Published by the IEEE CS n 1536-1268/10/$26.00 © 2010 IEEE PERVASIVE computing 21

C O N N E C T E D Y O U T H

T he critical nature of the obesity epidemic we face today has been widely documented, from ill health effects on the individual to the strain on the healthcare sys-

tem and society in general. Recently, the preva-lence of childhood obesity has increased mark-edly, along with the incidence of obesity-related diseases such as type-2 diabetes, cardiovascu-lar disease, and certain cancers. Fortunately, awareness of these risks has also increased, and various countries’ agendas aim to address the problem in today’s youth. If we modify the behaviors that lead to obesity early enough, we

can keep this problem from getting worse.

According to Nielsen re-search (http://blog.nielsen.com /nielsenwire /repor ts /nielsen_howteensusemedia_june09.pdf), adolescents and

teens daily on average spend more than three hours watching television, an hour on the com-puter, and half an hour playing video games, and send almost 100 text messages. So, we decided to tackle health-related problems by reaching kids through the media with which they’re the most intimate. We designed Time to Eat, a mo-bile-phone-based virtual-pet game intended to improve adolescents’ eating behavior. The game is firmly grounded in research in three popular areas: persuasive technology, mobile comput-ing, and health games. We conducted a study

to identify how a mobile phone game’s motiva-tional features support and encourage healthy eating habits in seventh and eighth graders.

Persuasive Technology and Mobile PhonesIn 2002, B.J. Fogg introduced the research community to the computer’s power to moti-vate and deter behaviors and study behavior change mechanisms.1 This research, a culmina-tion of psychology, persuasion, and computing, also provided a conceptual framework for re-searchers to further explore possibilities in this realm. Age-old psychology theories, including intrinsic and extrinsic motivation, social facili-tation, social comparison, and social cognitive theory, provided a firm grounding for what was then cutting-edge experimentation and practi-cal application of using computers to change behaviors.

Today, much of the brightest research in per-suasive technology centers on the most impor-tant technology of the last decade—the mobile phone. Mobile phones’ advancement and pro-liferation have been staggering; most people now carry a mobile phone, always switched on. More than 70 percent of teens own a phone by the time they’re 17; this number will only continue to grow (http://pewresearch.org/pubs/1315/teens-use-of-cell-phones). The mo-bile phone plays an important role in adoles-cents’ lives, providing an always-connected link to their social network and identity.

To motivate children to practice healthy eating habits, a mobile-phone-based game lets them care for a virtual pet by sending it photos of the food they consume.

J.P. Pollak, Geri Gay, Sahara Byrne, Emily Wagner, Daniela Retelny, and Lee HumphreysCornell University

It’s Time to Eat! Using Mobile Games to Promote Healthy Eating

22 PERVASIVE computing www.computer.org/pervasive

CONNECTED YOUTH

Furthermore, according to Juniper Research (www.juniperresearch.com/shop/viewwhitepaper.php?id=68& whitepaper=30), more than half of teens with mobile phones use con-nected services such as text messag-ing, cameras, multimedia messaging systems, games, and other apps.

Today, the average person can take notes, check the time, look up his or her schedule, fetch information from the Internet, listen to music, play games, and take pictures, all with a mobile phone. Many phones can even indicate the user’s precise location,

thanks to GPS. Mobile phones are our trusted companions, confidants, guides, assistants, and—potentially, through persuasive-technology appli-cations—coaches and mentors. Be-cause of these qualities, mobile phones are the perfect instrument for deliver-ing health-related messages at oppor-tune times.

Fogg describes this notion as the “kairos factor.” The myriad infor-mation that a mobile phone pos-sesses about its owner lets persuasive- technology designers deliver messages tailored to a user’s current location, cal-endar, habits, and preferences. It also leads to a very personal relationship between a phone and its owner; Fogg has even used the term “marriage” to describe a person’s relationship with his or her mobile phone.2 Because the mobile phone is always connected and, in many cases, has constant access to the Internet, users have confidence that the quality and relevancy of the infor-mation in phone messages are high. So,

users also deem the information trust-worthy and supportive.

