PreprintPDF Available

Effect of Values and Technology Use on Exercise: Implications for Personalized Behavior Change Interventions

Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Technology has recently been recruited in the war against the ongoing obesity crisis; however, the adoption of Health & Fitness applications for regular exercise is a struggle. In this study, we present a unique demographically representative dataset of 15k US residents that combines technology use logs with surveys on moral views, human values, and emotional contagion. Combining these data, we provide a holistic view of individuals to model their physical exercise behavior. First, we show which values determine the adoption of Health & Fitness mobile applications, finding that users who prioritize the value of purity and de-emphasize values of conformity, hedonism, and security are more likely to use such apps. Further, we achieve a weighted AUROC of .673 in predicting whether individual exercises, and we also show that the application usage data allows for substantially better classification performance (.608) compared to using basic demographics (.513) or internet browsing data (.546). We also find a strong link of exercise to respondent socioeconomic status, as well as the value of happiness. Using these insights, we propose actionable design guidelines for persuasive technologies targeting health behavior modification.
Content may be subject to copyright.
Effect of Values and Technology Use on Exercise:
Implications for Personalized Behavior Change Interventions
Yelena Mejova and Kyriaki Kalimeri
ISI Foundation, Turin, Italy,
Technology has recently been recruited in the war against the ongo-
ing obesity crisis; however, the adoption of Health & Fitness appli-
cations for regular exercise is a struggle. In this study, we present a
unique demographically representative dataset of 15k US residents
that combines technology use logs with surveys on moral views,
human values, and emotional contagion. Combining these data, we
provide a holistic view of individuals to model their physical exer-
cise behavior. First, we show which values determine the adoption
of Health & Fitness mobile applications, finding that users who
prioritize the value of purity and de-emphasize values of conformity,
hedonism, and security are more likely to use such apps. Further, we
achieve a weighted AUROC of .673 in predicting whether individual
exercises, and we also show that the application usage data allows
for substantially better classification performance (.608) compared
to using basic demographics (.513) or internet browsing data (.546).
We also find a strong link of exercise to respondent socioeconomic
status, as well as the value of happiness. Using these insights, we
propose actionable design guidelines for persuasive technologies
targeting health behavior modification.
Health; Exercise; Technology Use; Mobile; Moral Values
ACM Reference Format:
Yelena Mejova and Kyriaki Kalimeri. 2019. Effect of Values and Technology
Use on Exercise: Implications for Personalized Behavior Change Interven-
tions. In 27th Conference on User Modeling, Adaptation and Personalization
(UMAP ’19), June 9–12, 2019, Larnaca, Cyprus. ACM, New York, NY, USA,
10 pages.
Over the last half-century, the daily occupation-related energy expen-
diture of US workers has decreased by more than 100 calories, with
the proportion of jobs requiring at least moderate intensity physical
activity declining from 48% to 20% in 2008 [
]. Sedentary lifestyle
– one that includes TV watching and gaming and lacks vigorous
exercise – has also been found to be strongly related to the childhood
obesity epidemic [5].
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for components of this work owned by others than the
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specific permission
and/or a fee. Request permissions from
UMAP ’19, June 9–12, 2019, Larnaca, Cyprus
© 2019 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ACM ISBN 978-1-4503-6021-0/19/06. . . $15.00
Technology has been an essential factor in changing lifestyles.
Over the years, TV has been seen as a replacement of physical activ-
ity, a channel for advertisement of nutrient-poor food, and increasing
prevalence of “mindless” eating [
]. The arrival of mobile tech-
nology has been shown to contribute to the sedentary behaviors [
However, this has not stopped both entrepreneurs and public health
officials from attempting to use new technology to encourage behav-
ior change. Unfortunately, within the industry, user retention is an
ongoing struggle, with 62% of mHealth app publishers report digital
solutions with less than 1,000 monthly active users [
]. Indeed, the
latest research shows a complex relationship between psychology,
technology use, and exercise. Mediation analyses find that increased
physical activity associated with health app use is related to feelings
of self-efficacy [
], with yet other insights linking exercise to being
extroverted, neurotic, and less agreeable [
], as well as having
implications for mental health [
]. For a better understanding of
the relationship between psychology, technology, and exercise, it
is necessary to model users of new technologies [
] and to design
effective health interventions.
This study is a unique view of the interaction between technol-
ogy use, demographics, and value systems of a representative US
population sample, allowing for rich user modeling in the aims of
promoting exercise. Just over 15k participants filled in the ques-
tionnaires including the following psychometric measures: Moral
Foundations [
], Schwartz Basic Human Values [
], and Emo-
tional Contagion [
]. Along with these, 5,008 respondents agreed
to allow the capture of their desktop browsing data, whereas another
2,625 allowed to capture their mobile app usage. Using this data, we
contribute the following insights on psychological markers of health
app use and the actual exercise behavior:
Modeling “Health & Fitness” application use
in relation
to psychometric and demographic variables, we find a marked
difference in application usage between the two genders, as
well as significant negative relationship between the values of
tradition,conformity,hedonism, and security, while positive
for purity.
Predicting the engagement in physical exercise
via the
above variables, as well as browsing and application use data,
we show a marked increase in classification performance from
baseline demographic model with the addition of psychome-
tric features, as well as application usage data, but with a
smaller contribution of desktop browsing data.
Revealing determinants of exercise
among the types of vari-
able, confirming a significant effect of education and wealth
on healthy behaviors [
], as well as showing significant re-
lationships with the view that health is a choice, positive
arXiv:1903.11579v1 [cs.HC] 27 Mar 2019
UMAP ’19, June 9–12, 2019, Larnaca, Cyprus Mejova & Kalimeri
association with happiness emotional contagion and stimula-
tion value, with downloading a Health & Fitness app being
another strong predictor.
Comparing the exercise behaviors across mobile applica-
, we show those tracking a particular kind of exercise,
such as running or cycling, are associated with more users
reporting exercising regularly, than those for general health
tracking or women’s health tracking.
We conclude with concrete suggestions of employing this knowl-
edge in the design, personalization, and deployment of technologies
for an effective lifestyle change and health outcomes intervention.
Technology is now commonly used to monitor behavior and physical
activity [
]. Digital data from smartphones were initially used
as simple activity monitoring sensors [
]; however, over the last
years, their integration with user-generated content led to more
sophisticated personalized interventions aiming at motivating the
users to increase their physical activity level and encouraging a
healthier lifestyle [
]. Researchers have tried apps with different
messaging strategies [
], personalized exercise recommendation
], as well as utilizing machine learning via supervised learning
] and reinforcement learning [
]. Others help users
find exercise partners [
], provide educational materials [
and emotional support [
] (see [
] for a recent survey on
personalized health interventions).
To understand the impact of such interventions, researchers exam-
ined the role of individual characteristics, attitudes, and lifestyle of
users [
], demographic attributes such as gender, age, socioe-
conomic factors, and technology literacy [
]. For instance,
using self-reported technology use, [
] found that two-thirds of
their participants were using a smartphone. This subset was younger,
more likely to have a university degree with higher socioeconomic
status, and was more likely to engage physical activity. A substantial
proportion of their population was not engaged in Health & Fitness
apps; however, those who were were more motivated to change or
maintain a healthy lifestyle. A further association was confirmed
between smartphone use and health literacy [
], and an association
with age, with seniors (65 years and older) using digital health at
much lower (but steadily increasing) rates [
]. Unlike these
previous studies relying on self-reported data based on surveys, our
data present a snapshot of desktop and mobile use which provides
valuable ground truth for tech-related behaviors (as well as a com-
plementary rich demographic baseline).
Despite being an active research direction, the consideration of
the psychological aspects of the individual such as personal views,
values, and emotional states, in tech-driven intervention remains
hugely unexplored [
]. Human values are known to influence
people’s actions [
] but received little attention in studying their
relationship with attitudes regarding healthy lifestyles. Regarding
human and moral values in sports, Lee et al. [
] examined the
value-expressive function of attitudes and achievement goal theory in
predicting the moral attitudes of young athletes. Ball et al. [
] studied
the individual preferences for social support according to the values
system in following a healthy lifestyle. Apart from these studies,
human values were only considered in cases where an individual
deviated from a normative of healthy lifestyle [
] such as depression
] or mental disorders [
]. Among works closest to our intention,
Lathia et al. [
] assessed the relationship between physical activity
and happiness via a smartphone app concluding that people that
exercise more are happier. In this study, we contribute a unique
combination of psychometric measures, spanning morals, values,
and emotional contagion, to better understand technology use and
engagement in physical activity.
Data presented in this study spans 15,021 subjects in the United
States of America, selected using probabilistic, representative sam-
pling methodology, all of whom were incentivised to participate.
