Content uploaded by Gloria Mark
Author content
All content in this area was uploaded by Gloria Mark on Dec 23, 2016
Content may be subject to copyright.
Capturing the Mood: Facebook and Face-to-Face
Encounters in the Workplace
Gloria Mark1, Shamsi Iqbal2, Mary Czerwinski2, Paul Johns2
1Department of Informatics
University of California, Irvine
Irvine, CA 92697 USA
gmark@uci.edu
2Microsoft Research
One Microsoft Way
Redmond, WA 98052 USA
{shamsi,marycz,Paul.Johns}@microsoft.com
ABSTRACT
What makes people feel happy, engaged and challenged at
work? We conducted an in situ study of Facebook and face-
to-face interactions examining how they influence people’s
mood in the workplace. Thirty-two participants in an
organization were each observed for five days in their
natural work environment using automated data capture and
experience sampling. Our results show that online and
offline social interactions are associated with different
moods, suggesting that they serve different purposes at
work. Face-to-face interactions are associated with a
positive mood throughout the day whereas Facebook use
and engagement in work contribute to a positive feeling at
the end of the day. Email use is associated with negative
affect and along with multitasking, is associated with a
feeling of engagement and challenge throughout the day.
Our findings provide initial evidence of how online and
offline interactions affect workplace mood, and could
inform practices to improve employee morale.
Author Keywords
Mood; affect; Facebook; face-to-face interaction;
multitasking; email
ACM Classification Keywords
H.5.3 [Information Interfaces and Presentation (e.g., HCI)]:
Group and Organization Interfaces; K.4.m [Computers and
Society]: Miscellaneous.
INTRODUCTION
Mood in the workplace has long been a subject of interest
in fields ranging from organizational science to psychology.
In the field of CSCW, although some attention has been
given to how interactions in general influence mood (e.g.,
[44]), and how affect is conveyed through computer-
mediated interaction (e.g., [20]) and social media [4], little
attention has been given to how workplace interactions
affect mood. This is important, as mood has been shown to
impact performance in the workplace [3].
Understanding how such interaction affects mood in the
workplace is important for organizations as both online and
offline interactions continue to be integral parts of the daily
routine of workers. Social networking sites (as well as other
social media) are being deployed increasingly more in
organizations and they amplify opportunities for online
interactions.
In a work environment, it is an open question how face-to-
face and online interactions compare in affecting mood. To
our knowledge, this has never been investigated before. On
the one hand, we might expect online interactions to
positively affect mood, as has been found with face-to-face
interaction in field experiments [43, 44]. On the other hand,
interactions could be associated with work interruptions and
task demands which have been shown to increase stress [37,
38], which in turn could negatively impact mood. In this
paper we investigate how face-to-face and online social
network interactions influence mood in the workplace.
These interactions typically occur in an environment where
task switching and interruptions are prevalent [10, 17]; we
also investigate how this multitasking context affects mood.
To gain a deeper understanding of how online and offline
social interactions affect people’s mood at work, we
conducted an in situ study at a large U.S. corporation. Our
focus was to examine the influence of both face-to-face
interaction and Facebook use on a breadth of workplace
affective states. We chose Facebook because of its reported
versatility and fairly high adoption in the workplace [50].
This research is part of a larger project: WorkSense, which
has the goal of understanding people’s workplace behavior
via automated data capture and other methods. We found
that online and offline social interactions are associated
with different mood experiences suggesting that they serve
different purposes at work. Face-to-face interactions are
associated with people being happy throughout the day
whereas Facebook use and being engaged in work
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted with out 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 ow ned by others than ACM must be honored.
Abstracting with credit is permitted. T o copy otherwise, or republish, to
post on servers or to redistribute to lists, requires prior specific
permission and/or a fee. Request permissions from Permissions@acm.org.
CSCW'14, February 15 - 19 2014, Baltimore, MD, USA
Copyright 2014 ACM 978-1-4503-2540-0/14/02…$15.00.
http://dx.doi.org/10.1145/2531602.2531673
contribute to an overall positive feeling at the end of the
day. Our results showing the effects of online and offline
social interactions on mood may have implications for
workplace practices to improve employee morale.
EFFECTS OF SOCIAL MEDIA USAGE ON MOOD
Pew reports show that in 2012, 67% of internet users said
that they used at least one social network site (SNS) [14].
