3. Causes of Morbidity, US 2006
Heart Disease
Cancer
Diabetes
Stroke
Chronic lower
68% respiratory diseases
Accidents
Lifestyle diseases: Alzheimer's
Diseases Influenza and
caused, prevented or pneumonia
ameliorated by lifestyle
choices.
3
4. Self-monitoring predicts lifestyle
change.
● Self-monitoring: the process of
observing and recording a target
behavior
– Single most-effective behavioral change
tool
– Sustaining self-monitoring is hard
– Correlation between stopping self-
monitoring and reducing target
behavior
4
5. Contributions
● BALANCE: Examined and characterized
problems with nutrient-based food diaries
on mobile phones.
● Food Indexes for Dietary Self-Monitoring:
Examined, characterized and evaluated
goal-based, diet quality-oriented food
diaries.
● Design of POND: Designed and developed
pattern-oriented nutrition diary.
● Evaluation of POND: Insights based on
real-world evaluation.
5
6. Examined and characterized problems with database-based food diaries
on mobile phones.
Contribution 1: BALANCE
6
7. BALANCE Glanceable, real-time
energy
intake/expenditure
balance to support
timely decision
making
7
16. Evaluations
● Iterative, participatory design:
– 5 focus groups (3-5 people/group)
– Used Food Diary for 3 days
● Validation:
– 34 participants
– Used phone with BALANCE
software and MSB for 3 days
● ~50 people provided feedback
16
17. Users felt conflicted.
● Food diary on mobile phone is good.
● Took too long to enter a meal.
17
18. Finding a food in the database
● Choose what to search for
– Broad query yields too many results
– Specific query yields no results
● Enter text on mobile device
● Wait for query
– Magnified by context
● Choose which result
– How to choose which one to pick
● Self-prepared foods such as salads:
– Small amounts of many different foods
18
19. Users felt conflicted.
● Food diary on mobile phone is good.
● Took too long to enter a meal.
– Fuel gauge visualization rarely depicted
current energy intake/expenditure
balance
● Participants weren’t willing to
continue
19
20. How can we self-monitor dietary
intake without looking up foods?
20
33. Sample Task
D1
Spaghetti, Al Dente, Cooked
1 cup(s)
Green Peas, Frozen, Boiled, Drained
½ cup(s)
SARGENTO FANCY Shredded Parmesan Cheese
¼ cup(s)
PEPPERIDGE FARM Crusty Italian Bread, Garlic
1 serving(s)
Green Salad
¾ cup(s)
HIDDEN VALLEY The Original Ranch Dressing
2 tablespoon(s)
33
34. Time & Correctness
Time in Condition Correctness
25 10
Correctness Score
20 8
Minutes
15 6
10 4
5 2
0 0
HEI FBQI BAL HEI FBQI BAL
34
35. HEI
7
TLX Responses FBQI
BALANCE
6
5
4
Rating
3
2
1
0
Mental Demand
Physical Exertion
Discouraged, Irritated
Successful Quickly Easy to Use Easy to Learn
35
36. Which one will help you reach your
goal? I'm mostly interested in how balanced my diet is
and the HEI display seemed to do that best.
7
I'm not as focused on calories, but if I was, I
6 would like the BALANCE display best.
Number of Participants
5
The third just seemed to imprecise to be a valid
4 representation.
3
2
1
0
Lose weight Eat "better" Control portion sizes Eat less/more
BALANCE HEI
36
37. Summary
● It’s possible to simplify too much
(FBQI)
● There could be value in Food Index-
based tracking
● Lab studies don’t reflect the real
world
37
38. What happens in the real world?
Designed and developed pattern-oriented nutrition diary.
Contribution 3: Design of POND
38
46. Insights based on real-world evaluation.
Contribution 4: Evaluation of
POND
46
47. POND Evaluation
● Pilot Evaluation
– 12 ppts
● In-Lab Evaluation
– 24 ppts
– 4 days worth of meals
● In-situ Evaluation
– 20 ppts (from in-lab evaluation)
– 3 wks of using POND on their phone
47
48. Insights
● How: What strategies did people use
to make entries? (in-lab)
● What: What kinds of entries did
people make with the different
strategies? (in situ)
● When: When did people find it easy
or hard to make entries? (in situ)
48
49. How: Strategy to complete tasks (in
lab)
20
18 13
16 people 9 14
Number of tasks
14 used people people
12 this used used
10 strateg this this
8 y strateg strateg
6 y y
4
2
0
ppt13 ppt17 ppt20
+1_only lookup_only mixed
49
50. How people entered tasks
indicates…
● Some people value the speed and
estimates (+1 only)
● Some people value flexibility (mix)
● Some people value accuracy (lookup
only)
50
51. What: Participant Entry
Overview
Dairy
Solid
Fats
Fruit
Participant noted
finding an enjoyable Sodium
protein bar.
Participant noted trying
Refined
Grains to consume more Added
whole grains & dark Sugar
Whole
Grains green veggies
Oils
Other
Veggies
Dark Green &
Orange Veggies Protein
Day Day
51
52. What people entered
indicates…
● Where their attention was over time
– What they weren’t paying attention to:
• Solid fats
– What they were paying attention to:
• Grains
• Vegetables
52
53. When: People made entries based
on their routine.