After receiving informed consent from all participants for the col-
lection, storage, and analysis of the data, as well as the acceptance
of the privacy policy
, we administered a series of questionnaires
to gather demographic and psychometric data. Also, we asked the
participants for the access to either their basic mobile or desktop data
for one month, resulting in desktop activity data for 5,008 people
(2,823 women) and mobile activity data for 2,625 people (1,544
women). The latter subset with activity data has been discussed in
[30]. Below we describe the data collected and used in this study2.
3.1 Demographics
The intake survey covered basic demographic factors (age, gender,
ethnicity), geographic factors (home location, expressed at the zip
code level), socioeconomic factors (educational level, marital status,
parenthood, wealth, income), health-related factors (exercise, smoke,
and weight issues) and political orientation. Table 1 presents the
complete list of the demographic information gathered, along with
the respective range of values for all the 15,021 participants.
3.2 Psychometric Measures
Moral Foundations. To measure the values of the participants,
we employ the Moral Foundation Theory [
] which we opera-
tionalized via the Moral Foundations Questionnaire (MFQ) [
], a
validated measure of the degree to which individuals endorse each
of five dimensions:
care/harm, basic concerns for the suffering of others, includ-
ing virtues of caring and compassion;
fairness/cheating, concerns about unfair treatment, inequality,
and more abstract notions of justice;
loyalty/betrayal, concerns related to obligations of group
membership, such as loyalty, self-sacrifice, and vigilance
against betrayal;
authority/subversion, concerns related to social order and
the obligations of hierarchical relationships like obedience,
respect, and proper role fulfillment;
purity/degradation, concerns about physical and spiritual con-
tagion, including virtues of chastity, wholesomeness, and
control of desires.
For privacy considerations, the data will be made available upon request, exclusively
for the scientific community.
Effect of Values and Technology Use on Exercise UMAP ’19, June 9–12, 2019, Larnaca, Cyprus
Table 1: Complete list of the demographic attributes collected and their respective ranges for the entire sample of 15,021 participants.
Attribute Demographic Variables Sample size Attribute Demographic Variables Sample size
Range (N=15,021)Range (N=15,021)
Age 18-24 1,636 (10.8%) Political Party Democrat 6,227 (41.4%)
25-34 2,583 (17.1%) Republican 4,455 (29.6%)
35-49 3,770 (25%) Libertarian 429 (2.8%)
50-54 1,642 (10.9%) Independent 3,910 (26%)
55-64 2,707 (18%)
65+ 2,683 (17.8%)
Education College Graduate 4,854 (32.3%) Wealth 50k or less 5,520 (36.7%)
Post Graduate 3,409 (22.6%) 50k-100k 2,087 (13.8%)
Some College 3,810 (25.3%) 100k-250k 2,375 (15.8%)
High-school 1,832 (12.1%) 250k-500k 2,166 (14.4%)
Trade School 949 (6.3%) 500k-1000k 1,627 (10.8%)
1000k or more 1,246 (8.2%)
Ethnicity Asian 669 (4.4%) Weight Issues No 8,709 (57.9%)
African American 1,761 (11.7%) Yes 6,312 (42%)
White 11,042 (73.5%)
Hispanic 1,335 (8.8%)
Exercise No 6,631 (44.1%) Parent No 5,613 (37.4%)
Yes 8,390 (55.8%) Yes 9,408 (62.6%)
Gender Female 8,409 (55.9%) Smoker No 13,150 (87.5%)
Male 6,612 (44.1%) Yes 1,871 (12.4%)
Income 20k or less 1,384 (9.2%) Marital Status Divorced 1,409 (9.3%)
20k-30k 1,389 (9.2%) Single 3,509 (23.3%)
30k-50k 2,785 (18.5%) Married 8,037 (53.5%)
50k-75k 3,246 (21.6%) Living Together 1,444 (9.6%)
75k-100k 2,601 (17.3%)
100k-150k 2,386 (15.8%) High Blood No 12,025 (80%)
150k-200k 745 (4.9%) Pressure Yes 2,996 (20%)
200k or more 485 (3.2%)
The questionnaire is based on self-assessment evaluations and
consists of 30 items, resulting in a unique numerical value from 0-30
per person. According to the MFQ, six items (on a 6-point Likert
scale) per foundation were averaged to produce the individuals’
scores on each of the five foundations.
Schwartz Basic Human Values. We assess the Schwartz human
values employing the Portrait Values Questionnaire [
], whose
validity across cultures is validated in studies performed on 82 coun-
tries and samples belonging to highly diverse geographic, cultural,
linguistic, religious, age, gender, and occupational groups. The ques-
tionnaire is based on self-assessments resulting in a numerical value
per person for each of the ten basic values:
self-direction, independent thought, action-choosing, creating,
stimulation, need for variety and stimulation to maintain an
optimal level of activation;
hedonism, related to organismic needs and the pleasure asso-
ciated with satisfying them;
achievement, personal success through demonstrating compe-
tence according to social standards;
power, the attainment or preservation of a dominant position
within the more general social system;
security, safety, harmony, and stability of society, of relation-
ships, and self;
conformity, restraint of actions, inclinations, and impulses
likely to upset or harm others and violate social expectations
or norms;
tradition, symbols and practices or groups that represent their
shared experience and fate;
benevolence, concern for the welfare of close others in every-
day interaction;
universalism, this value type includes the former maturity
value type, including understanding, appreciation, tolerance,
and protection for the welfare of all people and nature.
UMAP ’19, June 9–12, 2019, Larnaca, Cyprus Mejova & Kalimeri
The questionnaire is based on self-assessment evaluations on
a 7-point Likert scale. Following [
], we average the respective
items per value, and we account for individual differences. The
above ten values can be clustered into four higher order values,
so-called quadrant values and into two dimensions, as the sum of
the individual items of which they consist: Openness to change
(self-direction, stimulation) vs. Conservation (security, conformity,
tradition) and Self-enhancement (universalism, benevolence) vs. Self-
transcendence (power, achievement). Therefore, the first dimension
captures the conflict between values that emphasize the indepen-
dence of thought, action, and feelings and readiness for change and
the values that highlight order, self-restriction, preservation of the
past, and resistance to change. The second dimension captures the
conflict between values that stress concern for the welfare and inter-
ests of others and values that emphasize the pursuit of one’s interests
and relative success and dominance over others. Hedonism shares
elements of both openness to change and self-enhancement.
Emotional Contagion. Emotional contagion is the phenomenon
that individuals tend to feel emotions, such as happiness, or sadness,
triggered by the feelings expressed by the people with whom they
interact [
]. In this study, we employ the well-established emotional
contagion scale (EC) [
]. The 15-item questionnaire is based on
self-assessment evaluations on a 5-point Likert scale. It assesses
mimetic tendency to five basic emotions (love, happiness, fear, anger,
and sadness), measuring the individual differences in susceptibility
to “catching” and empathizing the emotions of others.
3.3 Digital Data
Desktop Browsing Data. For the participants who permitted the
logging of their desktops’ web browsing data, 5,008 in total, we
capture: (i) the domain names, and (ii) the average time spent online
and (iii) the number of visits per day on each domain. All this
information is aggregated by day, and only the domain names (and
not the page or section of the websites) are stored, to ensure the
privacy of the participants. Users with fewer than
domains are discarded. We then assign to each domain name a
category label according to its content [57].
Mobile Data. Participants are also asked to download an appli-
cation which, upon agreement with the privacy policy, logs their
web browsing activity and application usage, and 2,625 agreed to be
Application Data. Application usage was captured whenever
the application was running in the “foreground”. Foreground
usage means an application is open on someone’s device, re-
gardless of whether the application is currently being engaged
with or not. Application usage data for each participant in-
cluded records of the date and time stamp, the local time zone,
and time spent on the application (in seconds). Moreover, we
assigned to each application the category label provided by
the Google Play Store3.
Mobile browsing Data. URL data was captured from the
native browser on the subject’s device (not any 3rd party
browsers). URLs for both secure and non-secure traffic were
The assignment was performed parsing the application data from the Google Play
Store using the following project.
captured, though only the URL domain was stored in consid-
eration of privacy. Similar to the desktop browsing data, users
with a number of visits fewer than
unique domains
are discarded from the analysis leaving us with a total of
2,406 participants.The domains are classified as above for the
Desktop users [57].
Noteworthy is the fact that “Mobile” and “Desktop” browsing
data provide the same information, they only express different modes
of web navigation, i.e. mobile vs desktop. See [
] for a detailed
description of the data.
We begin by examining the usage of mobile applications (apps) in
the Health & Fitness category. These apps include those associated
with particular wearables like Fitbit and Garmin Connect, activity
trackers like MapMyRun,RunKeeper,Nike+ Run Club, and weight
management including Lose It! and WW (Weight Watchers). The only
demographic variable associated with getting such an app (more
precisely, opening it at least once in the time of observation), is
gender, with females 45% more likely to get one than males. Though
the gender division is not evenly distributed across applications, with
those marketed for tracking running activity (such as RunKeeper
and MapMyRun) being 24.9% more likely to be adopted by females,
whereas those for walking (Walkroid and MapMyWalk) are 101%
more likely (that is, twice as likely). The distinction is even greater
for weight loss applications, with females 168% more likely to adopt
one than males. Notably, these gender distinctions have not been
revealed in recent surveys [19, 33].