Health GamesAlthough the mobile phone might be the perfect instrument for delivering persuasive messages, individuals must be adequately motivated before they change their behavior. This is where games come into play. The global re-tail market for mobile-phone-based games was US$5.4 billion in 2008 and should rise to $10 billion in 2013, ac-cording to Juniper Research (www.juniperresearch.com/shop/viewreport.

php?id=161). Clearly, the world is play-ing a lot of games on mobile phones. Although games have been implicated as a root cause for everything from the aforementioned obesity epidemic to in-creased youth violence, recent health-gaming research suggests otherwise. The Nintendo Wii showed the public that games could positively influence health. Over the last decade, quality re-search conducted at universities around the world has demonstrated that games can lead to more exercise, improved diet, and better management of chronic diseases such as diabetes.

One health games research genre fo-cuses on virtual-pet care. Virtual-pet games have made for widely successful retail products, including the Tama-gotchi and Nintendogs, but research also supports the notion of virtual-pet care as a means to change behavior. For starters, virtual pets are perfect companions for kids: failing to care for them doesn’t have real-world con-sequences, and they can offer more fre-

quent interaction.3 Furthermore, they introduce an element of social facili-tation. Research suggests that people will perform simple learned behaviors more readily when they feel they’re be-ing observed.4

Recent studies have demonstrated virtual-pet care’s potential health ben-efits: Ubifit Garden encourages players to exercise more to make a garden grow and flowers bloom.5 In Fish’n’Steps, players raise fish in an aquarium on the basis of the number of steps they take per day.6 In the StepGreen project, car-ing for a virtual polar bear led to more environmentally conscious behaviors.7

Time to EatWe decided early in the design process to mimic commercially successful vir-tual pets such as the handheld Ninten-dogs and Tamagotchi games, as well as online pets such as Webkinz and Neopets. These games focus on pet care as part of a child’s daily routine, which fits well with behavior change models in which repeated reinforce-ment is key to success. We worked with artists, graphic designers, and develop-mental psychologists to design pets and interfaces appropriate for seventh and eighth graders. We developed Time to Eat for the iPhone, which we selected for its large and outstanding touch screen, ease of use, and “cool factor” with kids. We extensively tested proto-types of the game with the target demo-graphic before deployment.

At the game’s start, players adopt a worm, dinosaur, dog, hippo, penguin, potato head, robot, stapler, or tree (see Figure 1). The choices are intended to capture a range of interests and were refined during play testing. Players then name their pets and provide login infor-mation. We let players pick and name their pet to provide a sense of control— one of the tenets of Mark Lepper and Thomas Malone’s classic work8—as well as a sense of attachment, which has proven important in virtual-pet-care games.7 A Time to Eat icon appears on the phone’s home screen and links the

Although the mobile phone might be the

perfect instrument for delivering persuasive

messages, individuals must be adequately

motivated before they change their behavior.

This is where games come into play.

JULY–SEPTEMBER 2010 PERVASIVE computing 23

user directly to the login screen, and players receive an e-mail address to let them communicate with their pet. Figure 2 depicts Time to Eat’s primary user interface.

Once the game is under way, pets send users healthy-eating reminders via e-mail, such as “Good morning! Re-member to eat breakfast today” or “A healthy breakfast is important.” The messages and the times they’re sent are configurable. In our study, we set up the game to send reminders before and after school, focusing on breakfast and snack time, respectively. All the e-mails sent to the players were written as if from the pets themselves. The reminders’ content changed depending on the day of the week but was always followed by “Have a glass of water and a healthy snack such as fruit or cheese and send us the picture! Log in to see how I’m doing!” or the equivalent for breakfast.

We could have conducted this interac-tion with text messages, a more popu-lar medium for communication among users in our target demographic. How-ever, during our study we had to block text messaging on the students’ phones to curb excessive costs. Before we blocked text messages, one student sent more than 800 messages in two days!