This is over double the percentage from a similar study in
2008 (as reported in Hampton et al, [18] ). Over half of
these SNS users (71%) were female, and while the largest
demographic continues to be 18-29 years of age, 52% of
50-64 year olds were using SNS sites and 32% were over
65. Of these sites, Facebook has been the dominant site
used, with 92% of SNS users reporting they were on
Facebook. The Pew report also documents that between 10-
26% of users update their status, comment, “like” or send
private messages at least once a day. But more specific to
our study, how does the use of Facebook influence users’
emotional states? The Pew report finds that Facebook users
are more trusting than others, have more close relationships,
get more social support and are more politically engaged
than most people [18]. Does this mean that they are happier
or turn to Facebook when they need this kind of support?
Pea et al. [43] used a large, online survey of girls from
North America aged 8-12 and examined the relationships
between social well-being and young girls' media use—
including video, games, music, reading/homework, e-
mailing/posting on social media sites, texting/instant
messaging, and talking on phones/video chatting and face-
to-face communication. These researchers found a strong
negative association between personal communication in
media and social well-being. For example, using the phone,
chatting, listening to music, reading and especially
watching video were highly correlated with negative social
well-being. Media multi-tasking was also correlated with
negative social well-being. On the other hand, face-to-face
communication was strongly associated with positive social
well-being. While these findings have direct implications
for pre-teens’ social well-being, do they translate into mood
effects in the workplace?
Kramer (2012) did a large-scale study of emotional
contagion by studying Facebook status updates [30]. His
research suggested that if someone from your social
network made a status update with emotional content, the
friends of the poster are more likely to post in a similar
emotional vein. His original group of participants included
one million random Facebook users. He found that this
effect was long-lasting (lasting up to 3 days after the
original post) and held when controlling for emotional
expressions by posters and their friends in the past. This
study was important because it may have demonstrated the
contagion of emotion through indirect textual posts in social
networks (though of course it could not control for face-to-
face interactions with close friends that might also be your
friend in Facebook). The speculation proffered by the
author is that these public expression strategies might help
users maintain psychological health, by inviting friends to
share in each other’s joy or by gaining social support when
needed. It was also conjectured that sharing negative
emotions on Facebook might be important for garnering
feelings of closeness.
There may also be gender differences to consider. For
instance, Kivran-Swaine et al. [29] analyzed the language
used on Twitter exchanges in addition to aspects of users'
networks, to analyze the influence of gender on emotion,
while controlling for the strength of connection between the
users. Their findings showed that women expressed positive
emotions more than men, especially when exchanging
tweets with other women.
According to reasoning by Hampton and Wellman [19],
Facebook might enhance current, place-based community
and help generate social capital, which may help explain
why so many people use it on a frequent basis. Another
study has shown that using Facebook may actually boost
your self-esteem, largely due to users and their social
networks portraying their best possible selves [16]. Yet it is
unclear whether these factors could also positively impact
mood in the workplace.
Research questions
In this study we ask the following two research questions.
1) How are Facebook (FB) use and face-to-face (F2F)
encounters related to mood as it fluctuates throughout the
day? Mood varies throughout the day depending on
workload changes, task engagement, interactions, and
breaks that one may take. Measuring mood at a single point
in time fails to capture the dynamic nature of interactions at
work and how they affect mood. Continually tapping into
mood as the environment and context change can provide
us with insight into how specific phenomena (interactions,
computer use) affect mood. It also enables us to capture
mood while the experience is still recent in memory [21].
Therefore, in this research question we examine how online
and offline interactions may impact mood changes
throughout the day. As interactions may occur in a context
of dynamic task-switching [17] we also examine how
multitasking influences mood throughout the day.
2) How are the amount of FB use and F2F encounters over
the course of the day related to one’s mood state at the end
of the day? In this research question, we are interested to
examine whether FB or F2F encounters have a longer
temporal effect such that at the end of the day, one's mood
might be affected. Mood at the end of the day is important
to consider as mood in work and home life can have
carryover effects [56]. In particular, we are interested in
whether the amount of FB or F2F encounters could
influence a change in mood over the course of the day.
Again, as interactions occur in a context of multi-tasking,
we also examine factors related to task-switching.