Considering the psychometric attributes, we run a linear model
) to predict the adoption of any Health & Fitness appli-
cation, with the coefficients plotted in Figure 1, whiskers marking
95% confidence intervals, and those significant at
in green. We find a negative relationship with values associated
with tradition,security,hedonism, and conformity, as well as with
concerns about hypocrisy increasing in the society, and a positive
relationship with purity value. Emotional contagion results show a
positive relationship with sadness but a negative with fear. These
trends point to people less concerned about societal traditions, who
are less influenced by caution or fear, and those striving towards
physical or spiritual purity.
Finally, we ask which health applications are most associated
with self-reported exercise (defined in more detailed in the following
section). To answer this question, we consider all applications in the
Health & Fitness category having at least ten users in our dataset
and compute the proportion of such users who self-report exercising.
The top 30 apps are shown in Figure 2, along with the number of
users the proportion is based on.
Running, cycling, and walking tracking applications dominate
the top, as well as Spark People, which provides a combination of
weight loss and fitness (although the proportion is based on eleven
respondents). Towards the bottom, we find generic health resources
like WebMD, as well as pregnancy apps (I’m Expecting). Thus, we
observe applications with an explicit activity to perform are better at
supporting regular exercise than, say, more generic pedometers or
health trackers.
Effect of Values and Technology Use on Exercise UMAP ’19, June 9–12, 2019, Larnaca, Cyprus
−4 −3 −2 −1 0 1
Figure 1: Coefficients of linear model predicting Health & Fit-
ness app use, with those significant at p<0.05 bolded in green.
Thus, we find value determinants in our study subjects’ willingness
to use the health applications, but we are interested in whether
such knowledge would help understand the step to exercise. Indeed,
we find that respondents who have downloaded such applications
are 42% more likely to say they exercise (
). However,
as we illustrate in the following sections, exercise behavior has a
multifaceted nature beyond health app usage.
5.1 Demographics of Exercise
In this study, we operationalize health-related activities of the user
via the questionnaire, mainly the reply to question “I exercise regu-
larly”, to which a binary yes/no reply is allowed. As a self-declared
assessment of action, the variable suffers from the biases endemic to
surveys, including acquiescence bias (tendency to reply positively),
social desirability bias (tendency to reply in line with perceived
expectations), and faulty recall. Participants may also have a unique
understanding of the frequency of exercise which may be consid-
ered “regular”. Since we are interested in comparing participants
within the study, we make an implicit assumption that the biases and
individual noise are uniformly distributed through the population
(more on this limitation in the Discussion section). In our data, 8,390
(55.8%) of respondents indicated they exercised regularly, the rest –
We begin by examining the basic demographic characteristics
of the two groups, shown in Figures 3. As the plots show the 95%
confidence intervals, we can discern some statistically significant
differences in the two groups. Mostly, we see no significant age
differences, except for in 34-49 range, when it is slightly more likely
that the users do not exercise. Similarly, there are slightly more males
indicating that they exercise than females. Education and income
prove to be a more discerning feature, with college graduates and
post-graduates exercising markedly more, and high-school graduates
I'm Expecting
Noom Coach
Lose It!
P Tracker
Instant Heart Rate
Noom Walk
WW Mobile
Express Scripts
Spark People
0.00 0.25 0.50 0.75 1.00
Figure 2: Applications in the Health & Fitness category ordered
by the proportion of respondents reporting exercising regularly,
with the nshown in white.
less. Similarly, the higher income individuals (household income of
$70k or more) reported exercising markedly more than those in the
lower brackets ($50k or less). A similar observation can be made
for the wealth variable, with those having a net worth of less than
$50k reporting to be exercising markedly less than those having
over $250k (plot omitted for brevity). These findings echo earlier
observed tendency of those in higher income stratum to engage in
higher levels of physical activity (as has been described in a literature
review [29] and later measured using accelerometers [55]).
5.2 Predicting Exercise
Next, we would like to determine whether it is possible to use this
data to predict engagement in physical activity. We formulate this
study as a supervised classification problem, aiming at predicting
whether participants exercise. We assess the predictive power of:
Demographics (ethnicity removed due to sparsity)
Psychometry: Moral Foundations and Schwartz Basic Human
Values, and Emotional Contagion
Health-related variables: including replies to survey questions
explicitly about health
Web domain categories: for both desktop and mobile users
Application categories: for mobile users only
Rate of usage of Health & Fitness applications
Note that we chose to single out the demographic and survey
variables having to do with health and health-related attitudes into
their category, as they tend to be highly correlated with exercise.
Focusing only on our “Mobile” dataset for which we have all
the above information (n=2620), we train a Random Forest (RF)
classifier [
] inferring each time from a richer set of predictors as
presented in Table 2. The choice of the classifier is motivated by
its ability to deal with the sparse web browsing activity data in our
UMAP ’19, June 9–12, 2019, Larnaca, Cyprus Mejova & Kalimeri
18−2425−3434−4950−5455−64 65+
female male
< High School
High School
Some College
Trade/Prof School
College Grad
Figure 3: Demographic distribution of respondents, broken down by whether they indicated they exercise, with 95% conf. intervals.
Table 2: Performance of random forest models predicting whether participant indicated exercise, measured using weighted AUROC,
along with p-value of significance in difference with Experiment 1 (Basic demographics).
Weighted Significance
Features included AUROC p-value
1. Basic demographics (only gender and age) .513 -
2. Advanced demographics .616 <0.0001
3. Advanced demographics + values/morals .623 <0.0001
4. Advanced demographics + values/morals + value health .650 0.0001
5. Advanced demographics + values/morals + value health + domains cat-s .654 <0.0001
6. Advanced demographics + values/morals + value health + app cat-s .671 <0.0001
7. Advanced demographics + values/morals + value health + domains cat-s + app cat-s .672 <0.0001
8. Advanced demographics + values/morals + value health + domains cat-s + app cat-s + H&F Time .673 <0.0001
9. Domains cat-s .546 0.07
10. Advanced demographics + domains cat-s .646 0.0002
11. App cat-s .608 0.001
12. Advanced demographics + app cat-s .646 <0.0001
13. Values/morals .575 0.002
14. Values/morals + value health .618 0.0001
dataset, and its performance in previous studies. We perform ten-
fold cross-validation procedure and report the average Area Under
the Receiver Operating Characteristic Curve (AUROC) weighted
statistic over all folds. In the last column of the table, we report
the statistical significance obtained by comparing the performance
of each model to the basic demographics baseline. Note that the
random baseline would achieve a weighted AUROC of .50 for all
experiments. For all experiments, our data fusion policy consisted of
“early” fusion at a feature level, concatenating the different feature
vectors for each respondent.
We begin by attempting to predict exercise inferring only on
the most basic demographic attributes – gender and age (Ex. 1 in
Table 2), finding performance to be not much beyond the random
baseline. However, adding richer demographic attributes, such as
wealth, income, and educational level, significantly improved our
prediction to 0.616 (at p<0.001).
Enhancing the baseline model with information about the moral
values (Ex. 3), we note an improvement in the performance. The
increase is even more pronounced - as expected - with the inclusion
of the health-related variables (Ex. 4). Adding web browsing domain
categories (Ex. 5) slightly improved the model, but it is when includ-
ing the categories of the applications used (Ex. 6) that the accuracy
is increased in a notable way to 0.671. Including both sets of fea-
tures - apps and domains categories - (Ex. 7), as well as the average
time people used a Health & Fitness application (Ex. 8) performs
statistically identical to Ex. 6, indicating that the mere knowledge of
application being opened once is enough.
Examining the predictive power of each variable type, we find
internet browsing domain categories to be the least useful (Ex. 9,10),
followed by application usage (Ex. 11,12), and the most valuable
(although also most difficult to obtain) the moral values, including
those about health (Ex. 13,14) and advanced demographics (Ex. 2)
including wealth and income.
Thus, we illustrate the utility of value beliefs in modeling ex-
ercise, which combined with demographics and technology usage
substantially outperform the baseline demographics model.
Effect of Values and Technology Use on Exercise UMAP ’19, June 9–12, 2019, Larnaca, Cyprus
Table 3: Logistic regression models predicting exercise using demographic (D), values (V), health (H), url domain (U), and app
(A) features, applied to users who shared PC activity (P), mobile activity (M) or neither (N). Only features significant at p<0.01
level shown (before Bonferroni adjustment), alongside their coefficient estimate and their corresponding p-values (now Bonferroni-
adjusted). Confidence levels: p<0.001 ***, p<0.01 **, p<0.05 *.