The game prompts each participant to take a picture of his or her meal or snack (see Figure 3) and submit it. The user then receives a score ranging from −2 to 2, based on two factors: whether he or she ate anything at all and the healthiness of the food in the photo. Nutritionists trained our research team on healthy breakfast eating guidelines so that they could examine each photo and assign a score based on a simple rubric. Users who submit no breakfast picture receive −2. The feedback asso-

ciated with this score is often, “Did you miss breakfast? I am hungry.” Other scores are accompanied by responses specific to the meal the child submit-ted, designed to provide some informa-tion about which foods are and aren’t healthy. For example, “Yesterday’s breakfast wasn’t great, and neither is two doughnuts. Can you try something healthier tomorrow, like oatmeal with fruit?” A score of 2 could be accom-panied by, “Yogurt with strawberries makes me happy! I had plenty of en-ergy until lunch.”

A meal’s score also determines a pet’s

emotional state when its player logs in. Every pet begins in a neutral state. Af-ter the user sends a photo, the game e-mails feedback corresponding to the meal’s score, and the pet’s appearance changes on the basis of that score. Each score from −2 to 2 has a corresponding state for each pet; negative scores pro-duce a sad pet, whereas positive scores produce a happy one (see Figure 4).

Changing Behavior with a Virtual PetOur first version of Time to Eat cen-tered on encouraging kids to eat a

(b)(a)

www.cornellhci.org/tim...

Today Sweetie is..

AT&TAT&T 12:08 PM12:08 PMTime To EatTime To Eat

AT&TAT&T 12:08 PM12:08 PMTime To EatTime To Eat

Google www.cornellhci.org/tim... Google

Awesome!

See What I’ve Done

Logout

Most Recent Photo:

Great Breakfast! Cereal with milk and a piece of fruitis the perfect way to start your day. Keep up the good

work!Photos you submitted:

View More History

Back to Sweetie

Logout

Figure 1. The adoptable pets in Time to Eat. Players choose to adopt a pet from a selection ranging from dogs to robots to staplers. We developed these pets by working closely with designers and appropriately aged potential users.

Figure 2. The Time to Eat user interface: (a) the home screen, depicting the pet’s current emotional state and (b) the feedback screen with the corresponding food photo. The pet’s emotional state reflects the quality of meals the player has recently eaten and submitted. Navigating to the feedback screen lets players see recent photos and their corresponding feedback and hopefully make a connection between their actions and their pet’s state.

24 PERVASIVE computing www.computer.org/pervasive

CONNECTED YOUTH

healthy breakfast. For initial studies, we selected a single healthy eating be-

havior to try to manage complexity and focus on the outcomes of design

changes. We chose eating a healthy breakfast for two main reasons. First, because it normally happens at home, subjects can more easily control it. Second, and perhaps more impor-tant, skipping breakfast is becom-ing common in the US9 and has been linked to numerous health-compro-mising behaviors and lifestyles among adolescents.10

In a recent study, we gave iPhones loaded with Time to Eat to 53 seventh and eighth graders at a rural mid-dle school in upstate New York for one month.11 In some respects, this study’s novelty is that it took place at all with middle school students. Such a study involves obvious difficul-ties, such as getting school district, teacher, and parent permissions and dealing with problem kids. Further-more, the cost and logistics of pro-viding iPhones to many teenagers are somewhat prohibitive. (Amazing side note: only two phones were lost over the month.) Finally, configuring more than 50 iPhones, including setting up mail, games, and parental controls, is a daunting task not to be overlooked when you’re planning a study.

This study’s findings were conclu-sive: kids who played Time to Eat ate a healthy breakfast more frequently than those who didn’t. Children playing the game ate a healthy breakfast 52 percent of the time; kids who didn’t play it ate a healthy breakfast approximately only 20 percent of the time. We also found that kids needed to receive both nega-tive and positive feedback from their pet, suggesting that emotional and so-cial realism were key to their experience and the game’s motivational abilities.11 Interestingly, we found no significant gender effects.

