Rationale: Testing indicated that children respond most positively to very visual design choices. Parents remarked that children relate to body benefit conceptualizations of food most easily.

MARKET OPPORTUNITY

LittleMoochi tackles an $11.4B sector of our economy.

LittleMoochi is tackling at least a $3.9 billion market, appealing to 29 million children between the ages of 5-10. Furthermore, the U.S. educational technology & weight loss markets are two of the fastest-growing segments of our economy, expected to reach $73 billion by 2025.

We calculated $6.9M projected revenue over 3 years.

We decided to monetize LittleMoochi in the following ways:

1) In-game purchases— to customize Moochi's appearance & decorate its space (WIP)
2) Customizable content— allowing parents to create custom goals for children
3) [Eventually] a Moochi doll— for children to connect to LittleMoochi outside of the app

Through these means as well as social media advertising, we calculated $6.9M in projected revenue by 2022.

BACKEND DEVELOPMENT

Our backend developers created a machine learning algorithm and trained it on 500+ recipes.

Our two backend developers built a deep, multi-modal neural network that is capable of recognizing an ingredient and its corresponding nutritional elements through an image. This was achieved by training a ML network on thousands of recipes from around the world and pairing it with a deep NLP network to capture the relationships between images of food and ingredients.

VENTURE COMPETITION

We won $25,000 in the McGinnis Venture and Forte Power Pitch Competitions!

We pitched LittleMoochi to a panel of expert judges in Carnegie Mellon's McGinnis Venture Competition. We were honored to win a $15k investment after months of hard work!
LittleMoochi also won first place and $10k in the Forte Power Pitch competition in June 2019.

REFLECTION

Pivots are unavoidable.

We pivoted several times throughout this project— first, we designed LittleMoochi for adults; later, we switched from a mobile to tablet version of the app. Ultimately, our pivots made our story much stronger, showing the breadth and depth of research we'd conducted to support our decisions.

Don't be self-referential.

When I joined the LittleMoochi project, our CEO had based our initial target audience of adults off of her own experience. It was later that we discovered that the children's health app market was a much more appropriate place for our idea. It's essential to remember that you are not the user throughout the design process.

< PREVIOUS PROJECT
NEXT PROJECT >
OVERVIEWPROBLEM SPACESOLUTIONWHAT I DIDGENERATIVE RESEARCHCOMPETITIVE ANALYSISITERATIONSUSER TESTINGDESIGN DECISIONSMARKET OPPORTUNITYBACKEND DEVELOPMENTVENTURE COMPETITIONREFLECTION

"Let the beauty of what you love be what you do." –Rumi

© Molly Vierhile 2019