D (N+P+M) D+V (N+P+M) D+V+H (N+P+M) D+V+H+U (P) D+V+H+U (M) D+V+H+A (M) D+V+H+U+A (M)
n=15021 n=15021 n=15021 n=4995 n=2260 n=2260 n=2260
M F =0.031 R2
M F =0.047 R2
M F =0.074 R2
M F =0.704 R2
M F =0.870 R2
M F =0.868 R2
M F =0.875
(Intercept) -0.6837 *** -1.4740 *** 0.1931 -0.0914 1.1420 0.6129 1.0410
education 0.1334 *** 0.1372 *** 0.1164 *** 0.1313 *** 0.1243 0.1294 0.1239
gender -0.0161 -0.0058 -0.0819 -0.0898 0.1450 0.0729 0.0981
income 0.0561 *** 0.0626 *** 0.0442 * 0.0761 0.0833 0.0729 0.0696
parent -0.1035 -0.1112 -0.1091 -0.0265 -0.3953 -0.3543 -0.4137
wealth 0.1745 *** 0.1709 *** 0.1518 *** 0.1387 *** 0.1714 ** 0.1453 * 0.1743 **
age -0.0795 *** -0.0858 *** -0.0708 *** -0.0630 -0.0441 -0.0052 -0.0126
marital_status_married -0.0766 -0.0712 -0.0793 -0.3038 -0.1720 -0.2266 -0.1229
political_party_vote 0.0253 0.0385 0.0304 0.0396 0.0403 0.0465 0.0507
EC_Happiness 0.0541 *** 0.0162 0.0063 0.0420 0.0479 0.0331
EC_Sadness -0.0269 -0.0207 0.0086 -0.0698 -0.0665 -0.0689
MFT_authority -0.0089 -0.0169 -0.0015 -0.0071 -0.0084 -0.0105
MFT_loyalty 0.0160 0.0244 *** 0.0144 -0.0043 0.0007 -0.0027
SWV_achievement -0.0594 -0.0594 -0.0442 -0.0788 -0.0675 -0.0776
SWV_benevolence -0.1262 * -0.1444 ** -0.0643 -0.1547 -0.1210 -0.1481
SWV_hedonism -0.1418 *** -0.1304 *** -0.1027 -0.1832 -0.1547 -0.1671
SWV_power -0.1655 *** -0.1083 ** -0.1337 -0.1553 -0.1321 -0.1462
SWV_security -0.1871 *** -0.2111 *** -0.1488 -0.1920 -0.1771 -0.1692
SWV_selfDirection -0.0633 -0.0709 -0.0841 0.0099 0.0416 0.0230
SWV_stimulation 0.1131 ** 0.1221 *** 0.1247 0.1581 0.1726 0.1698
SWV_tradition -0.1554 *** -0.1371 *** -0.1389 -0.2205 -0.1608 -0.1980
hypocrisy_increasing -0.0253 -0.0304 -0.0182 -0.0619 -0.0557 -0.0578
blood_pressure_high -0.4244 *** -0.5468 *** -0.4903 * -0.5326 ** -0.5493 *
chronic_disease -0.4653 *** -0.4767 *** -0.3628 -0.3190 -0.3262
smoker -0.2274 ** -0.3870 -0.2631 -0.1437 -0.1423
HQ_1_health_plans -0.1579 *** -0.1065 -0.1080 -0.0718 -0.0847
HQ_4_habit_choice -0.2547 *** -0.3755 *** -0.4382 *** -0.4013 *** -0.4491 ***
HQ_5_health_is_gift 0.0605 ** 0.0473 0.0096 0.0241 0.0326
HQ_6_avoid_test_results 0.0576 ** 0.0801 0.0084 0.0419 0.0138
Education_Reference -0.0040 0.0022 0.0029
General_News 0.0002 -0.0033 -0.0035
Interactive_Web_Applications 0.0477 0.9753 1.1660
Internet_Radio_TV 0.0252 0.1096 0.1118
Malicious_Sites -0.1538 0.0120 0.0103
Motor_Vehicles -0.0143 -0.0341 -0.0353
Online_Shopping -0.0026 -0.0019 -0.0021
Personal_Network_Storage 0.1282 0.3129 0.2914
Personal_Pages 0.0366 -0.3409 -0.3620
Recreation_Hobbies 0.0161 0.0213 0.0229
Search_Engines 0.0014 -0.0015 -0.0017
Sports 0.0061 0.0131 0.0119
total_web_visits 0.0011 -0.0001 0.0014
total_app_time 0.0000 0.0000
ENTERTAINMENT 0.0011 0.0012
HEALTH_AND_FITNESS 0.0095 *** 0.0095 ***
HOUSE_AND_HOME -0.0137 -0.0110
LIBRARIES_AND_DEMO -0.5137 -0.5396
LIFESTYLE 0.0051 0.0047
MEDICAL -0.0168 -0.0162
MUSIC_AND_AUDIO 0.0037 0.0038
5.3 Determinants of Exercise
In the aim of understanding the contribution of individual variables
to whether a person exercises, we employ multivariate logistic regres-
sion analysis, building models from demographics-based baseline.
The resulting models are shown in Table 3. In consideration of space,
only variables which have a coefficient significant at
shown; note however that the p-value markers (stars) shown in the
table have been Bonferroni-adjusted to ameliorate the multiple com-
parison problems. Intuitively, the Bonferroni correction “punishes”
the significance of features in a larger model, allowing fewer oth-
erwise significant tests pass the adjusted
threshold (as is visible
in right-most columns of the table). Also at the top of each model,
we show the number of users having non-empty fields available for
the data (
) and McFadden’s
, which relates the (maximized)
likelihood value from the current fitted model to null model [40].
In baseline model using only demographics we find a strong
positive relationship between exercise and education,income, and
wealth, and a negative one with age (with age echoing findings of
UMAP ’19, June 9–12, 2019, Larnaca, Cyprus Mejova & Kalimeri
previous studies [
]). However, the explanatory power of this
model is low, according to
. Adding the moral values, we find
exercise to be highly related with happiness and stimulation, and
negatively related to hedonism,power,security, and tradition. Note
the difference between these and values associated with download-
ing a health app from Section 4, now with an negative association
with power (“attainment of a dominant position”) and positive with
happiness (latter is well documented to accompany exercise [35]).
Next, the addition of health-related variables unsurprisingly pro-
duces highly significant coefficients. There is a robust negative re-
lationship between exercise and having high blood pressure, some
other chronic disease, and being a smoker. Interestingly, we also find
strong effects in the belief statements of respondents. Those who
exercise are more likely to agree with the statement “When I think
of making plans for the future, my health is something I strongly
consider” (HQ 1), “We all have a choice about how to lead our lives,
and healthy habits are just one example of that” (HQ 4), but are more
likely to disagree with “Health is a gift and there is not much I can
do about it” (HQ 5) and “Sometimes I avoid getting my test results
if I think it will be bad news” (HQ 6). These findings underscore the
importance of the individual’s belief in their innate ability to achieve
goals or self-efficacy [38].
Upon adding the internet browsing data (available for both Desk-
top and Mobile cohorts), we find the most important domains to be
in the areas of Personal Network Storage (content management),
Malicious Sites (may include adult content), and Interactive Web
Applications (document readers, calendar), though none had a par-
ticularly significant p-value after Bonferroni correction. Though in
combination with baseline variables, the model achieves
0.704 and 0.870 for Desktop and Mobile cohorts respectively.
Finally, as we consider the use of applications (listed in caps) for
the respondents who shared their mobile activity, we find several
application classes beneficial to the model, the most significant
of these being Health & Fitness. Note that we have also included
aggregate technology usage statistics, including a total number of
web visits in the observed time, the total app time and (not shown due
to insufficient significance) the total web browsing time. However,
we find these to be not highly related to exercise behavior.
As the models increase in complexity (see the last two columns
of Table 3), the Bonferroni correction becomes more strict, which
reveals the variables most important in modeling exercise: a combi-
nation of demographics (wealth), attitude (HQ4 “Health is a choice”),
and technology use (Health & Fitness Apps).
This study is a contribution to an exciting area of research in User
Modeling and Personalization into the user-centered design of per-
suasive technologies and behavior change interventions to improve
health and well-being. In the past, psychological features have been
used to model healthy shopping habits [
], cooking [
], and engage-
ment in physical activity [
]. Thus far, existing UMAP literature
focused on personality traits which could be leveraged for personal-
ization [
]. Here, we show that values held by the individuals
also affect their health behavior. In a sense, it can be seen as a re-
sponse to a recent study on strategies to encourage diet and physical
exercise by Radha et al. [
], who called for the study of factors that
could “explain attitude towards the feasibility level of a recommen-
dation”. Here, we illustrate the use of validated, finer-grained value
theories in the modeling of technology users, while also contributing
an observational technology use and rich demographics.