What is it about Time to Eat that mo-tivates kids to change their behavior? In interviews, the theme of eating healthy breakfasts for the pet emerged:

Student: I just didn’t take pictures of breakfast because I don’t eat breakfast.

Figure 3. Various breakfast photos submitted by kids playing Time to Eat. This collage represents a small sample of the photos we received from players during the study. Kids submitted photos of meals ranging from unhealthy choices, such as doughnuts, to well-balanced choices, such as cereal, fruit, and milk.

Figure 4. The dog and tree pets in various emotional states ranging from happiest (left) to saddest (right). The pets’ emotional state changes from sad to happy or vice versa based on the healthiness of meals in recently submitted photos. The depictions of emotional states were piloted with potential players to ensure recognition of the different states.

JULY–SEPTEMBER 2010 PERVASIVE computing 25

Interviewer: You don’t eat breakfast. Do you feel like [the pet game] changed the way that you eat at all?

Student: A little.

Interviewer: A little?

Student: Yeah.

Interviewer: What kind of change did you make?

Student: For the pet, I ate breakfast.

Furthermore, the fact that the pet “lives” in the kids’ phones is important. As we mentioned earlier, our mobile phones are always with us and switched on; some kids have even reported sleep-ing with theirs (http://pewsocialtrends.org/assets/pdf/millennials-confident-connected-open-to-change.pdf). In some cases, the students’ attachment to their phone encouraged their attach-ment to the pet:

Interviewer: Do you feel attached to the pet?

Student: Well, not attached, but I’m attached to the iPhone.

Interviewer: And you like the pet because of the iPhone; they go well together?

Student: I’m going to be sad when I have to give it back.

Because phones could be constantly with the students, the pet could be with them at all times—a constant companion, always available to send and receive messages and provide feed-back on recent meals. Placing the vir-tual pet on the phone might be the only way to accomplish this feat—certainly real pets, toys, or computer-based vir-tual pets wouldn’t be with the kids with such great frequency.

Also, the phone the pet lived in is obviously inherently a communica-tion device, which might have led to more natural interactions with the pet through other modes such as e-mail. We found much evidence of kids e-mailing their pet directly, calling it by name, and interacting with it as a social being. After receiving positive feedback about breakfast, one subject e-mailed the pet well-wishes: “thank u skittles, happy monday!” Similar evi-dence emerged in the interview data:

Interviewer: What do you like about [the pet game]?

Student: It interacts with me. It does what I do.

Such behavior also indicates a certain element of realism or at least the play-ers’ suspension of disbelief.

Will these children’s behavior changes continue beyond the study? To determine the answer, we’ve planned a six-month follow-up study.

Looking ForwardDespite this success, we can do more with the game and the mobile platform to improve the kids’ experience as well as the motivational element. We can im-prove the players’ experience with Time

to Eat in several ways, some of which are in the current version but were un-available at the time of the study.

Ultimately, games should be fun, and in a virtual-pet game, children should look forward to taking care of and in-teracting with their pet. Obviously, children will eventually become bored with this game unless we add features

to engage and excite them beyond the game’s initial appeal. Adding certain features will lead to a more personal-ized, interactive, and social game. Per-sonalizing the pet and the game would let children customize the system to best fit their lifestyle, eating habits, and interests. Social features such as a chat function, competitions between peers, and the ability to view and comment on another’s pet will let kids interact with their peers, while increasing the game’s effectiveness.

Rather than feeling like eating health-fully is a daily job or chore, the satis-faction of raising a happy pet should motivate children to want to maintain a healthy diet. This emphasizes the im-portance of building a strong relation-ship between children and their pet. By incorporating features similar to other games children are used to playing, such as reaching different levels based on how the pet is doing, the game could increase the pet’s interaction capabili-ties. For example, in higher levels, us-ers could have the option to play fetch or other games with their pet, walk it, or bathe it. Similarly, users could un-lock certain items once the pet reaches a certain state. For example, after eating three healthy balanced meals, a child could unlock a new outfit for the pet or choose a new toy for it. By providing in-

cremental rewards, the game will pro-mote a stronger relationship between children and their pet, thereby increas-ing their desire to eat healthily to keep the pet healthy.

