Conference PaperPDF Available

Food and Mood: Just-in-Time Support for Emotional Eating


Abstract and Figures

Behavior modification in health is difficult, as habitual behaviors are extremely well-learned, by definition. This research is focused on building a persuasive system for behavior modification around emotional eating. In this paper, we make strides towards building a just-in-time support system for emotional eating in three user studies. The first two studies involved participants using a custom mobile phone application for tracking emotions, food, and receiving interventions. We found lots of individual differences in emotional eating behaviors and that most participants wanted personalized interventions, rather than a pre-determined intervention. Finally, we also designed a novel, wearable sensor system for detecting emotions using a machine learning approach. This system consisted of physiological sensors which were placed into women’s brassieres. We tested the sensing system and found positive results for emotion detection in this mobile, wearable system.
Content may be subject to copyright.
Food and Mood: Just-in-Time Support
for Emotional Eating
Erin A. Carroll
University of Rochester, Computer Science
Rochester, NY 14627 USA
Mary Czerwinski, Asta Roseway,
Ashish Kapoor, Paul Johns, Kael Rowan
Microsoft Research
Redmond, WA 98052
{marycz, astar, akapoor, paul.johns,
m.c. schraefel
University of Southampton
Southampton, Hampshire, UK
Abstract—Behavior modification in health is difficult, as
habitual behaviors are extremely well-learned, by definition.
This research is focused on building a persuasive system for
behavior modification around emotional eating. In this paper,
we make strides towards building a just-in-time support system
for emotional eating in three user studies. The first two studies
involved participants using a custom mobile phone application for
tracking emotions, food, and receiving interventions. We found
lots of individual differences in emotional eating behaviors and
that most participants wanted personalized interventions, rather
than a pre-determined intervention. Finally, we also designed
a novel, wearable sensor system for detecting emotions using a
machine learning approach. This system consisted of physiological
sensors which were placed into women’s brassieres. We tested the
sensing system and found positive results for emotion detection
in this mobile, wearable system.
We eat not just because we are hungry and craving nutrients
but also for a host of emotional and habitual reasons. There is
no single term that encompasses the combination of lifestyle,
hedonic, emotional, or habitual over-eating that leads to obe-
sity. For this paper, we will use the term “emotional eating”
as a placeholder for non-homeostatic eating (i.e., eating that
is not physiologically required). Persuasive technologies have
been used within the health space to assist people in developing
new habits around health, from making better eating choices
to encouraging a more active lifestyle. These systems have,
for the most part, relied on fixed contexts to offer support:
alarms at particular times of the day; showing performance
updates when one glances at their mobile phones [1]; and
user-activated support [2,3]. However, one cannot rely solely
upon fixed solutions for persuasive technology to effect long-
term behavioral change. An alternative approach is to intervene
proactively to prevent the behavior from happening. In the case
of emotional eating, our goal is to provide an intervention
before the person turns to food for emotional support.
We begin to investigate just-in-time interventions to support
behavior modification for emotional eating. Designing such
a system is an ambitious endeavor: it involves exploring the
emotional triggers of eating, developing elaborate technol-
ogy for automatically detecting emotions, and investigating
intervention approaches for emotional eating. In this paper,
we present three user studies that move us towards the goal
of developing a just-in-time support system for emotional
eating. The purpose of the first study was to gather data
about emotional eating patterns. The second study investigated
the feasibility and benefits of emotional eating interventions.
Finally, the third study focused on emotion detection with
machine learning and custom-built wearable sensors.
A. Persuasive Health Technologies
Persuasive technologies for health and weight management
have become ubiquituous over the past few years with several
off-the-shelf fitness technologies like FitBit
and BodyBugg
helping to make the logging of activity and physiological
state more accessible. These fitness gadgets are designed to
help people measure physical activity and sleep quality and
to motivate increased movement. Amft and Troster [4] built
sensors for monitoring chewing and swallowing, and they
proposed how these sensors could be included in an elaborate,
(yet uncreated) persuasive system for health. Purpura et al.
also proposed a similar, uncreative persuasive system [5].
The technology in their system included, to name a few
parts: an earpiece for monitoring chewing and swallowing;
augmented reality glasses for capturing foods consumed; and
heart rate recording for sensing exercise. This proposed system
also connected through a mobile phone application in order
to process the data and for the person to receive feedback.
The hypothetical feedback, whether from a social network,
a close friend, or pre-recorded messages, served as a health
intervention to encourage the person to be more active or
consume less food.
Not at all persuasive systems need to be as elaborate as
those discussed above. A persuasive system with peer support
alone can help people be more successful in weight loss.
Mutsaddi et al. [6] utilized text messages as a form of social
support for encouraging more physical activity. They found
that even after the novelty of text messages wore off, the text
messages were still beneficial because they served as reminders
to participants to be more active.
B. Eating for Non-Homeostatic Reasons
There are many theories around why we eat. One is largely
homeostatic: we eat because we need fuel to survive; we
crave certain foods because we require certain nutrients to
function [7]. Other reasons are more nuanced, though these
too are intertwined in physiological responses. For instance,
many people reach for calorically dense foods, like donuts,
when stressed. Gilhooly et. al characterized these responses as
“instinctive” in that at some point these behaviors may have
served a survival function [8]. Stress releases hormones, which
trigger a fight or flight response; thus, grabbing high-energy,
available foods (like the sugary donut) would be effective
for energy production. Likewise, when confronted with the
variety of a buffet, it seems a food scarcity mentality kicks
in, and more food is taken in than necessary. The challenge
in a food abundant culture is that these instinctive responses
are no longer helpful. An important insight around over-eating
behavior is that non-homeostatic eating patterns can be re-
educated [8]. In other words, both our physiological responses
(the release of ghrelin to cue stomach grumbling at particular
times [7]) and psychosocial responses (eat in the presence
of food or in response to stress) are malleable. Therefore,
technology that is used to intervene before the maladaptive
behavior happens could provide some assistance towards long-
term, behavioral change.
C. Supporting Behavior Change
An increasingly well-regarded approach to support eating
behavior change is Cognitive Behavioral Therapy (CBT) [9].
The focus of CBT is to help a person become aware of their
“maladaptive behaviors” and then replace these with adaptive
ones “by modifying their antecedents and consequences and by
behavioral practices that result in new learning” [10]. A com-
mon approach to supporting discovery of these antecedents to
maladaptive behaviors, such as one’s cues for emotional eating,
is to have participants in CBT do work to identify triggers.