In particular, insights obtained in this study lead us to recommend
the following considerations when designing persuasive technologies
for health behavior modification:
Focus on a
particular activity
to track. We show that applications
having most users exercising are marketed for tracking a particular
activity, such as walking, running, or cycling. Those centered
around particular wearables, for instance, fare less well.
Create interventions with
socioeconomic status
of the users in
mind. We find that wealth is one of the greatest determinants of
regular exercise. More should be done to understand the barriers
of the less wealthy to leading a healthy lifestyle. For instance, it is
curious that it is wealth, not income, that remains most predictive
in the regression model.
Incorporate discussion of
in the interaction or interface.
One of the strongest predictors for exercise is the belief that
healthy habits are a conscious choice. Making this choice explicit
may reinforce this value and encourage engagement.
Encourage the expression of
and offer emotional re-
wards. We find people engaging in exercise identifying more
strongly with the value of happiness, and although the direction
of causation may point both ways, associating positive emotions
with physical activity may reinforce the connection. Note that
our findings that the value of power is negatively associated with
exercise suggest that competitions and leader boards may not be
appropriate for many users.
Use application usage in
predictive analytics
, in the absence of
detailed demographic or value information. In our classification
experiments, we find mobile application usage to be much more
useful in predicting exercise than internet browsing.
As mentioned earlier, the greatest limitation of this study is the
reliance on self-reporting when measuring exercise, as many biases
are possible. In future studies, we encourage researchers also to
obtain permission to gather user physical activity (which can be done
unobtrusively via pedometers and heart rate monitors). Further, we
realize administering scientifically validated surveys to technology
users may be infeasible. However, attempts are being made to detect
moral judgments in social media [
] and associated with images
], with potential for automatic value detection in the future. Also,
although we capture one month of technology use, the study is not
longitudinal – many people may have downloaded and used the
apps at some time, but not in our window of observation. Thus
conclusions on application adoption/retention should be made while
considering the small eventual sample size per application.
Finally, we would like to reiterate the privacy precautions taken
in this study, with respondent anonymization, data aggregation, and
URL cleaning, which is performed to limit the exposure of partici-
pants. Similar precautions should be taken if or when an inference
of values or other personal information is performed, such that the
user is given greatest possible control over his or her information, as
enforced by, for instance, EU General Data Protection Regulation
Effect of Values and Technology Use on Exercise UMAP ’19, June 9–12, 2019, Larnaca, Cyprus
Y.M. and K.K., acknowledge support from the “Lagrange Project”
of the Institute for Scientific Interchange (ISI) funded by the Fon-
dazione Cassa di Risparmio di Torino (CRT).
Giovanni Acampora, Diane J Cook, Parisa Rashidi, and Athanasios V Vasilakos.
2013. A survey on ambient intelligence in healthcare. Proc. IEEE 101, 12 (2013),
Ifeoma Adaji, Kiemute Oyibo, and Julita Vassileva. 2018. The Effect of Gen-
der and Age on the Factors That Influence Healthy Shopping Habits in E-
Commerce. In Proceedings of the 26th Conference on User Modeling, Adap-
tation and Personalization (UMAP ’18). ACM, New York, NY, USA, 251–255.
Stephanie Alley, Cally Jennings, Ronald C Plotnikoff, and Corneel Vandelanotte.
2016. Web-based video-coaching to assist an automated computer-tailored physi-
cal activity intervention for inactive adults: a randomized controlled trial. Journal
of Medical Internet Research 18, 8 (2016).
Gordon Willard Allport and Philip Ewart Vernon. 1960. A study of values. (1960).
Ross E Andersen, Carlos J Crespo, Susan J Bartlett, Lawrence J Cheskin, and
Michael Pratt. 1998. Relationship of physical activity and television watching
with body weight and level of fatness among children: results from the Third
National Health and Nutrition Examination Survey. Journal of the American
Medical Association 279, 12 (1998), 938–942.
Stacy C Bailey, Rachel O’conor, Elizabeth A Bojarski, Rebecca Mullen, Rachel E
Patzer, Daniel Vicencio, Kara L Jacobson, Ruth M Parker, and Michael S Wolf.
2015. Literacy disparities in patient access and health-related use of I nternet and
mobile technologies. Health Expectations 18, 6 (2015), 3079–3087.
Elizabeth A Baker, Laura K Brennan, Ross Brownson, and Robyn A Houseman.
2000. Measuring the determinants of physical activity in the community: current
and future directions. Research Quarterly for Exercise and Sport 71, sup2 (2000),
Kylie Ball, Robert W Jeffery, Gavin Abbott, Sarah A McNaughton, and David
Crawford. 2010. Is healthy behavior contagious: associations of social norms with
physical activity and healthy eating. International Journal of Behavioral Nutrition
and Physical Activity 7, 1 (2010), 86.
Jacob E Barkley and Andrew Lepp. 2016. Mobile phone use among college stu-
dents is a sedentary leisure behavior which may interfere with exercise. Computers
in Human Behavior 56 (2016), 29–33.
Rebecca Boulos, Emily Kuross Vikre, Sophie Oppenheimer, Hannah Chang, and
Robin B Kanarek. 2012. ObesiTV: how television is influencing the obesity
epidemic. Physiology & Behavior 107, 1 (2012), 146–153.
[11] Leo Breiman. 2001. Random forests. Machine Learning 45, 1 (2001), 5–32.
Audun Brunes, Liv Berit Augestad, and Sigridur Lara Gudmundsdottir. 2013.
Personality, physical activity, and symptoms of anxiety and depression: the HUNT
study. Social Psychiatry and Psychiatric Epidemiology 48, 5 (2013), 745–756.
Jennifer K Carroll, Anne Moorhead, Raymond Bond, William G LeBlanc, Robert J
Petrella, and Kevin Fiscella. 2017. Who uses mobile phone health apps and does
use matter? A secondary data analytics approach. Journal of Medical Internet
Research 19, 4 (2017).
Federica Cena, Amon Rapp, Cataldo Musto, and Pasquale Lops. 2018. Towards
a Conceptual Model for Holistic Recommendations. In Adjunct Publication of
the 26th Conference on User Modeling, Adaptation and Personalization. ACM,
Guanliang Chen, Dan Davis, Claudia Hauff, and Geert-Jan Houben. 2016. On the
Impact of Personality in Massive Open Online Learning. In Proceedings of the
2016 Conference on User Modeling Adaptation and Personalization (UMAP ’16).
ACM, New York, NY, USA, 121–130.
Timothy S Church, Diana M Thomas, Catrine Tudor-Locke, Peter T Katzmarzyk,
Conrad P Earnest, Ruben Q Rodarte, Corby K Martin, Steven N Blair, and Claude
Bouchard. 2011. Trends over 5 decades in US occupation-related physical activity
and their associations with obesity. PLOS ONE 6, 5 (2011), e19657.
Damien L Crone, Stefan Bode, Carsten Murawski, and Simon M Laham. 2018.
The Socio-Moral Image Database (SMID): A novel stimulus set for the study of
social, moral and affective processes. PLOS ONE 13, 1 (2018), e0190954.
R William Doherty. 1997. The emotional contagion scale: A measure of individual
differences. Journal of Nonverbal Behavior 21, 2 (1997), 131–154.
Clemens Ernsting, Stephan U Dombrowski, Monika Oedekoven, Julie LO, et al
2017. Using smartphones and health apps to change and manage health behaviors:
a population-based survey. Journal of Medical Internet Research 19, 4 (2017).
Bruce Ferwerda, Marko Tkalcic, and Markus Schedl. 2017. Personality Traits
and Music Genres: What Do People Prefer to Listen To?. In Proceedings of the
25th Conference on User Modeling, Adaptation and Personalization (UMAP ’17).
ACM, New York, NY, USA, 285–288.
Suparna Ghanvatkar, Atreyi Kankanhalli, and Vaibhav Rajan. 2019. User Models
for Personalized Physical Activity Interventions: Scoping Review. Journal of
Medical Internet Research mHealth and uHealth 7, 1 (2019), e11098.
Jesse Graham, Brian a Nosek, Jonathan Haidt, Ravi Iyer, Spassena Koleva, and
Peter H Ditto. 2011. Mapping the moral domain. Journal of Personality and
Social Psychology 101, 2 (Aug. 2011), 366–85.
Jonathan Haidt and Jesse Graham. 2007. When morality opposes justice: Con-
servatives have moral intuitions that liberals may not recognize. Social Justice
Research 20, 1 (2007), 98–116.
Jonathan Haidt and Craig Joseph. 2004. Intuitive ethics: How innately prepared
intuitions generate culturally variable virtues. Daedalus 133, 4 (2004), 55–66.