VARIABLES TO ASSESS MOOD IN THE WORKPLACE
An assumption in studying mood or "feeling" is that it can
be broken down into different dimensions that are
consciously accessible [46]. A range of variables have been
used to assess mood [46]. As our focus was on interaction
and mood in the workplace, we considered the following
dimensions to be the most relevant: valence, engagement,
and challenge. Since personality's effects have been well
studied on mood, we felt it was important to consider
personality traits in our exploration. Finally, given the
potential of disruption to ongoing work caused by
interactions, we incorporate prior knowledge on how task-
switching behavior may impact a user's affective state.
Dimensions of mood
We selected Valence as a measure to capture the positive
(e.g., happy, upbeat) and negative (e.g., sad, gloomy)
affective dimensions of an emotion [5]. Informed by the
study of Pea et al [43], who examined media and face-to-
face interactions, we deemed it important to understand the
type of valence associated with offline and online
interactions. The valence measure has been validated in
assessing positive and negative affect. Steptoe et al. [51]
measured affect using a self-report sampling method (as in
the current study) and validated it with physiological
measures of cortisol samples and heart rate monitoring.
Positive and negative affect have been examined in a range
of different contexts in the workplace. See [3, 21] for
reviews.
We selected Engagement as a measure to capture the degree
to which people felt involved in or distracted from their
work, particularly relevant in information work where
people are constantly interrupted and switch tasks [10, 17].
We follow the definition of Schaufeli et al. [48] who
defines engagement as "a positive, fulfilling, work-related
state of mind that is characterized by vigor, dedication, and
absorption". Experience-sampling studies typically query
subjects with either 'engagement', 'concentration', or
'involvement' in an activity to capture the degree of
engagement [21]. Others have studied and visualized self-
reports of engagement to help users reflect upon their work
activities and mood using various sensors [40]. We were
motivated to build upon and extend this work. For a review
of how engagement has been used in self-reports, see [21].
We also focus on Challenge, as we felt that when someone
in the workplace is faced with juggling interactions and
work demands, then this could at times create a feeling of
being challenged, e.g., due to time pressure [37]. This may
especially be the case if online or offline interactions are
viewed as distractions as opposed to benefits [41]. One's
perception of 'challenge' can be dynamic, depending on the
activity; interacting with someone face-to-face during a
coffee break might yield a different feeling of challenge
compared to dealing with an interruption amidst other task
activities. We define challenge as the amount of mental
effort involved in performing an activity. The construct of
challenge in the workplace has long been emphasized for its
significant contribution to job satisfaction (e.g., [22]). The
feeling of challenge has been studied extensively and has
been validated as part of the experience of "flow", or
optimal experience [21].
Effects of personality on mood
Personality, as related to interactions and mood, has long
been a topic of interest, with recent studies beginning to
address online interactions. The Big Five dimensions of
personality have been widely employed as a measure [39],
as they are well-validated, and consistent and
comprehensive in scope [13]. The Big Five characterizes
personality using five different traits [39]. Agreeableness
refers to cooperative behavior, as well as deferring to others
during a conflict. Conscientiousness refers to the propensity
for planning and to seek high achievement. Extraversion is
the tendency to want to be with others, to have strong social
skills, and to seek social stimulation. Openness to
Experience refers to being open to change and variety and
seeking diversity. Neuroticism is the tendency to feel guilty,
depressed or anxious.
In terms of mood, Neuroticism is generally associated with
negative affective states while Extraversion is associated
with positive affective states (cf [54]). Some research has
addressed the relationship of personality with face-to-face
and online communication. Extraversion was shown to be
positively correlated with using the Internet to maintain
both face-to-face and remote friendships [53]. Facebook
users were found to score high on Extraversion [47], and
social network site use [6]. Conscientiousness has been
shown to be negatively correlated with time spent on
Facebook [47] while Neuroticism was found to be
positively correlated with time spent on Facebook [47].
Openness to Experience was positively correlated with
social network site use [6].
Thus, though positive and negative affect are generally
related to specific personality traits, the results relating
personality to communication, especially online social
network use, is mixed. As prior studies show correlations
with Extraversion, Conscientiousness and Openness, we
examined these personality traits in our study.
Effects of interruptions, email and task switching on mood
Broadly speaking, both Facebook and face-to-face
interactions typically interrupt the flow of ongoing work at
the workplace, which can in turn affect mood. A large body
of prior work has explored the effects of interruption on a
user’s affective state, mostly focusing on negative affect
such as frustration, anxiety and annoyance caused by
inopportune interruptions [1, 2, 24, 57]. McFarlane and
Latorella demonstrated that interruptions that are
unpredictable and cannot be controlled result in more stress
and affect task performance [41]. Mark et al. [37] also
showed that interruptions can lead to increased stress and
higher frustration.