Typical approaches include keeping daily food and mood logs
or journals. In some cases, the goals of the logs are to high-
light particularly positive states (“benefit finding”) to enhance
success with new, adaptive behaviors [11]. One particularly
relevant process has been called “real-time self-monitoring”
around eating, related behaviors, and feelings [12]. The ratio-
nale for the criticality of monitoring is that “it helps patients
to be more aware of what is happening in the moment so that
they can begin to make changes to behavior that may have
seemed automatic or beyond their control” [12]. An important
aspect of CBT is bringing the automatic or thoughtless actions
into consciousness for deliberate engagement/change.
D. Implicit Emotion Detection
Emotion detection with sensor data has been carried out
in the past. McDuff et al. used a variety of signals (i.e.,
electrodermal activity, posture, facial expressions, etc.) to
detect emotions using a machine learning approach with self-
reported ratings of emotion serving as the ground truth in their
emotion classifiers [13]. They found that the electrodermal
activity (EDA) signal was the most beneficial, which is a
measure of the eccrine sweat glands [14,15]. McDuffs et al.s
detection system involved sensors that required users to be
tethered to their desk (e.g., Kinect, web cam, etc.).
The goal of our system was to perform emotion detection
in a mobile, wearable system, which allows us to collect data
as users move about their day. Chang et al. created a mobile
system for detecting user emotion but this was done via activity
modeling and speech prosody tracking [16]. In contrast, the
goal of our system was to use on-body sensors with EDA and
electrocardiogram (EKG), since these sensors have been shown
to be reliable for emotion detection [14,17,18].
Designing a system to provide just-in-time interventions
for emotional eating is an ambitious endeavor. Consider the
following hypothetical scenario:
Sally has been home from work for a few hours, and she
finds herself rather bored. An application on Sally’s mobile
phone has also detected that she is bored by reading her
physiological state through wearable sensors. Since this mobile
application has previously learned that Sally is most suscep-
tible to emotional eating when she is bored, the application
provides an intervention to distract Sally and hopefully prevent
her from eating at that moment.
From this scenario, we see three key requirements for a
just-in-time support system for emotional eating. First, the
application has to be aware of the user’s emotional eating
patterns. Does Sally emotionally eat only when she is bored?
Second, the system needs to be able to implicitly detect
emotions. This involves wearable physiological sensors that are
connected to the mobile phone. Implicit detection of emotions
would then be possible through machine learning classification,
which requires training on large amounts of users’ data.
Finally, it is critical (and perhaps the most challenging) to
determine how to intervene. What type of intervention do we
design? How often do we intervene? How do we prevent it
from becoming an annoyance to the user?
Our approach to researching a just-in-time support system
for emotional eating was to make strides towards addressing
these three requirements. We studied these requirements across
three user studies, which have been summarized below.
Study 1: Gather Emotional Eating Patterns. We
investigated eating behaviors and corresponding emo-
tions of participants by having them self-report their
emotions and log their eating patterns using a custom
built application called EmoTree. The goal was to
understand their emotions associated with eating.
Study 2: Investigate An Intervention Technique.
The purpose of Study 2 was to learn about a particular
intervention technique for emotional eating. We proto-
typed implicit intervention by triggering an interven-
tion based on self-reported ratings of emotions. This
allowed us to gather early feedback about interven-
tions before implementing an automatic system. Are
users aided by the intervention? Was the intervention
sent at the appropriate time? What other types of
interventions would interest users?
Study 3: Emotion Detection with Wearables. This
work was a first step in building an automatic system.
We investigated the feasibility of using physiological
sensor data, combined with machine learning, to au-
tomatically detect emotions in a mobile system. We
also present the design of our wearable system.
EmoTree is a custom designed Windows 7 mobile phone
application, consisting of four screens. The default screen
provides an overview of the user’s logging activity and overall
sentiment (Figure 1A). The tree on this screen eventually
populates its leaves over time, as the user interacts with the
app. Each circle represents a day’s worth of activity and the
color green is used to indicate positive decisions based on food
intake. The user’s sentiment is aggregated as the background
sky color to indicate positive or negative valence of affect.
Fig. 1. EmoTree Mobile Phone Application. Main Page (A) and Intervention
Page (B)
As usage continues, the user may assess overall progress or
history by selecting the history icon at the bottom of the tree.
Inspiration for the interface design came from the notion of
tending a garden and using that as a metaphor to visually guide
users to make better choices [1]. Since the application was
designed for long-term use, the idea of a tree that gradually
grows based on your daily health decisions served best as a
visual guidance system that could invoke encouragement: the
healthier your choices, the healthier the tree.
To start using the application, the user goes to the main
screen (Figure 1A) and glances at their current mood. By
swiping to the left, the user is taken to a screen that asks: “How
do you feel?” (Figure 2B) Self-reporting of emotion was based
on the Russell’s Circumplex model [15], widely used across
the affective computing community [13,19], in which emotion
is represented two-dimensionally; valence on the x-axis and
arousal on the y-axis. Figure 1A shows the emotion self-report
tool that participants used. In our user interface, we used the
terminology of “Negative” to “Positive” to describe valence,
and “Pumped” to “Relaxed” to describe arousal, as in McDuff
et. al [13]. After indicating their current emotional state, the
users also indicated how engaged they were with their current
task, on the same screen. Reminders to self-report emotions
were sent to the users’ mobile phones every hour, on the hour.
Users were also instructed to swipe right from the main
screen to log any food that had been consumed (see Figure 2A)
for what they were asked to log. Logging food populated
the leaves on the main screen. After logging food, the user
was immediately sent to the self-report page to indicate their
emotions prior to eating. The task engagement question was
not asked after logging food.
Lastly, we introduce the intervention screen (Figure 1B).
which consisted of a deep breathing exercise. Users were
instructed to tap on the screen for each breath that they took
for 10 seconds (or 10 taps). The intervention was employed in
Study 2 and will be discussed further in Section VI.
In Study 1, we conducted a user study (N=12, 2 males)
to investigate emotional eating patterns using EmoTree. In
recruitment, all participants self-identified as emotional eaters.
Fig. 2. EmoTree Mobile Phone Application. Nutrition Log (A) and Self-
Report Tool (B)
Prior to the study, users filled out a pre-experiment survey
that asked about their emotional eating patterns, called the
Emotional Eating Scale [20]. All participants were given a
demonstration of the EmoTree app, and they were given
instructions on how and when to use the software. This study
took place over a period of four days (Tues-Fri). Participants
were asked to use EmoTree to self-report emotions every hour,
on the hour, in addition to reporting food. We collected at least
10 self-reports of emotion per day from participants.
A. Analysis and Results
Our pre-survey results on the Emotional Eating Scale
showed that the emotions marked as causing a moderate to
strong urge to eat included: on edge, nervous, upset, worried,
nothing to do, bored, irritated, restless, and discouraged. These
findings support research on emotional eating as reported
elsewhere [21].