Sarah Hales, Gabrielle Turner-McGrievy, Arjang Fahim, Andrew Freix, Sara
Wilcox, Rachel E Davis, Michael Huhns, and Homayoun Valafar. 2016. A mixed-
methods approach to the development, refinement, and pilot testing of social
networks for improving healthy behaviors. Journal of Medical Internet Research
Human Factors 3, 1 (2016).
Christina N Harrington, Lauren Wilcox, Kay Connelly, Wendy Rogers, and Jon
Sanford. 2018. Designing Health and Fitness Apps with Older Adults: Examining
the Value of Experience-Based Co-Design. In Proceedings of the 12th EAI Inter-
national Conference on Pervasive Computing Technologies for Healthcare. ACM,
Elaine Hatfield, John T Cacioppo, and Richard L Rapson. 1993. Emotional
contagion. Current Directions in Psychological Science 2, 3 (1993), 96–100.
Heather A Hausenblas and Peter R Giacobbi Jr. 2004. Relationship between exer-
cise dependence symptoms and personality. Personality and Individual Differences
36, 6 (2004), 1265–1273.
Andrew T Kaczynski and Karla A Henderson. 2007. Environmental correlates
of physical activity: a review of evidence about parks and recreation. Leisure
Sciences 29, 4 (2007), 315–354.
Kyriaki Kalimeri, Mariano G Beiró, Matteo Delfino, Robert Raleigh, and Ciro
Cattuto. 2018. Predicting demographics, moral foundations, and human values
from digital behaviours. Computers in Human Behavior (2018).
Kyriaki Kalimeri, Aleksandar Matic, and Alessandro Cappelletti. 2010. RFID:
Recognizing failures in dressing activity. In 2010 4th International Conference on
Pervasive Computing Technologies for Healthcare. IEEE, 1–4.
Charlotte Kerner and Victoria A Goodyear. 2017. The motivational impact of
wearable healthy lifestyle technologies: a self-determination perspective on Fitbits
with adolescents. American Journal of Health Education 48, 5 (2017), 287–297.
Paul Krebs and Dustin T Duncan. 2015. Health app use among US mobile phone
owners: a national survey. Journal of Medical Internet Research mHealth and
uHealth 3, 4 (2015).
[34] Neal Lathia, Veljko Pejovic, Kiran K Rachuri, Cecilia Mascolo, Mirco Musolesi,
and Peter J Rentfrow. 2013. Smartphones for large-scale behavior change inter-
ventions. IEEE Pervasive Computing 3 (2013), 66–73.
Neal Lathia, Gillian M. Sandstrom, Cecilia Mascolo, and Peter J. Rentfrow. 2017.
Happier People Live More Active Lives: Using Smartphones to Link Happiness
and Physical Activity. PLOS ONE 12, 1 (01 2017), 1–13.
Martin J Lee, Jean Whitehead, Nikos Ntoumanis, and Antonis Hatzigeorgiadis.
2008. Relationships among values, achievement orientations, and attitudes in
youth sport. Journal of Sport and Exercise Psychology 30, 5 (2008), 588–610.
David M Levine, Stuart R Lipsitz, and Jeffrey A Linder. 2016. Trends in se-
niorsâ ˘
Z use of digital health technology in the United States, 2011-2014. Journal
of the American Medical Association 316, 5 (2016), 538–540.
Leib Litman, Zohn Rosen, David Spierer, Sarah Weinberger-Litman, Akiva Gold-
schein, and Jonathan Robinson. 2015. Mobile exercise apps and increased leisure
time exercise activity: a moderated mediation analysis of the role of self-efficacy
and barriers. Journal of Medical Internet Research 17, 8 (2015).
Cyril FM Marsaux, Carlos Celis-Morales, Katherine M Livingstone, Rosalind
Fallaize, Silvia Kolossa, Jacqueline Hallmann, Rodrigo San-Cristobal, Santiago
Navas-Carretero, Clare B O’Donovan, Clara Woolhead, et al
2016. Changes
in physical activity following a genetic-based internet-delivered personalized
intervention: randomized controlled trial (Food4Me). Journal of Medical Internet
Research 18, 2 (2016).
Daniel McFadden et al
1973. Conditional logit analysis of qualitative choice
behavior. (1973).
Nancy McLennan and Jannine Thompson. 2015. Quality physical education
(QPE): Guidelines for policy makers. UNESCO Publishing.
Harm op den Akker, Valerie M. Jones, and Hermie J. Hermens. 2014. Tailoring
real-time physical activity coaching systems: a literature survey and model. User
Modeling and User-Adapted Interaction 24, 5 (01 Dec 2014), 351–392. https:
// 9146-y
Kiemute Oyibo. 2016. Designing Culture-based Persuasive Technology to Pro-
mote Physical Activity Among University Students. In Proceedings of the 2016
Conference on User Modeling Adaptation and Personalization (UMAP ’16). ACM,
New York, NY, USA, 321–324.
Michael Pratt, Olga L Sarmiento, Felipe Montes, David Ogilvie, Bess H Marcus,
Lilian G Perez, Ross C Brownson, Lancet Physical Activity Series Working Group,
et al
2012. The implications of megatrends in information and communication
UMAP ’19, June 9–12, 2019, Larnaca, Cyprus Mejova & Kalimeri
technology and transportation for changes in global physical activity. The Lancet
380, 9838 (2012), 282–293.
Mashfiqui Rabbi, Min Hane Aung, Mi Zhang, and Tanzeem Choudhury. 2015.
MyBehavior: automatic personalized health feedback from user behaviors and
preferences using smartphones. In Proceedings of the 2015 ACM International
Joint Conference on Pervasive and Ubiquitous Computing. ACM, 707–718.
Mustafa Radha, Martijn C. Willemsen, Mark Boerhof, and Wijnand A. IJssel-
steijn. 2016. Lifestyle Recommendations for Hypertension Through Rasch-based
Feasibility Modeling. In Proceedings of the 2016 Conference on User Model-
ing Adaptation and Personalization (UMAP ’16). ACM, New York, NY, USA,
Research2Guidance. 2018. mHealth Economics - How mHealth App Publishers
Are Monetizing Their Apps.
RE Rhodes and NEI Smith. 2006. Personality correlates of physical activity:
a review and meta-analysis. British Journal of Sports Medicine 40, 12 (2006),
Arbindra Rimal. 2002. Association of nutrition concerns and socioeconomic status
with exercise habits. International Journal of Consumer Studies 26, 4 (2002),
Markus Rokicki, Eelco Herder, Tomasz Ku´
smierczyk, and Christoph Trattner.
2016. Plate and Prejudice: Gender Differences in Online Cooking. In Proceedings
of the 2016 Conference on User Modeling Adaptation and Personalization (UMAP
’16). ACM, New York, NY, USA, 207–215.
Sohrab Saeb, Mi Zhang, Christopher J Karr, Stephen M Schueller, Marya E
Corden, Konrad P Kording, and David C Mohr. 2015. Mobile phone sensor
correlates of depressive symptom severity in daily-life behavior: an exploratory
study. Journal of Medical Internet Research 17, 7 (2015).
Daniel Sanchez-Valdes and Gracian Trivino. 2015. Linguistic and emotional
feedback for self-tracking physical activity. Expert Systems with Applications 42,
24 (2015), 9574 – 9586.
Shalom H Schwartz. 2012. An overview of the Schwartz theory of basic values.
Online Readings in Psychology and Culture 2, 1 (2012), 11.
CE Short, Amanda Rebar, EL James, MJ Duncan, KS Courneya, RC Plotnikoff,
R Crutzen, and C Vandelanotte. 2017. How do different delivery schedules of
tailored web-based physical activity advice for breast cancer survivors influence
intervention use and efficacy? Journal of Cancer Survivorship 11, 1 (2017),
Kerem Shuval, Qing Li, Kelley Pettee Gabriel, and Rusty Tchernis. 2017. Income,
physical activity, sedentary behavior, and the ‘weekend warrior’ among US adults.
Preventive Medicine 103 (2017), 91–97.
Aaron Smith and Dana Page. 2015. US smartphone use in 2015. Pew Research
Center 1 (2015).
Gaurav Sood. 2015. Content Categories for Unique Domains in comScore brows-
ing data.
Livia Teernstra, Peter van der Putten, Liesbeth Noordegraaf-Eelens, and Fons
Verbeek. 2016. The morality machine: tracking moral values in Tweets. In Inter-
national Symposium on Intelligent Data Analysis. Springer, 26–37.
Jerry CC Tseng, Bo-Hau Lin, Yu-Feng Lin, Vincent S Tseng, Miin-Luen Day,
Shyh-Chyi Wang, Kuen-Rong Lo, and Yi-Ching Yang. 2015. An interactive health-
care system with personalized diet and exercise guideline recommendation. In
Technologies and Applications of Artificial Intelligence (TAAI), 2015 Conference
on. IEEE, 525–532.