On the other hand, external interruptions to ongoing work
caused by emails and instant messages have been shown to
have positive benefits including supporting near instant
communication [9, 11, 32], maintaining awareness of
peripheral information [34], reminding upcoming activities
[12] or helping users perform complex tasks [33, 45]. Such
benefits can lead to positive affect. Self-interruptions can
also have positive or negative impact on an individual [27].
Other studies have looked at how interruptions from digital
media such as email or IM notifications, or physical
interruptions such as phone calls or face-to-face
interactions, cause people to continue to be distracted by
switching tasks with effects on task performance [10, 25,
26]. Notifications also are associated with emotional
experiences and are correlated with types of notifications
people would like to receive in the future [42]. Email use in
particular has been found to be associated with an increase
in stress [37, 38].
Thus, it is unclear how FB and F2F interactions, and digital
media use in general, might influence mood in the
workplace. To the best of our knowledge, no study has
looked at the effects of social interactions on mood in the
context of the workplace. As it is an open question, based
on this prior body of work, we examine FB and F2F
interaction, as well as task switching of applications and
documents, along with email usage. Our broader goal is to
understand factors that impact mood at work.
RESEARCH STUDY
To understand effects of FB and F2F interactions on mood
in the workplace, we conducted an in situ study. The
research was conducted in the fall of 2012 at a large U.S.
corporation.
Participants
Participants were recruited primarily from volunteers who
responded to email advertising done throughout a research
division in the company. The rest of the participants were
recruited through convenience sampling or snowball
sampling, (i.e., recommendations of names from people
who participated.). Our criteria for recruitment were that
people should be FB users and that they use the Windows 7
operating system (which was compatible with our logging
software). Thirty-two people (17 females, 15 males),
volunteered to participate. This sample is over double that
of samples used in similar in situ work tracking studies [10,
17, 38]. We feel that the data collected from 32 people over
five full work days was sufficient for enabling us to gain a
representative sample of their mood in their daily work
environment.
All participants were knowledge workers. Most participants
were involved in research (15), but there were also
managers, admins, engineers, a department director, a
designer, and consultant.
Methodology
Following a paradigm of precision workplace shadowing
[17, 38], our goal was to capture as complete a picture as
possible about online workplace behaviors and mood. Each
participant was observed for a period of five days. For most
participants, this was Monday through Friday, i.e., a regular
work week. Some participants traveled during the week or
missed a day for other reasons; in these cases, they made up
the missed day (in most cases) the following week.
We used mixed methods for our data collection: computer
logging was used to capture online actions, a wearable
SenseCam camera [23] was utilized to capture face-to-face
interactions, experience sampling probes were used to
capture self-reported moods throughout the day, and a
series of surveys were collected for other subjective and
demographic measures. On the day before the participant
began the study (usually a Friday), the computer logging
software and experience sampling software were installed
on their computer. The study and tracking approaches were
explained to the participant. Participants were assured that
their data would be private and aggregated, that no content
would be retraced to their information, and that they would
remain anonymous. They were also assured that their data
would be deleted in a timely manner.
Participants were instructed to work as they normally would
throughout the workday. They were instructed to answer
the (experience sampling) probes when the probe windows
popped up on their computer screens, but they also could
cancel the probe window when they chose. We emphasized
that it was important to answer the probe questions as
accurately as possible. Participants were told that at any
time they could turn off the SenseCam camera. In the post-
study interviews, 11 participants reported that they turned
the camera off when they left the office for brief periods
(e.g., for meetings, a demo, and a doctor’s office). In these
cases, we did not collect probe or logging data either. For
two people we lack SenseCam data for ½ day each.
Measures
Below we explain the details of the measures taken. Table 1
provides a summary of the measures and their explanations,
along with abbreviations used to refer to them in the paper.