From the four days of data collection, we created two
scatterplots for each person (2 plots x 12 participants). The
first scatterplot was of a person’s emotional ratings, which
visualized their emotional state over the duration of the study.
The second scatterplot was of a person’s emotional ratings
that were associated with food, showing us their emotional
state just before eating. It was not surprising to find that
there were a lot of individual differences and that people ate
when they were experiencing a variety of emotional states. In
fact, most users had eating events that occurred in all four
quadrants of the Circumplex model (Figure 2B). However,
we did find that six participants tended to eat when they
were predominantly stressed with the majority of eating events
occurring in the upper-left quadrant (i.e., negative & pumped).
We also observed across all participants that few eating events
happened when the users were in a calm/serene state, so
very few eating events were in the lower-right quadrant (i.e.,
positive & relaxed).
B. Intervention Technique
We decided to intervene based on when users provided self-
reports occuring in the upper-left quadrant of the Circumplex
model (i.e., negative & pumped). Our goal was to explore if our
intervention could help move participants from a stressful state
to a calmer state. Therefore, we developed a deep breathing
activity (Figure 1B), where a little bird is displayed that
counts users’ breaths with each tap on the screen. Several
components are at work in our design decisions. First, stress
has been identified as a contributor to obesity and to emotional
eating [7,21]. Second, we found in the analysis of EmoTree’s
emotional eating data that 6 out of 12 participants were
primarily stress eaters, and the Pre-Survey results also show
that participants primarily eat in response to feelings that were
associated with stress (e.g., on edge, upset, worried). Finally,
we leverage a CBT strategy for reducing emotional eating,
as described in Section II-C: by targeting stressful moments
that may be precursors to an emotional eating episode, the
breathing exercise may break the focus on stress, allowing for
further cognitive processing of what one may be about to do,
thus offering the opportunity to choose a different, and more
positive, action.
In Study 2, we released a new version of EmoTree to
participants, which included a deep breathing intervention
that was triggered in response to self-reported emotions of
stress/anxiety. This study was focused on the usefulness of the
deep breathing intervention technique.
This study was comprised of the same participants from
Study 1 with the exception of one who withdrew due to travel.
This study also took place over a period of four days (Tues-
Fri). During this time, participants (N=11, 1 male) continued
to use EmoTree for tracking food and self-reporting emotions
throughout the day, using the same instructions from Study 1.
We investigated intervention effectiveness with Pre- and Post-
Experiment surveys. The Pre-Survey included questions about
participant awareness of eating behaviors and emotions as a
result of tracking emotions and nutrition in Study 1. The Post-
Survey asked questions about the success of the deep breathing
activity (i.e., effectiveness in preventing emotional eating and
for reducing stress/anxiety); whether triggering the intervention
based on stress/anxiety was effective; and solicited suggestions
for alternative intervention techniques. The Post-Survey sought
to understand whether we should build a fully automated
system, and if so, how a particular intervention should be
implemented. Specifically, we asked participants questions
about whether they would be interested in the system; how
successful the deep breathing activity was (i.e., effectiveness
in preventing emotional eating and for reducing stress/anxiety);
whether intervening based on stress/anxiety was effective; and
suggestions for alternative intervention techniques.
A. Analysis & Results
In the Pre-Survey, all participants reported that they became
more aware of their eating behaviors, and 87.5% reported
that they became more aware of their emotions. Qualitative
feedback also indicated that logging in EmoTree was an
excellent awareness tool:
“I was eating without being aware of it, but by having
to log both my eating habits and my emotions, I
became aware of triggers for emotional eating, and
also more aware of the health (or lack thereof) in my
“I became more conscious when I was about to eat
or drink and self-reflected on why I was consuming
“I noticed more how I was feeling when I ate, and this
was not something you normally stop to think about.
However, only 37.5% reported that their eating behaviors
changed as a result of Study 1, so while logging made
participants more aware, it seemed that most needed something
extra to incent real change.
The Post-Survey was given to participants after completing
Study 2. The Post-Survey asked participants how accurate the
stress/anxiety self-reporting trigger was for the intervention
timing. The average accuracy was 2.80 (SD=1.30) on a scale of
1 (Not At All Accurate) to 5 (Extremely Accurate). This result
was not surprising given that we knew from Study 1 that our
participants ate for a variety of reasons, not just when stressed.
We also asked how effective the deep breathing intervention
was at reducing stress (M=3.0, SD=1.87) and the effectiveness
of deep breathing in preventing emotional eating (M=2.20,
SD=1.30). These last two questions were on a scale of 1 (Not
At All Effective) to 5 (Extremely Effective). We were pleased
that the deep breathing worked moderately at reducing stress,
but its effectiveness at preventing emotional eating was not as
strong as we might have hoped for.
Participants were also asked about suggestions for other
intervention techniques. We learned that personalization is
desirable for intervention, not only in regards to individual
differences in eating patterns but also for personal preferences
about interventions. For instance, 3 out of 11 participants
were enthusiastic about the deep breathing activity and tried
to incorporate it throughout their day, but the others wanted
alternatives. The alternative interventions that they suggested,
included: presenting something funny to alleviate boredom
or stress; having them fill out a gratitude questionnaire (i.e.,
reminding them what they have to be thankful for); meditation;
suggestions to take a walk; brain teasers/games; asking if they
are really hungry; reading or writing; and calling a friend.
From these results, personalizing the intervention would be the
best solution. However, offering a menu of these interventions
could break the food focus and prove to be a useful approach,
which is consonant with intervention work in CBT.
In Study 1 and 2, we explored emotional eating patterns
and investigated the feasibility and benefits of developing an
elaborate, integrated system, as described in the scenario. We
found some initial results that suggested providing awareness
and just-in-time support for emotional eating could work with
better personalization on timing and intervention. To move
towards the goal of a personalized, integrated system, this
study was focused on investigating the feasibility of using
physiological sensors to implicitly detect emotions. While
implicit emotion detection has been done in the past [13],
this is the first study, that we are aware of, that makes use
of wearable, mobile sensors for detecting emotions. In this
section, we discuss our novel sensing system, including the
sensors employed and the wearables. We also discuss our
methodology for collecting data. Finally, we present positive
results for classifying emotions using wearable sensors.
Fig. 3. Sensor placement outline, showing where EKG and EDA was
collected and connected to the custom board.
A. Emotion Sensing System
1) Sensors: We used custom boards called GRASP
(Generic Remote Access Sensing Platform) which is a real-
time system and is comprised of a physical sensor board,
corresponding firmware, software libraries, and an API. The
GRASP board has an MSP-430 microprocessor and is powered
by a Lithium-Ion polymer (3.7V) battery. GRASP can sample
up to eight bio-signal channels (TI80S1298) simultaneously.