Corneel Vandelanotte, Camille Short, Ronald C Plotnikoff, Cindy Hooker, D
Canoy, Amanda Rebar, Stephanie Alley, Stephanie Schoeppe, W Kerry Mum-
mery, and Mitch J Duncan. 2015. TaylorActive–Examining the effectiveness of
web-based personally-tailored videos to increase physical activity: a randomized
controlled trial protocol. BMC Public Health 15, 1 (2015), 1020.
Elad Yom-Tov, Guy Feraru, Mark Kozdoba, Shie Mannor, Moshe Tennenholtz,
and Irit Hochberg. 2017. Encouraging physical activity in patients with diabetes:
intervention using a reinforcement learning system. Journal of Medical Internet
Research 19, 10 (2017).
... Attempts to inform potential interventions have been made by measuring the importance of community support during a weight loss journey (Cunha et al., 2016) and the effect of intervention messaging on those affected by anorexia (Yom-Tov et al., 2012). Observational studies of exercise in particular through specialized exercise applications have shown that information about other people's routine may affect one's own (Aral & Nicolaides, 2017) and that gender plays an important role in the continued use of such apps (Mejova & Kalimeri, 2019). Further, a combination of web search and wearables data has been used to show the health impact of applications not necessarily meant for exercise, such as Pokémon Go, which resulted in potentially years worth of life spans added to the fans of the game (Althoff et al., 2016). ...
Full-text available
Computational social science has had a profound impact on the study of health and disease, mainly by providing new data sources for all of the primary Ws—what, who, when, and where—in order to understand the final “why” of disease. Anonymized digital trace data bring a new level of detail to contact networks, search engine and social media logs allow for the now-casting of symptoms and behaviours, and media sharing informs the formation of attitudes pivotal in health decision-making. Advances in computational methods in network analysis, agent-based modelling, as well as natural language processing, data mining, and time series analysis allow both the extraction of fine-grained insights and the construction of abstractions over the new data sources. Meanwhile, numerous challenges around bias, privacy, and ethics are being negotiated between data providers, academia, the public, and policymakers in order to ensure the legitimacy of the resulting insights and their responsible incorporation into the public health decision-making. This chapter outlines the latest research on the application of computational social science to epidemiology and the data sources and computational methods involved and spotlights ongoing efforts to address the challenges in its integration into policymaking.
... Artificial intelligence (AI) applications are revolutionising our everyday lives in a wide range of domains like: agriculture [27], disaster prevention and response [10], education and productivity [9,18,11], health care and well-being [30,20,24,32], and ubiquitous personalisation [19]. Substantial increase in the future uses of AI will reinforce human efficiency, bringing economic and social benefits shared across society. ...
Full-text available
Artificial Intelligence (AI) is an integral part of our lives with AI systems to revolutionise our daily practices. At the same time, the rapid pace of AI innovations entails inherent risks that can range from cyber-crime to social discrimination. Here, we administered a large scale survey (\(n=1298\)) assessing peoples’ concerns and expectations regarding AI’s influence on society in the future decade. The AI concerns employed in this study, originate from the “One hundred year study on Artificial Intelligence” project. Taking Norway as a case study, we discuss the participants’ prioritisation of concerns for their socio-demographic characteristics. Our findings show a divide in the society; with younger generations to expect a positive impact of AI on our lives in the future decade. More sceptical groups are afraid of structural changes in the economy and job losses, while supporters see opportunities that will improve our life quality. These findings can inform both academics and policymakers that should work closely to ensure fairness, explainability and maintain a trusting relationship between AI and society.
... Leveraging on the immense amount of digital data produced daily, the field of Digital Demography emerged, addressing vital research questions of demographic research via innovative data sources. These new sources of data are demonstrated to be particularly powerful in monitoring a series of demographic phenomena such as birthrates [5], mortality [4], unemployment [10], daily commuting [6], international and internal migration [55], but also modelling more complex socio-demographic issues such as psychological well-being and attitudes towards health [24,30]. Digital data are particularly useful in cases where official data are sparse, incomplete, or even impossible to obtain. ...
Online social network is a platform that plays an essential role in identifying the emotional values of user-generated content such as blogs, posts, and comments along with their influential factors. Especially on Twitter, network users are growing worldwide day by day and creating a massive amount of data that is not analyzed effectively in a quick way. Identifying the most influential persons on the social network is also a challenging task over the wide range of real-time applications like recommendation systems. Now, to handle these situations, this article proposes a novel approach for prediction of information diffusion that includes emotion recognition with sarcasm detection based influence spreader identification (PID-ERSDISI). The proposed method uses the user-generated posts for emotion recognition in tandem with sarcasm detection both implicitly and explicitly. This approach helps to gauge the leverage that influences spreaders and also enhances the prediction accuracy of information diffusion in a better way. The implementation of the proposed work executed their task one after another in the following way, namely, sarcasm detection, emotional-level computation, breakpoint computation, breakpoint validation, influential user generation, and information diffusion. After the successful implementation of this proposed PID-ERSDIS, it produced prominent results against other state-of-art methods.
Online social networks (OSNs) is a platform that plays an essential role in identifying misinformation like false rumors, insults, pranks, hoaxes, spear phishing and computational propaganda in a better way. Detection of misinformation finds its applications in areas such as law enforcement to pinpoint culprits who spread rumors to harm the society, targeted marketing in e-commerce to identify the user who originates dissatisfaction messages about products or services that harm an organizations reputation. The process of identifying and detecting misinformation is very crucial in complex social networks. As misinformation in social network is identified by designing and placing the monitors, computing the minimum number of monitors for detecting misinformation is a very trivial work in the complex social network. The proposed approach determines the top suspected sources of misinformation using a tweet polarity-based ranking system in tandem with sarcasm detection (both implicit and explicit sarcasm) with optimization approaches on large-scale incomplete network. The algorithm subsequently uses this determined feature to place the minimum set of monitors in the network for detecting misinformation. The proposed work focuses on the timely detection of misinformation by limiting the distance between the suspected sources and the monitors. The proposed work also determines the root cause of misinformation (provenance) by using a combination of network-based and content-based approaches. The proposed work is compared with the state-of-art work and has observed that the proposed algorithm produces better results than existing methods.
Full-text available
Background: Fitness devices have spurred the development of apps that aim to motivate users, through interventions, to increase their physical activity (PA). Personalization in the interventions is essential as the target users are diverse with respect to their activity levels, requirements, preferences, and behavior. Objective: This review aimed to (1) identify different kinds of personalization in interventions for promoting PA among any type of user group, (2) identify user models used for providing personalization, and (3) identify gaps in the current literature and suggest future research directions. Methods: A scoping review was undertaken by searching the databases PsycINFO, PubMed, Scopus, and Web of Science. The main inclusion criteria were (1) studies that aimed to promote PA; (2) studies that had personalization, with the intention of promoting PA through technology-based interventions; and (3) studies that described user models for personalization. Results: The literature search resulted in 49 eligible studies. Of these, 67% (33/49) studies focused solely on increasing PA, whereas the remaining studies had other objectives, such as maintaining healthy lifestyle (8 studies), weight loss management (6 studies), and rehabilitation (2 studies). The reviewed studies provide personalization in 6 categories: goal recommendation, activity recommendation, fitness partner recommendation, educational content, motivational content, and intervention timing. With respect to the mode of generation, interventions were found to be semiautomated or automatic. Of these, the automatic interventions were either knowledge-based or data-driven or both. User models in the studies were constructed with parameters from 5 categories: PA profile, demographics, medical data, behavior change technique (BCT) parameters, and contextual information. Only 27 of the eligible studies evaluated the interventions for improvement in PA, and 16 of these concluded that the interventions to increase PA are more effective when they are personalized. Conclusions: This review investigates personalization in the form of recommendations or feedback for increasing PA. On the basis of the review and gaps identified, research directions for improving the efficacy of personalized interventions are proposed. First, data-driven prediction techniques can facilitate effective personalization. Second, use of BCTs in automated interventions, and in combination with PA guidelines, are yet to be explored, and preliminary studies in this direction are promising. Third, systems with automated interventions also need to be suitably adapted to serve specific needs of patients with clinical conditions. Fourth, previous user models focus on single metric evaluations of PA instead of a potentially more effective, holistic, and multidimensional view. Fifth, with the widespread adoption of activity monitoring devices and mobile phones, personalized and dynamic user models can be created using available user data, including users' social profile. Finally, the long-term effects of such interventions as well as the technology medium used for the interventions need to be evaluated rigorously.
Conference Paper
Full-text available
Identifying useful features of health and fitness technologies has the potential to lead to more effective pervasive technology interventions. Through collaborative creative processes, co-design provides useful insight and direction for the design of these technologies from populations such as older adults who are often neglected in design. Additionally, extensive technology experience prior to co-design may elicit richer feedback from users that may otherwise be unfamiliar with the intervention of interest. In this paper, we examine the value of experience-based co-design among older adults. We report on seven design sessions conducted with a total of 25 older adults, following a 10-week study of fitness app effectiveness. Sessions were grouped into users and non-users of the assigned app following the study. We found that continued use with the assigned application led to more robust, detailed feedback in design sessions, suggesting that long-term, prior use of sample technologies is an important prerequisite to ideating useful features for new health technology. We draw on these findings to make design recommendations, and discuss the values and guidelines of implementing experience-based co-design.