● Depending on routine.
– Routines varied from person to person
– Routines changed over time (in 3wk
period)
– Routines were linked to location
● What prevented you from entering in a
timely manner?
– “My phone was charging in another room”
– “I was doing something else when eating”
53
54. Takeaways from POND
● There is no “one-size-fits-all”
approach for dietary self-monitoring
● Search still matters
● Appreciation of the analysis
– Motivation to eat fruits, veggies and
whole grains
54
55. Contributions
● BALANCE: Examined and characterized
problems with database-based food diaries
on mobile phones.
● Food Indexes for Dietary Tracking:
Examined, characterized and evaluated
goal-based, diet quality-oriented food
diaries.
● Design of POND: Designed and developed
pattern-oriented nutrition diary.
● Evaluation of POND: Insights based on
real-world evaluation.
55
57. Overall observations
Support POND +1s enabled easy capture of whole, self-
target prepared, healthy foods (recipes), while
behavior. BALANCE made it time consuming.
The POND overview provided a target for
changing dietary intake (“more vegetables”).
57
58. Overall observations
Support POND +1s enabled easy capture of whole, self-
target prepared, healthy foods (recipes), while
behavior. BALANCE made it time consuming.
The POND overview provided a target for
changing dietary intake (“more vegetables”).
Food Querying a food database is easy if you
queries. know what to search for.
58
59. Overall observations
Support POND +1s enabled easy capture of whole, self-
target prepared, healthy foods (recipes), while
behavior. BALANCE made it time consuming.
The POND overview provided a target for
changing dietary intake (“more vegetables”).
Food Querying a food database is easy if you
queries. know what to search for.
Routines. People expect tools to fit into their current
routine.
Routines vary from person to person, but
they’re observable, and people talk about the
consistency of the routine in terms of location.
59
60. Future Work
● How different people use POND
– Motivation
– Health literacy
– General literacy
60
61. Future Work
● How different people use POND
● What happens after the intervention
– When people stop using POND, is the
reduction the same as when people stop
using BALANCE?
61
62. Future Work
● How different people use POND
● What happens after the intervention
● The “Health and Wellness
Ecosystem”
With tracking physical activity
and weight, detail of food
tracking can vary.
62
63. Gaetano Borriello. James Fogarty.
Julie Kientz. Wanda Pratt.
Glen Duncan. Deonna Hughes. Heather Snively. Jonathan Lester. Karl Koscher. Tammy
Denning. Waylon Brunette. Barbara Breummer. Amy Karlson. A.J. Brush. Beverly
Harrison. Lydia Musher. Lindsay Michimoto! Kayur Patel. Katie Kuksenok. Saleema
Amershi. Morgan Dixon. Hao Lu. Ryder Ziola. Yaw Anokwa. Brian DiRenzi. Carl
Hartung. Alan Liu. Rohit Chaudhri. Sunny Consolvo. Tammy Toscos. Pedja Klasnja.
Jon Froehlich. Kate Everitt. James Landay. Magda Balazinska. David Notkin. Linda
Shapiro. Anna Cavendar. Brian Van Essen. Dan Goldman. Kasia Wilamowska. Suporn
Pongnumkul. Scott Saponas. Colin Dixon. YongChul Kwon. Jonah Cohen. Emily Ryan.
Meredith Skeels. Andrea Grimes. Dave Thurman. Mary Frances Lembo. Liz Jurrus.
Alan Chappell. Andrew Cowell. Michelle Gregory. Shuli Gilutz. Laura Pina. Mark van
der Helm. Dave Grundy. Vanessa Chan. Karen Wilcox. Rebecca Morss. Benko. Jen
Cohen. Carol Strohecker. Edith Ackermann. Mitch Resnick. Justine Cassell. Jan
Borchers. Eytan Adar. Mira Dontcheva. Maya Rodrig. Richard Davis. Ben Lerner. John
Kim. Debbie Gracio. Judi Thompson. Kate Smith. …
63
Thanks for comingPhD CandidateUW Computer Science & EngineeringResearch Focus: HCI & Ubiquitous ComputingSpecifically: for mobile Health and Wellness
I’m going to start this talk out on a morbid foot– the causes of death in the US in 2006.
As we see, a full two-thirds of deaths were caused by Heart Disease, Cancer and Diabetes– all of which are believed to be at least somewhat preventable and treatable by lifestyle, specifically diet. The incidence of lifestyle diseases in our country (and globally) indicates the need for tools to support lifestyle changes.
Changing your lifestyle behaviors, such as eating and exercise, is hard, and research shows that the most effective tool to support change is self-monitoring. Sustaining self-monitoring is hardCorrelation between stopping self-monitoring and reducing target behavior
My dissertation is focused on self monitoring of nutrition behaviors. I offer 4 contributions.