The GRASP boards in this study were configured to capture
heart rate and respiration with an electrocardiogram (EKG)
sensor; skin conductance with an electrodermal activity (EDA)
sensor; and movement with a 3-axis accelerometer and a 2-
axis gyroscope. The board transmitted the sensor stream data
to a mobile phone application using Bluetooth, which was then
stored remotely into a Microsoft Azure Cloud.
2) Design of Wearables: The design motivation for the bra
sensing system (Figure 3) was driven by a few key factors.
First, we needed a form factor that would be comfortable when
worn for long durations. We also needed a way to gather both
EKG and EDA signals; so ideally, we wanted to collect those
signals from the same wearable. The bra form-factor was ideal
because it allowed us to collect EKG near the heart. Ultimately,
we chose to leverage the functionality and wearability of a
bra, but had to consider that each participant’s chest and rib
size would vary greatly. Rather than build each participant
their own embedded sensor bra, we aimed for a much more
modest solution: conductive pads that could be inserted or
removed. This provided us flexibility in recruiting participants,
in addition to resolving the wash-ability aspect. Three pads for
each participant were required in order to capture both EKG
and EDA (Figure 3). For EKG, two pads were designed to fit
snug against each side of the ribs. These pads were designed
with more surface space to help reduce noise in the signal. The
EDA pad was initially designed to fit on the backside of the
bra; however, we were unable to gather sufficient signal due
to low sweat levels in the back. We therefore had participants
relocate the EDA pad into the bra cup, just under the breast.
It is worth nothing that collecting EDA from this part of the
body is non-standard; however, EDA was still a useful signal.
The board sits at the center of the user’s sternum (Figure 3),
encased in a fabric pouch that was stitched to the outside of
the bra. Each sensor pad is comprised of 2.5mm of laser cut
neoprene sandwiched with batting and laser cut conductive
silver ripstop fabric (Figure 4). A laser cut outline of cotton
was top stitched by a sewing machine to enclose all the
Fig. 4. EKG sensor pads, connected to a GRASP board (A); Conductive
thread insulated by scotch tape (B)
material. The insulated wire turned out to be inflexible and
obstructive to the wear, so we explored ways in which we
could make an electrical connection using conductive thread.
For insulation, we sandwiched the thread with two pieces of
scotch tape. This method worked remarkably well in that it
retained the soft and flexible nature of the thread but was also
not noticed when worn. One end of the conductive thread was
sewn into the pad and then covered by a top stitch, while the
other end was tied to thin wire, heat wrapped, and crimped to
a connector on the GRASP board. The average conductivity
between the sensor and connector was approximately 12 ohms.
3) Methodology: We conducted a user study to see if
emotion detection through physiological sensing was possible
in a mobile system. To this end, we had four women from
our research lab participate in this study. The study involved
these women wearing the three conductive fabric pads inside
their bras, as previously described. They also used the mobile
application, EmoTree, in order to self-report their emotions,
which was used as labeled data in our machine learning
classifier (ground truth).
Participants wore the bra sensing system and reported
their emotions for about 4-6 hours a day over a period of
approximately four days. It was very tedious for participants
to wear our prototyped sensing system, as the boards had to
be recharged every 3-4 hours, which resulted in participants
having to finagle with their wardrobe throughout the day.
4) Analysis and Results: Similar to McDuff et. al [13], we
took a machine learning approach for predicting emotions. The
sensor data went through several pre-processing stages before
we could perform classification. First, we applied two different
filters to the raw EKG signal: a 2Hz low-pass filter to get
respiration and a 2Hz high-pass filter to get heart rate. Next,
we normalized all 8 signals (heart rate, respiration, EDA, 3-
axis accelerometer, 2-axis gyroscope) between 0-1. Finally, we
extracted features from the signal by creating 5 bands that fell
within our normalized values. These bands were: 0 to <0.2,
0.2 to <0.4, 0.4 to <0.6, 0.6 to <0.8, and >0.8. We used
these bands to determine within a 10-minute time period the
proportion of data that occurred within each band. This feature
extraction process resulted in 40 features (8 signals x 5 bands)
for each person. The labeled data for classification was self-
reported ratings of emotion (x-valence and y-arousal), and the
attributes were the 40 extracted features.
We used a classification framework based on Gaussian
Process Regression (GPR) [22], which has been shown to
be very effective for multimodal affect recognition [23] and
fits well into the context of our classification task. In par-
ticular, we define a similarity (a kernel) function between
two observations x
and x
using a Radial Basis Function:
, x
) = exp k x
. GPR considers this
similarity when classifying points and assigns labels to test
cases such that similar labels are assigned to similar points.
We refer readers to Kapoor et. al [23] for further details
on affect classification using Gaussian Processes. For testing
the classifiers, we used a leave-one-example-out methodology.
Specifically, we consider data recorded from all but one data
point as the training corpus and then tested the algorithm on the
left-out observation. This process was repeated for all points.
We were able to classify arousal at 75.00% and valence
at 72.62% accuracy. We observe that for both arousal and
valence the recognition accuracy is significantly better than
chance. Furthermore, we would like to highlight that the
recognition accuracy achieved here is at par with other affect
recognition systems [13]. Based on these results, we conclude
that building a wearable, physiological system is feasible.
However, we will continue to explore how to build a robust,
real-world system that stands up to every day challenges with
regards to battery life, comfortability, and being suitable for
both men and women. Since these classification results were
based on log files, rather than real-time sensor data, our next
iteration will also run using real-time sensor data that is able to
predict emotions and show an appropriately time, personalized
In this paper, we began to investigate just-in-time interven-
tion to support behavioral modification in emotional eating.
Designing a just-in-time support system for emotional eating
is an ambitious endeavor, and to this end, we presented three
user studies that moved us towards the goal of developing
a fully integrated system. We found in Study 1 from user
logging that 6 out of 12 participants primarily ate when
they were stressed. In Study 2, we found that participants
were enthusiastic about interventions for emotional eating, but
that most wanted personalized interventions, rather than the
single, pre-determined deep breathing interventions. Finally, in
Study 3, we investigated using wearable sensors to implicitly
detect emotions while participants were mobile. Using log
files, we were able to detect arousal at 75.00% and valence
at 72.62%.
There are several future research directions that are essen-
tial to building an integrated system for just-in-time support.
First, we have already expanded our implicit emotion detection
system to work using real-time sensor data, opposed to log
files. While the wearable sensors in the brassiere form factor
only allowed women to participate in Study 3, we have now
moved towards using the Affectiva Q
sensor bracelets for this
collection. Our pilot results have been quite promising for men
and women. Finally, we are currently exploring the just-in-
time intervention space in which we test different intervention
approaches in longitudinal studies. We consider research in
intervention design to be critical in future iterations, including
personalized methods of responding to the user, in addition to
investigating formal theories of behavior change [24].