We must address several scalability issues if we hope to conduct larger stud-ies and ultimately make the game avail-able to the public. In the game’s current

By providing incremental rewards,

the game will promote a stronger relationship

between children and their pet,

thereby increasing their desire to eat healthily.

26 PERVASIVE computing www.computer.org/pervasive

CONNECTED YOUTH

version, the pet’s state doesn’t change until a human coder uploads all the kids’ images and feedback to the server. This causes a lag between the time the child eats the food and the time the pet

actually reacts to the food, and clearly is a strain on resources.

We’d like to improve the system’s ef-ficiency and scalability by letting peers rate each other’s photos or implement-

ing machine vision and image recogni-tion tools to provide immediate feed-back. By letting peers rate photos, the game would send users almost instan-taneous feedback. Although peers ob-viously lack the expertise of trained nu-tritionists, the game would need fewer nutritionists, and even the photo rating could become a game, harnessing the wisdom of the crowd.12 Eventually, machine vision and image recognition tools could automatically rate photos and change the pet’s status. If based on an appropriate rubric, such as avoid-ance of sugary foods, inclusion of fruit, and so on, this should be quite feasible. However, even the state of the art in this area can’t accurately interpret different foods and portion sizes. Although both these mechanisms will eventually im-prove the system’s scalability, they have weaknesses we must address before we can include them in the game.

Time to Eat can also act as a power-ful data collection tool. Much research has demonstrated the importance of test subjects keeping diaries when re-searchers track routine behaviors such as eating. Our inability to accurately re-call and report on what and how much we eat makes it difficult for nutrition-ists and care providers to assess how well we’re adhering to dietary guide-lines. For diligent Time to Eat players, however, a record exists of not only the food they consumed but also when they ate it and even where they were. This in itself is valuable, and combining photo diary data with geotagging—or placing photos onto a map—might reveal inter-esting patterns about eating habits and problem areas. A system such as this could make it easy to identify, for ex-ample, that teens are consuming large quantities of soda and candy on days they go to the mall after school. The system could also send prompts, warn-ings, and alternate-food recommenda-tions on the basis of a player’s location.

Whereas Time to Eat emphasizes eating behavior, we intend to study the use of these game mechanisms to im-prove behavior in other areas, includ-

the AUTHORSJ.P. Pollak is a PhD student in information science at Cornell University, with a visiting appointment at Weill Cornell Medical College. His research focuses on creating social mobile phone applications that encourage healthy behavior. Pollak has an MS in human-computer interaction from Cornell. Contact him at jpp9@cornell.edu.

Geri Gay is the Kenneth J. Bissett Professor and chair of communication at Cor-nell University and a Stephen H. Weiss Presidential Fellow. Her research focuses on social and technical issues in the design of interactive communication tech-nologies, specifically social navigation, affective computing, social networking, mobile computing, and design theory. Gay has a PhD in educational psychol-ogy from Cornell University. Contact her at gkg1@cornell.edu.

Sahara Byrne is an assistant professor of communication at Cornell University. Her research interests include media effects, strategic communication, and cognitive development. Byrne has a PhD in communication from the Univer-sity of California at Santa Barbara. Contact her at seb272@cornell.edu.

Emily Wagner is a senior in Cornell University’s College of Engineering. Her research interests include mobile technologies and health gaming. She’s a member of the Information Science Student Association and the Society of Women Engineers. Contact her at eew27@cornell.edu.

Daniela Retelny is working toward a BS in information science at Cornell University, with concentrations in human-computer interaction and social sys-tems. Her research uses mobile phone applications to investigate the mecha-nisms of mobile and social persuasion to improve health behaviors. She’s on the executive board of the Information Science Student Association. Contact her at dr326@cornell.edu.