BALANCE is a tool designed to provide glanceable, real-time feedback about the user’s current energy intake/expenditure balance over the course of the day. Energy expenditure (calories burned) is measured by the MSB. MSB is self-contained unitContains accelerometers and other sensors common in phones todayThe MSB is worn on the waistEnergy intake (calories eaten) is collected via a food diary and database on a cell phoneThe viz/feedback is provided on a mobile phone. In order to provide timely, relevant feedback that enables users to make good decisions, they need to enter what they eat, when they eat it, and as correctly/accurately as possible. Calorie expenditure calculation was validated independentlyFOCUS ON: Feedback from users about the food diary
I’ll start with a quick example of how the BALANCE food diary works.
I start with a food query.
I use the QWERTY keyboard (hardware) to begin entering the food I ate. Suggestions of common or previously used foods populate on the right
As I keep typing, the list is filtered.
I find the food I that most closely matches what I ate.
I specify how much of it I ate.
And it appears on my list for today, along with the number of calories.
And then I repeat the process to also add Butter, Syrup, and the berries to my breakfast.
For this talk, I’m going to focus on just the evaluations that resulted in user feedback about the food diary portion of the project. 2 phases of evaluationValidation:Compared what was entered with a 24-hr recall to estimate quality of food entries
So, what was some of the user feedback? Food diary on mobile is good: They could “play” on the phoneIt could be used in pockets of timeNo one knows what they’re doing
Not sure if this goes here or elsewhere: The database was clearly causing angst and increasing time. If we got rid of it, could we “reduce the amount of resources” required to keep track of food intake? If we could reduce the resources/overhead, would people use it longer? Are people not continuing to use food diaries because it takes too much work?
So, what was some of the user feedback? Food diary on mobile is good: They could “play” on the phoneIt could be used in pockets of timeNo one knows what they’re doing
How quick/easy can we make food entry, while still providing value?
Taking a step back, we considered the original goal of tracking what you eat: change eating behaviors.
Something here or nearby:
Food indexes are a tool used by the nutrition research community to formalize characterizations of what people eat. Multiple componentsUsually combined to generate an overall scoreGenerally used to compare diet patterns of populationsAdequacy means that the scoring supports eating at least a certain amount of something and encourages more; Moderation means that the scoring supports restricting intake of a component.
FBQI
HEI
BALANCE
BAL versus 2 non-db approaches. I was afraid to run this test, because we had already heard people tell us how bad BALANCE was. I didn’t think it’d be a fair comparison! BALANCE was the strawman. SURPRISE! People LIKED BALANCE in the lab!
BAL versus 2 non-db approaches. I was afraid to run this test, because we had already heard people tell us how bad BALANCE was. I didn’t think it’d be a fair comparison! BALANCE was the strawman. SURPRISE! People LIKED BALANCE in the lab!
BAL versus 2 non-db approaches. I was afraid to run this test, because we had already heard people tell us how bad BALANCE was. I didn’t think it’d be a fair comparison! BALANCE was the strawman. SURPRISE! People LIKED BALANCE in the lab!
That raised the question: well, they said they didn’t like BALANCE in the field; they said they do like it in the lab. They also kinda liked HEI in the lab. But, really, what happens in the real world? Do they still like HEI, BALANCE? Given the choice, do they switch?
I’ll give you a short walk through of the POND tool. Android appEach row represents an HEI-05 componentThe gray blocks represent a daily goalA filled block shows you’ve consumed something in that component todayThe +1 button on the right allows you to increment that component A long press on the +1 allows you to add ½ a block.
You can also look up
The record for a food from the database gives an idea of how to count it in terms of the components. Not all foods in the database give a good lookupShows a “what if” analysis: Colored blocks show the current foodLight gray blocks show what’s been eaten todayDark gray blocks show your goal
Here we have an overview of everything a participant entered, over the entire 3 weeks. Each column represents one day. As in the POND application, the gray boxes indicate the goal, and the colored boxes indicate what was entered. I’m showing you this to you primarily because it shows us where the participant’s attention was throughout the study.
Remember the HEI + BAL transition
Since this is a practice defense talk, I’m going to begin with a quick overview of my contributions and then dive into the content.
Supporting the target behavior: *
Querying the database is easy if you know what to search for. BALANCE: in situ queries were hardIn-lab comparison study: given queries were easyPOND: searching for packaged food is natural
Routines.BALANCE: “I waited until I got home at night”POND: It depended where I was, what I was doing, if I was at home or work
Next paper in this topic,
Next paper in this topic,
Next paper in this topic,
Stephen Intille. MarshiniChetty. Sinan al-Saffar. Liz Tseng. Erin Solovey. EunyeeKoh. John John. Shaun Kane. KirstieHawkey. Shwetak Patel. KoriInkpen. Michael Bernstein. Raphael Hoffman. RosaliaTungaraza. Roxana Geambasu. StefShoenmackers. never-grads. Matt Kay. Jared Bauer. Harlan Hile. Steve Stein. Curt West. Elsa Augustenburg. Jon Barr. Sadie Johnson. Jared Chase. Tony Bladek. Sean Munson. Alan Au. Brian Smith.