Affectiva Q:
[1] S. Consolvo, P. Klasnja, D. W. McDonalds, D. n. Avrahami,
J. Froehlich, R. Libby, K. Mosher, and J. A. Landay, “Flowers or a
robot army? Encouraging awareness & activity with personal, mobile
displays, in ACM UbiComp 2008, pp. 54–63, 2008.
[2] P. Y. Chi, J. H. Chen, and J. L. Lo, “Enabling Calorie-Aware Cooking
in a Smart Kitchen, Persuasive 2008, pp. 116–127, 2008.
[3] D. Mankoff, G. Hsieh, H. C. Hung, S. Lee, and E. Nitao, “Using Low-
Cost Sensing to Support Nutritional Awareness,” in Proceedings of ACM
UbiComp 2002, pp. 371–387, 2002.
[4] O. Amft and G. Troster, “On-body sensing solutions for Automatic
Dietary Monitoring, IEEE Pervasive Computing, vol. 8, no. 2, pp. 62–
70, 2009.
[5] S. Purpura, V. Schwanda, K. Williams, W. Stubler, and P. Sengers,
“Fit4Life: The design of persuasive technology promoting healty behav-
ior and ideal weight, in Proceedings of ACM CHI 2011, pp. 423–434,
[6] A. Mutsaddi and K. Connelly, “Text Messages for Encouraging Physical
Activity: Are they effective after the novelty effect wears off?, in
Proceedings of Pervasive Health, 2012.
[7] J. F. Davis, D. L. Choi, D. J. Clegg, and S. C. Benoit, “Signaling through
the ghrelin receptor modulates hippocampal function and metal behavior
in mice, Physiology & Behavior, vol. 103, no. 1, pp. 39–43, 2008.
[8] C. H. Gilhooly, S. K. Das, J. K. Golden, M. A. McCrory, G. E.
Dallal, E. Saltzman, F. M. Kramer, and S. B. Roberts, “Food cravings
and Energy Regulation: The characteristics of craved foods and their
relationship with eating behaviors and weight change during 6 months
of dietary energy restriction, International Journal of Obesity, vol. 31,
no. 12, pp. 1849–1859, 2007.
[9] S. G. Hoffman, A. T. Sawyer, and A. Fang, “The empirical status of
the “new wave” of CBT, Psychiatr Clin North Am, vol. 33, no. 3,
pp. 701–710, 2010.
[10] C. Otte, “Cognitive behavioral therapy in anxiety disorders: Current
state of the evidence, Dialogues in Clinical Neuroscience, vol. 13,
no. 3, pp. 413–421, 2011.
[11] A. J. Guastella and M. R. Dadds, “Cognitive-behavioural emotion
writing tasks: A controlled trial of multiple processes, Journal of
Behavioural Therapy and Experimental Psychicatry, vol. 39, no. 4,
pp. 558–566, 2008.
[12] R. Murphy, S. Straebler, Z. Cooper, and C. G. Fairburn, “Cognitive
behavioral therapy for eating disorders, Psychiatr Clin North Am,
vol. 33, no. 3, pp. 611–627, 2010.
[13] D. McDuff, A. Karlson, A. Kapoor, A. Roseway, and M. Cszerwinski,
AffectAura: An Intelligent System for Emotional Memory, in Pro-
ceedings of ACM CHI 2012, pp. 849–858, 2012.
[14] W. Boucsein, Electrodermal Activity. New York: Springer-Verlag, 1992.
[15] J. A. Russell, “A circumplex model of affect, Journal of Personality
and Social Psychology, vol. 39, no. 6, pp. 1161–1178, 1980.
[16] K. Chang, D. H. Fisher, and J. Canny, “From Affect to Mental Health:
A Speech Analysis Library for Unobtrusive Monitoring on Mobile
Phones, in 2nd International Workshop on Sensing Applications on
Mobile Phones (PhoneSense), 2011.
[17] R. L. Mandryk and M. S. Atkins, “A Fuzzy Physiological Approach
for Continuously Modeling Emotion During Interaction with Play
Environments, International Journal of Human-Computer Interaction,
vol. 6, no. 4, pp. 329–347, 2007.
[18] R. W. Picard, Affective computing. Cambridge, MA, USA: MIT Press,
[19] C. Latulipe, E. A. Carroll, and D. Lottridge, “Love, Hate, Arousal,
and Engagement: Exploring Audience Responses to Performing Arts,
in Proceedings of ACM CHI 2011, pp. 1845–1854, ACM Press, May
[20] A. M. Koball, M. R. Meers, A. Storfer-Isser, S. E. Domoff, and D. R.
Musher-Eizenman, “Eating When Bored: Revision of the Emotional
Eating Scale With a Focus on Boredom, Health Psychology, vol. 31,
no. 4, pp. 521–524, 2012.
[21] D. J. Wallis and D. M. Hetherington, “Stress and eating: The effects
of ego-threat and cognitive demand on food intake in restrained and
emotional stress, Appetite, vol. 43, no. 1, pp. 39–46, 2004.
[22] C. E. Rasmussen and C. Williams, GPML: Gaussian Processes for
Machine Learning. Cambridge, MA: MIT Press, 2006.
[23] A. Kapoor, W. Burleson, and R. W. Picard, “Automatic Predictions
of Frustration, International Journal of Human-Computer Interaction,
vol. 65, no. 8, pp. 724–736, 2007.
[24] E. B. Hekler, P. Klasnja, J. E. Froehlich, and M. P. Buman, “Mind the
theoretical gap: interpreting, using, and developing behavioral theory in
HCI research, in Proceedings of ACM CHI ’13, pp. 3307–3316, ACM
Request Permissions, 2013.
... It is unclear, however, the extent to which HIT interventions for clinical eating disorders is applicable to nonclinical populations, which is the focus of this research study. Carroll et al. (2013) has carried out what is perhaps the most ambitious preliminary HCI research on ESRE. The research comprised three studies through which emotional eating was addressed through the just-in-time interventions of a smartphone app that supported food and emotion tracking, the provision of deep-breathing relaxation exercises via smartphone app, and testing of physiological sensors designed to detect emotions. ...
... Additionally, if interventions should be tailored for users, are there any user traits that predict which kind of interventions are most likely to be accepted by and effective for the user? Notwithstanding the contributions of Carroll et al. (2013)'s preliminary work we still lack a comprehensive understanding from which to develop HIT interventions for overweight/obese women struggling with ESRE. ...