Conference Paper
Full-text available
In this paper we introduce the concept of holistic recommendations, namely a set of suggestions generated by exploiting a more comprehensive representation of the user that relies on the personal information coming from different heterogeneous data sources (e.g., social networks, wristbands, smartphones, etc.) and considers the diverse relations and constraints among the data encoded in the profiles. Specifically, in this article we provide the following contributions: i) we outline a conceptual model for providing holistic recommendations built on the ground of such richer user profiles; ii) we present some challenges related to holistic recommendations that can inspire further research in the field.
Conference Paper
Full-text available
People typically eat what they shop for; if consumers shop for healthy foods, they will likely eat healthy foods. In order to influence healthier eating habits among consumers, it is important to identify the factors that influence them to shop for healthy foods. To contribute to ongoing research in this area, we explore the influence of commonly used e-commerce strategies: personality, persuasive strategies, social support, relative price, and perceived product quality on healthy shopping habits among e-commerce shoppers. Research has shown that personalizing these strategies makes them more effective in achieving the desired behavior change among users. Age and gender have been identified as factors that can be used for group-based personalization. We thus investigate the moderating effect of age and gender on the factors that influence healthy shopping habits in e-commerce shoppers. To achieve this, we carried out an online study of 244 e-commerce shoppers. Using partial least squares structural equation modeling (PLS-SEM), we developed a path model using the commonly used e-commerce factors: personality, persuasive strategies, social support, relative price, and perceived product quality. The result of our analysis suggests that social support, relative price and perceived product quality significantly influence healthy shopping habits in e-commerce shoppers. In addition, females are more influenced by social support to adopt healthy shopping habits compared to male e-shoppers. Furthermore, older shoppers are more influenced by social support to adopt healthy shopping habits, while the younger shoppers are more influenced by the relative price of products.
Full-text available
A major obstacle for the design of rigorous, reproducible studies in moral psychology is the lack of suitable stimulus sets. Here, we present the Socio-Moral Image Database (SMID), the largest standardized moral stimulus set assembled to date, containing 2,941 freely available photographic images, representing a wide range of morally (and affectively) positive, negative and neutral content. The SMID was validated with over 820,525 individual judgments from 2,716 participants, with normative ratings currently available for all images on affective valence and arousal, moral wrongness, and relevance to each of the five moral values posited by Moral Foundations Theory. We present a thorough analysis of the SMID regarding (1) inter-rater consensus, (2) rating precision, and (3) breadth and variability of moral content. Additionally, we provide recommendations for use aimed at efficient study design and reproducibility, and outline planned extensions to the database. We anticipate that the SMID will serve as a useful resource for psychological, neuroscientific and computational (e.g., natural language processing or computer vision) investigations of social, moral and affective processes. The SMID images, along with associated normative data and additional resources are available at
Full-text available
Personal electronic devices such as smartphones give access to a broad range of behavioral signals that can be used to learn about the characteristics and preferences of individuals. In this study we explore the connection between demographic and psychological attributes and digital records for a cohort of 7,633 people, closely representative of the US population with respect to gender, age, geographical distribution, education, and income. We collected self-reported assessments on validated psychometric questionnaires based on both the Moral Foundations and Basic Human Values theories, and combined this information with passively-collected multi-modal digital data from web browsing behavior, smartphone usage and demographic data. Then, we designed a machine learning framework to infer both the demographic and psychological attributes from the behavioral data. In a cross-validated setting, our model is found to predict demographic attributes with good accuracy (weighted AUC scores of 0.90 for gender, 0.71 for age, 0.74 for ethnicity). Our weighted AUC scores for Moral Foundation attributes (0.66) and Human Values attributes (0.60) suggest that accurate prediction of complex psychometric attributes is more challenging but feasible. This connection might prove useful for designing personalized services, communication strategies, and interventions, and can be used to sketch a portrait of people with similar worldviews.
Full-text available
Background Regular physical activity is known to be beneficial for people with type 2 diabetes. Nevertheless, most of the people who have diabetes lead a sedentary lifestyle. Smartphones create new possibilities for helping people to adhere to their physical activity goals through continuous monitoring and communication, coupled with personalized feedback. Objective The aim of this study was to help type 2 diabetes patients increase the level of their physical activity. Methods We provided 27 sedentary type 2 diabetes patients with a smartphone-based pedometer and a personal plan for physical activity. Patients were sent short message service messages to encourage physical activity between once a day and once per week. Messages were personalized through a Reinforcement Learning algorithm so as to improve each participant’s compliance with the activity regimen. The algorithm was compared with a static policy for sending messages and weekly reminders. Results Our results show that participants who received messages generated by the learning algorithm increased the amount of activity and pace of walking, whereas the control group patients did not. Patients assigned to the learning algorithm group experienced a superior reduction in blood glucose levels (glycated hemoglobin [HbA1c]) compared with control policies, and longer participation caused greater reductions in blood glucose levels. The learning algorithm improved gradually in predicting which messages would lead participants to exercise. Conclusions Mobile phone apps coupled with a learning algorithm can improve adherence to exercise in diabetic patients. This algorithm can be used in large populations of diabetic patients to improve health and glycemic control. Our results can be expanded to other areas where computer-led health coaching of humans may have a positive impact. Summary of a part of this manuscript has been previously published as a letter in Diabetes Care, 2016.
Full-text available
Background: Considerable numbers of young people are not meeting physical activity guidelines. Wearable fitness devices can provide opportunities for physical activity promotion. Purpose: The aim of the study was to explore whether wearable healthy lifestyle technologies impacted on adolescents’ (13- to 14-year-olds) motivation for physical activity. Methods: The study was a mixed method sequential design. Participants were 84 adolescents (44 girls, 40 boys) from 6 physical education classes. Pupils were issued with a Fitbit to wear for 8 weeks and completed pre-/posttest questionnaires that assessed motivational regulation and psychological need satisfaction. Adolescents also engaged in focus group interviews after wearing the Fitbit for 8 weeks. Quantitative data were analyzed using a repeated measures multivariate analysis of variance (MANOVA) to explore differences between gender and time. Qualitative data analysis was conducted deductively using self-determination theory. Results: The quantitative findings identified significant reductions in need satisfaction and autonomous motivation and significant increases in amotivation after 8 weeks. Qualitative evidence suggested short-term increases in motivation through feelings of competition, guilt, and internal pressure. Discussion: Findings suggest that healthy lifestyle technology may have negative motivational consequences. Translation to Health Education Practice: Certified Health Education Specialists should support young people to personalize health targets in order to critically engage with normalized health targets.
The present study examines the association between income and physical activity intensity along the entire continuum using accelerometry in a nationally representative sample of U.S. adults. Specifically, we assessed the relationship between annual household income, sedentary behavior, light, and moderate-vigorous intensity physical activity, and meeting physical activity guidelines over a brief, 2-day period (‘weekend warrior’), and during the entire week. The sample consisted of 5206 National Health and Examination Survey adult participants (2003–2006) who wore accelerometers and completed pertinent survey questions. Ordinary Least Square models were computed to examine the relationship between income and the dependent variables (sedentary behavior, light, and moderate to vigorous intensity activity) adjusting for covariates. Logistic regression was employed to examine the association between income and meeting physical activity guidelines during a 2-day and 7-day time-period. Results indicate that individuals with an annual income of ≥$75,000 engaged in 4.6 more daily minutes of moderate to vigorous activity (p-value < 0.01), in comparison to the reference group (<$20,000 annual income). Those in the high-income strata were 1.6 and 1.9 times more likely to meet physical activity guidelines during a 2 and 7-day period (respectively) than their lower income counterparts (p < 0.05 for both). Further, those in the high-income strata spent 11.8 more minutes daily being sedentary than their lower income counterparts (p-value < 0.01). In conclusion, higher annual household income is related to more intense, less frequent (per week) patterns of physical activity and more daily sedentary time.
Conference Paper
Personality-based personalized systems are increasingly gaining interest as personality traits has been shown to be a stable construct within humans. In order to provide a personality-based experience to the user, users' behavior, preferences, and needs in relation to their personality need to be investigated. Although for a technological mediated environment the search for these relationships is often new territory, there are findings from personality research of the real world that can be used in personalized systems. However, for these findings to be implementable, we need to investigate whether they hold in a technologically mediated environment. In this study we assess prior work on personality-based music genre preferences from traditional personality research. We analyzed a dataset consisting of music listening histories and personality scores of 1415 users. Our results show agreements with prior work, but also important differences that can help to inform personalized systems.