Lee Humphreys is an assistant professor of communication at Cornell Univer-sity. Her research examines mobile technology’s role in urban environments. Humphreys has a PhD in communication from the University of Pennsylvania. Contact her at lmh13@cornell.edu.

JULY–SEPTEMBER 2010 PERVASIVE computing 27

ing smoking cessation or prevention, addiction, and management of chronic diseases such as diabetes. However, these tools shouldn’t be limited to health behaviors. We also envision en-couraging children to complete home-work assignments or become more socially responsible by improving en-ergy conservation and recycling habits. The future is bright for this and other mobile-phone-based health games. We can use mobile phones and custom-designed mobile phone games to pro-vide different types of support for indi-viduals and peer groups. People tend to stick with and learn new healthy habits if they know they’re being monitored; through good design, we can capitalize on mobile tools to encourage healthy behaviors.

ACKNOWLEDGMENTSThe Robert Wood Johnson Foundation’s Pioneer Program funded this research.

REFERENCES 1. B.J. Fogg, Persuasive Technology: Using

Computers to Change What We Think and Do, Morgan Kaufmann, 2002.

2. B.J. Fogg and D. Eckles, Mobile Persua-sion: 20 Perspectives on the Future of Behavior Change, Stanford Captology Media, 2007.

3. L.R. Bloch and D. Lemish, “Disposable Love: The Rise and Fall of a Virtual Pet,” New Media & Society, vol. 1, no. 3, 1999, pp. 283–303.

4. R.B. Zajonc, “Social Facilitation,” Sci-ence, vol. 149, 1965, pp. 269–274.

5. S. Consolvo, D.W. McDonald, and J.A. Landay, “Theory-Driven Design Strate-gies for Technologies That Support Behav-ior Change in Everyday Life,” Proc. 27th Int’l Conf. Human Factors in Computing Systems (CHI 09), ACM Press, 2009, pp. 405–414.

6. J.J. Lin et al., “Fish’n’Steps: Encourag-ing Physical Activity with an Interactive Computer Game,” Ubicomp 2006: Ubiq-uitous Computing, LNCS 4206, Springer, 2006, pp. 261–278.

7. T. Dillahunt et al., “Motivating Environ-mentally Sustainable Behavior Changes with a Virtual Polar Bear,” Perva-sive 2008 Workshop Proc., 2008, pp. 58–62; www.pervasive2008.org/Papers/Workshop/w2-04.pdf.

8. M.R. Lepper and T.W. Malone, “Intrin-sic Motivation and Instructional Effec-tiveness in Computer-Based Education,” Aptitude, Learning and Instruction: Cog-nitive and Affective Process Analyses, vol.

3, R.E. Snow and M.J. Farr, eds., Rout-ledge, 1987, pp. 255–286.

9. A.M. Siega-Riz, B.M. Popkin, and T. Car-son, “Trends in Breakfast Consumption for Children in the United States from 1965–1991,” Am. J. Clinical Nutrition, vol. 67, 1998, pp. 748S–756S.

10. A. Keski-Rahkonen et al., “Breakfast Skipping and Health-Compromising Behaviors in Adolescents and Adults,” European J. Clinical Nutrition, vol. 57, 2003, pp. 842–853.

11. S. Byrne et al., “‘When I Eat So Bad, My Pet Looks So Sad’: An Initial Test of the Persuasive Effect of Feedback from a Mobile Phone Virtual Pet on Adolescent Breakfast Intake,” to be published in Proc. 2010 Ann. Conf. Int’l Communi-cation Assoc. (ICA 10), 2010.

12. L. Von Ahn and L. Dabbish, “Labeling Images with a Computer Game,” Proc. Int’l Conf. Human Factors in Comput-ing Systems (CHI 04), ACM Press, 2004, pp. 319–326.

Selected CS articles and columns are also available for free at http://ComputingNow.computer.org.

Learn about computing history and the people who shaped it.

COMPUTING THEN

http://computingnow.computer.org/ct