... To the author's knowledge, Carroll et al. (2013) is the only experimental technology study that contemplated monitoring user stress level or emotion state with ESRE as its primary consideration. Carroll et al. (2013) tested technology in which users tracked their own emotions and diet, and tested another technology in which stress levels were monitored and measured by wearable physiological sensors, though not considered along with dietary intake. ...
Extensive research shows that negative emotions and stress can prompt eating behavior that is in excess of physiological nutritional needs. Additionally, research indicates that women are more likely than men to cope with negative emotions and stress by overeating. There is a dearth of Human-Computer Interaction (HCI) research related to Health Information Technology (HIT) interventions that address overeating in context of negative emotions and stress. As a result, there is little guidance for HCI design and evaluation in this area. The study uses a convergent mixed methods design to understand how HIT can support overweight/obese women curb emotion- and stressed-related eating (ESRE), with the ultimate goal of sustained weight management. In the interview study strand, cross sectional semi-structured interviews (N = 22) explore ESRE behavior in overweight/obese women (BMI ≥ 25). The survey study strand, consisting of a questionnaire (N = 430) administered to overweight/obese women, comprises data about user characteristics, stress, ESRE, and coping, and digital access and skills. The thesis found that that overweight/obese women who engage in ESRE encountered stressors that spanned from daily hassles, persistent challenges, and life-changing losses; they also experience stressors related to serving as caretakers and social support providers. They used food as a coping response to the stress they encounter, and tend to associate food with social support. Furthermore, some have characteristics that make them particularly vulnerable to ESRE behavior. This thesis suggests that HIT should assist users before their coping eating response is triggered. Additionally, HIT should support women in becoming aware of their tendencies to associate food with social support. The thesis also found that overweight/obese women who engage in ESRE need to be supported in both the acute and chronic dimensions of their ESRE behavior. Their acute needs include instrumental support for eating awareness in- the- moment as they are making food choices that could be ESRE, as well as in the form of a just-in-time distraction intervention to prevent them from engaging in ESRE. Their chronic needs include support for holistic goals and motivation to address their ESRE, emotional support for encouragement in weight loss efforts, and informational support for appraisal to understand ESRE and change thought patterns for lasting behavior change. This thesis suggests that HIT needs to allowi for more self-experimentation and tailoring opportunities. Finally, the thesis found that stress and self-blame contribute to ESRE behavior, and that the relative influence stress and self-blame had on ESRE differs by racial groups. This thesis suggests that HIT avoid content and design choices that may incite feelings of self-blame. The thesis’ contribution is that it fills a gap in the literature by using an interpretivist approach to understand the ESRE experiences of overweight/obese women, which permits insight into previously-undescribed aspects of the experience. Additionally, the thesis relates the lived experience of ESRE to HIT design, and highlights ESRE behavior in context of socioeconomic factors. It also makes the contribution of applying the concept of self-blame to a sample of overweight/obese women who are largely not diagnosed with an eating disorder. Finally, it explores how self-blame could be taken into account in the design of HIT for weight management.
... A possible intervention approach for adults living with obesity and EE is to educate individuals to understand and recognise EE, instead of restricting dietary intake which often exacerbates the issue [84]. Both CBT and Acceptance-based interventions may reduce EE episodes by recognizing stressors and resultant emotions and replacing the urge to eat with alternative positive actions [38,86]. CBT is a common approach to support change in EE by keeping a food and mood diary to identify the triggers and situational context of eating [87,88]. ...
Full-text available
Background: Emotional eating (EE) may be defined as a tendency to eat in response to negative emotions and energy-dense and palatable foods, and is common amongst adults with overweight or obesity. There is limited evidence regarding the effectiveness of interventions that address EE. Objectives: To synthesize evidence on the effectiveness of EE interventions for weight loss and EE in adults living with overweight or obesity. Methods: This is a systematic review and meta-analysis. Adhering to the PRISMA guidance, a comprehensive electronic search was completed up to February 2022. Random effects meta-analysis was carried out to determine the percentage change in weight and EE scores. Results: Thirty-four studies were included. The combined effect size for percentage weight change was −1.08% (95% CI: −1.66 to −0.49, I2 = 64.65%, n = 37), once adjusted for publication bias. Similarly, the combined effect size for percentage change in EE was −2.37%, (95% CI: −3.76 to −0.99, I2 = 87.77%, n = 46). Cognitive Behavioural Therapy showed the most promise for reducing weight and improving EE. Conclusions: Interventions to address EE showed promise in reducing EE and promoted a small amount of weight loss in adults living with overweight or obesity.
... The eating behavior is not only related to homeostatic reasons. In fact, an important factor that influences people's need and choice of food is represented by the emotional state (Carroll et al., 2013). In fact, it is commonly known that emotional states can have major effects on eating behavior. ...
Conference Paper
Overweight and obesity are the first leading risk related to nutrition for global deaths, in the last few years it outranked the famine. Obesity increases the risk of several debilitating, and deadly diseases, including diabetes, heart disease, and some cancers. Due to the many health risks associated with obesity, the financial burden that the treatment of this disease exercises on the European healthcare system is enormous. For this reason, the best strategy relies in prevention. In particular, the pervasiveness of technology can leverage an important advantage for the promotion of healthy behaviors in the new generations. This paper introduces PEGASO, a technological multidisciplinary project funded by the European Commission that aims at creating an ecosystem that can enable teenagers to adopt healthy habits leading to a healthy life-style. The ICT system plays an important role in the PEGASO ecosystem. This behavior change support system integrates a Virtual Individual Model that allows characterizing the physiological status, physical condition and the psychological status for each user. This allows the elaboration of tailored interventions aiming at promoting the adoption of healthy habits by the users. This paper describes this concept introducing the Virtual Individual Model and discusses the possible interventions related to the promotion of physical exercise and of healthy dietary habits. At the end of the paper, some indications about the future development of the PEGASO project are provided.
... Connecting a text messaging service with a person's calendar or social media apps could be one way of collecting information about user availability. Another way of making message delivery more timely would be through mood or activity tracking via mobile sensing [25]. By continuously monitoring a person's context and behavior through sensor streams (e.g., location, phone activity, motion), there is potential to infer how a person is feeling and how willing they may be to engage with an intervention [98,145]. ...
Full-text available
One-way text messaging services have the potential to support psychological wellbeing at scale without conversational partners. However, there is limited understanding of what challenges are faced in mapping interactions typically done face-to-face or via online interactive resources into a text messaging medium. To explore this design space, we developed seven text messages inspired by cognitive behavioral therapy. We then conducted an open-ended survey with 788 undergraduate students and follow-up interviews with students and clinical psychologists to understand how people perceived these messages and the factors they anticipated would drive their engagement. We leveraged those insights to revise our messages, after which we deployed our messages via a technology probe to 11 students for two weeks. Through our mixed-methods approach, we highlight challenges and opportunities for future text messaging services, such as the importance of concrete suggestions and flexible pre-scheduled message timing.
... The decision rule uses tailoring variables to identify the current state of vulnerability and specifies when it is appropriate to offer intervention [21]. Owing to substantial individual variability in what tailoring variables and at which thresholds indicate a heightened state of lapse risk, a machine learning algorithm informs the decision rule in this JITAI for lapses [6,[24][25][26]. In formative work to develop this JITAI, a supervised machine learning approach was used to train an algorithm using previously collected data on tailoring variables and dietary lapses. ...
Background Behavioral obesity treatment (BOT) is a gold standard approach to weight loss and reduces the risk of cardiovascular disease. However, frequent lapses from the recommended diet stymie weight loss and prevent individuals from actualizing the health benefits of BOT. There is a need for innovative treatment solutions to improve adherence to the prescribed diet in BOT. Objective The aim of this study is to optimize a smartphone-based just-in-time adaptive intervention (JITAI) that uses daily surveys to assess triggers for dietary lapses and deliver interventions when the risk of lapse is high. A microrandomized trial design will evaluate the efficacy of any interventions (ie, theory-driven or a generic alert to risk) on the proximal outcome of lapses during BOT, compare the effects of theory-driven interventions with generic risk alerts on the proximal outcome of lapse, and examine contextual moderators of interventions. Methods Adults with overweight or obesity and cardiovascular disease risk (n=159) will participate in a 6-month web-based BOT while using the JITAI to prevent dietary lapses. Each time the JITAI detects elevated lapse risk, the participant will be randomized to no intervention, a generic risk alert, or 1 of 4 theory-driven interventions (ie, enhanced education, building self-efficacy, fostering motivation, and improving self-regulation). The primary outcome will be the occurrence of lapse in the 2.5 hours following randomization. Contextual moderators of intervention efficacy will also be explored (eg, location and time of day). The data will inform an optimized JITAI that selects the theory-driven approach most likely to prevent lapses in a given moment. Results The recruitment for the microrandomized trial began on April 19, 2021, and is ongoing. Conclusions This study will optimize a JITAI for dietary lapses so that it empirically tailors the provision of evidence-based intervention to the individual and context. The finalized JITAI will be evaluated for efficacy in a future randomized controlled trial of distal health outcomes (eg, weight loss). Trial Registration NCT04784585; International Registered Report Identifier (IRRID) DERR1-10.2196/33568
One-way text messaging services have the potential to support psychological wellbeing at scale without conversational partners. However, there is limited understanding of what challenges are faced in mapping interactions typically done face-to-face or via online interactive resources into a text messaging medium. To explore this design space, we developed seven text messages inspired by cognitive behavioral therapy. We then conducted an open-ended survey with 788 undergraduate students and follow-up interviews with students and clinical psychologists to understand how people perceived these messages and the factors they anticipated would drive their engagement. We leveraged those insights to revise our messages, after which we deployed our messages via a technology probe to 11 students for two weeks. Through our mixed-methods approach, we highlight challenges and opportunities for future text messaging services, such as the importance of concrete suggestions and flexible pre-scheduled message timing.
In this scoping review, we aimed to summarize and analyze the latest research developments of persuasive design for healthy eating behavior and explore future design opportunities. This paper initially collected 1231 papers from 2011 to 2021 in three different databases: the Association for Computing Machinery (ACM) digital library, IEEE Xplore and SpringerLink databases. Based on a selection process, 28 papers that mainly focused on addressing dietary health by persuasive designs were eventually included in final analysis. These 28 papers were sorted by three characteristics: research specifications, methodologies, and design rationales. Our data analyses revealed that the reviewed papers primarily utilized persuasive technologies for eating behaviors monitoring, recording, and healthy eating suggestion. Moreover, six types of design applications were commonly implemented, including mobile applications, persuasive messages, digital products and service systems, wearable devices, chatbots/assistants, and public devices. Our review showed that persuasive design, as a generic approach for promoting healthy eating, lacked research investigations on personalized solutions for particular user groups such as office workers and teenagers. Future works could explore persuasive design strategies by applying the research factors of user experience and examining the efficacy of persuasive technology tools to effectively promote healthy eating behaviors for various user groups in different contexts.
Full-text available
Deviant eating behavior such as skipping meals and consuming unhealthy meals has a significant association with mental well-being in college students. However, there is more to what an individual eats. While eating patterns form a critical component of their mental well-being, insights and assessments related to the interplay of eating patterns and mental well-being remain under-explored in theory and practice. To bridge this gap, we use an existing real-time eating detection system that captures context during meals to examine how college students’ eating context associates with their mental well-being, particularly their affect, anxiety, depression, and stress. Our findings suggest that students’ irregularity or skipping meals negatively correlates with their mental well-being, whereas eating with family and friends positively correlates with improved mental well-being. We discuss the implications of our study in designing dietary intervention technologies and guiding student-centric well-being technologies.
Full-text available
Background Despite their reliability- and validity-related challenges, self-reports remain the most common data collection method in nutrition research, food-related consumer and marketing research. The rapid development of technology has nevertheless inspired attempts to overcome the challenges of self-reports by applying technological solutions capable of capturing objective data. Scope and approach We reviewed objective measurement technologies applicable in nutrition research, food-related consumer and marketing research, spanning the continuum from food-evoked emotions to food choice and dietary intake. Focusing on non-invasive solutions, we categorised identified technologies according to five study domains: 1) detecting food-related emotions, 2) monitoring food choices, 3) detecting eating actions, 4) identifying the type of food consumed, and 5) estimating the amount of food consumed. Additionally, we considered technologies not yet applied in the targeted research disciplines but worth considering in future research. Key findings and conclusions Within each domain, several variables have been measured using diverse technologies or combinations of technologies. These technologies cover wearable and remotely applied solutions that collect data on the individual or provide indirect information on consumers’ food choices or dietary intake. The key challenges of the reviewed technologies concern their applicability in real-world settings; capabilities to produce accurate, reliable, and meaningful data with reasonable resources; participant burden, and privacy protection. We provide recommendations for researchers and practitioners in nutrition research, food-related consumer and marketing research to work around the key challenges. For fruitful use of available technologies, we encourage collaboration between technology developers and experts in nutrition, consumer, and marketing sciences.
Full-text available
The COVID-19 related lockdown made it much more difficult for people to control their eating behaviours and body weight with the methods and means they had used before. This is reflected in reports that show that eating behaviours deteriorated significantly during the COVID-19 pandemic (including in Poland). Therefore, it is important to determine what factors may be conducive to healthy eating behaviours among people with different BMI. As previous studies show, the use of healthy eating related-apps and training programs may be a protective factor against the development of unhealthy eating behaviours. Therefore, it is worth checking whether their action will be a protective factor during COVID-19. The aim of this cross sectional study was to analyse whether the current use of healthy eating-related apps and previous participation in training in this field (educational activities) as well as body mass index may play a role in eating motives and behaviours among women during COVID-19. Our final sample included 1,447 women (age: M = 31.34 ± 11.05). Participants completed: the Eating Motivation Survey, the Emotional Overeating Questionnaire, the Mindful Eating Questionnaire, socio-demographic survey and questions about healthy eating-related apps and training (educational activities). Referring to the selected significant results, our study shows that during COVID-19, the use of healthy eating-related apps alone, as well as the use of apps and prior training participation promote healthy eating motives and behaviours. It suggests that promoting the use of healthy eating applications and the acquisition of knowledge and skills in this field could be one way of shaping resources that can be effectively used to deal with crisis situations. PLOS ONE PLOS ONE |
Full-text available
We present AffectAura, an emotional prosthetic that allows users to reflect on their emotional states over long periods of time. We designed a multimodal sensor set-up for continuous logging of audio, visual, physiological and contextual data, a classification scheme for predicting user affective state and an interface for user reflection. The system continuously predicts a user's valence, arousal and engage-ment, and correlates this with information on events, communications and data interactions. We evaluate the interface through a user study consisting of six users and over 240 hours of data, and demonstrate the utility of such a reflection tool. We show that users could reason forward and backward in time about their emotional experiences using the interface, and found this useful.
Full-text available
Factor-analytic evidence has led most psychologists to describe affect as a set of dimensions, such as displeasure, distress, depression, excitement, and so on, with each dimension varying independently of the others. However, there is other evidence that rather than being independent, these affective dimensions are interrelated in a highly systematic fashion. The evidence suggests that these interrelationships can be represented by a spatial model in which affective concepts fall in a circle in the following order: pleasure (0), excitement (45), arousal (90), distress (135), displeasure (180), depression (225), sleepiness (270), and relaxation (315). This model was offered both as a way psychologists can represent the structure of affective experience, as assessed through self-report, and as a representation of the cognitive structure that laymen utilize in conceptualizing affect. Supportive evidence was obtained by scaling 28 emotion-denoting adjectives in 4 different ways: R. T. Ross's (1938) technique for a circular ordering of variables, a multidimensional scaling procedure based on perceived similarity among the terms, a unidimensional scaling on hypothesized pleasure–displeasure and degree-of-arousal dimensions, and a principal-components analysis of 343 Ss' self-reports of their current affective states. (70 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Full-text available
A plethora of studies have examined the efficacy and effectiveness of cognitive-behavioral therapy (CBT) for adult anxiety disorders. In recent years, several meta-analyses have been conducted to quantitatively review the evidence of CBT for anxiety disorders, each using different inclusion criteria for studies, such as use of control conditions or type of study environment. This review aims to summarize and to discuss the current state of the evidence regarding CBT treatment for panic disorder, generalized anxiety disorder, social anxiety disorder, obsessive-compulsive disorder, and post-traumatic stress disorder. Overall, CBT demonstrates both efficacy in randomized controlled trials and effectiveness in naturalistic settings in the treatment of adult anxiety disorders. However, due to methodological issues, the magnitude of effect is currently difficult to estimate. In conclusion, CBT appears to be both efficacious and effective in the treatment of anxiety disorders, but more high-quality studies are needed to better estimate the magnitude of the effect.
Conference Paper
We give a basic introduction to Gaussian Process regression models. We focus on understanding the role of the stochastic process and how it is used to define a distribution over functions. We present the simple equations for incorporating training data and examine how to learn the hyperparameters using the marginal likelihood. We explain the practical advantages of Gaussian Process and end with conclusions and a look at the current trends in GP work.
Conference Paper
Researchers in HCI and behavioral science are increasingly exploring the use of technology to support behavior change in domains such as health and sustainability. This work, however, remain largely siloed within the two communities. We begin to address this silo problem by attempting to build a bridge between the two disciplines at the level of behavioral theory. Specifically, we define core theoretical terms to create shared understanding about what theory is, discuss ways in which behavioral theory can be used to inform research on behavior change technologies, identify shortcomings in current behavioral theories, and outline ways in which HCI researchers can not only interpret and utilize behavioral science theories but also contribute to improving them.
Conference Paper
Many studies have found text messaging to be a promising medium for healthcare delivery. However, since the studies that successfully used text messages for encouraging physical activity were all short term (10 days to 3 weeks) and conducted with a small sample (n≤15), we do not know if people will not be motivated by these technologies after the novelty effect dies. In this paper, we present the results from a study conducted for a longer term (3 months) with a larger sample size (n=28) to discover if text messages are effective for encouraging physical activity once the novelty effect of the technology wears off. We chose a population of young adults (age 18–24) as they are one of the heaviest users of text messages. Measures of analysis included number of steps, message ratings, level of motivation and interviews. Our findings suggest that text messages are a good way for encouraging physical activity in young adults, even after the novelty effect wears off.
In this paper, various on-body sensors can gather vital information about an individual's food intake. Such data can both help weight-loss professionals personalize programs for clients and inform nutrition research on eating behaviors.
Predicting when a person might be frustrated can provide an intelligent system with important information about when to initiate interaction. For example, an automated Learning Companion or Intelligent Tutoring System might use this information to intervene, providing support to the learner who is likely to otherwise quit, while leaving engaged learners free to discover things without interruption. This paper presents the first automated method that assesses, using multiple channels of affect-related information, whether a learner is about to click on a button saying “I’m frustrated.” The new method was tested on data gathered from 24 participants using an automated Learning Companion. Their indication of frustration was automatically predicted from the collected data with 79% accuracy (chance=58%). The new assessment method is based on Gaussian process classification and Bayesian inference. Its performance suggests that non-verbal channels carrying affective cues can help provide important information to a system for formulating a more intelligent response.
We report on a controlled trial of three structured writing paradigms that engage the writer with cognitive-behavioural emotion-processes: exposure, devaluation, and benefit-finding. University students (N = 198) wrote once a week for three weeks about their most upsetting experience. The long-term effects of these structured writing procedures were compared to an unstructured emotion writing condition and control. Outcomes indicated that exposure writing sped the reduction of intrusive and avoidant symptoms, while benefit-finding writing increased reports of positive growth. Results suggest the use of these paradigms to study emotion-processing mechanisms and, potentially, in practice to enhance coping in process-specific ways.