Data collection: sensors and experience sampling
In collecting data we had to decide between using highly
precise sensor-based data and human observation data
which is more effective in capturing the context of the
participant's work. Previous ethnographic studies that track
workplace behavior, e.g. [17], while capturing rich data, are
very labor intensive for capturing precision data, i.e., users’
actions tracked to the second. Automatic data collection by
sensors, though lacking contextual information and
intentional data that ethnography provides, enables the
collection of a wider array of data, with more participants
simultaneously, and at a higher precision level than human
ethnographic observation. Thus, the choice of using sensors
to capture human behavior as opposed to ethnographic
approaches involves a tradeoff between precision and
contextual richness. Since for this study the precision of
data was central to understanding mood changes we chose
for our methodology the use of sensors. To compensate for
the lack of context that human observation could have
provided, we collected participant self-reports to
supplement our automated data with the participant's
perspective. We collected the following data:
Mood throughout the day: Experience sampling
To collect mood data throughout the day we used the
experience sampling method. Experience sampling is
designed to capture people's perception of daily life as it
changes throughout the day. This method has been proven
to have internal validity [8] as well as external validity [21].
Experience sampling works well paired with computer
logging as it provides information on how the participant is
experiencing the events and context. Experience sampling
has been used in a large number of studies and particularly
has been used for measuring mood in the workplace: some
example studies examined work and home life balance [56]
and time pressure [52]. For a review see [21].
Experience sampling was done with custom built software
that presented a probe, a small pop-up window on the
computer screen, to participants using predetermined
sampling rules. We used a hybrid interval-contingent and
event-contingent sampling approach [21]. The sampling
occurred whenever a user left email after being active in
that application for at least three consecutive minutes, or in
FB after a full minute of uninterrupted use. Sampling also
Measure
Explanation
Abbrev.
Computer Logging
FB use (seconds)
The amount of time spent in a web browser where a Facebook page
is in the foreground tab.
FB
Email use (seconds)
The amount of time an email is open and in the foreground, whether
one is reading a received mail, composing a new mail, or replying or
forwarding.
Email/Use
Email/Calendar
(seconds)
The amount of time any part of the Email/Calendar Application is
open and in the foreground, including email, calendar, contacts,
tasks, etc. Email usage in Email/Cal is distinct from usage in
Email/Use. Email/Use refers to operations of reading and writing,
whereas Email/Cal refers to simply viewing the inbox.
Email/cal
Document switches
(counts, per unit of
time)
The number of document switches within an application, e.g. within
Word or Excel or Internet Explorer. In web browsing, each new
page is a document switch.
Doc
Application switches
(counts, per unit of
time)
The number of switches between applications, e.g., from Internet
Explorer to Word.
App
SenseCam
F2F interactions
(counts)
The SenseCam captured images on average once every 15
seconds.
F2F
Experience Sampling Probe
Valence (-200 to
+200. Neutral = 0)
Ranging from negative to positive affect.
Valence
How Engaged
(0=not at all to 5=high)
The extent to which people feel involved with or distracted from
work
Engaged
How Challenged
(0=not at all to 5=high)
The amount of mental effort involved in performing an activity
Challenged
Surveys
PANAS
Rating scale of mood of positive (PA) and negative (NA) affect
dimensions [55]; deployed at beginning (BEG) and end (END) of
each day
PA-BEG, NA-BEG
PA-END, NA-END
Big 5 traits
Personality inventory [39]
Extroversion
Neuroticism
Conscientiousness
Demographic info
Age, gender, job role, education
Table 1. Summary table of measures, explanations and their abbreviations.
was triggered whenever a user logged into Windows or
unlocked the screen saver (event-contingent). If fifteen
minutes passed without a sampling, then a probe was
triggered (interval-contingent).
The probe presented the instructions “Please rate how you
feel right now”. We used rating scales as is commonly used
in experience sampling approaches [21]. To measure
Valence, participants saw a sliding scale which
corresponded to a range of -200 (negative affect) to +200
(positive affect) and were asked to click with their cursor on
that point that best expresses their feeling "right now". To
measure Engagement, participants were asked 'How
Engaged Were You?' using a 6-point Likert scale (0=Not at
All; 5=Extremely). To measure Challenge, participants
were asked 'How Challenged Were You?' using the same
Likert scale. Participants were asked if they just had any
face-to-face interactions, and if so, a second screen was
shown, asking whether the participant had a scheduled
meeting. The timestamp when participants submitted the
probe was recorded.
Beginning and end of day mood: PANAS
To answer the second research question, we needed to
collect overall mood data at the beginning of the day and at
the end of the day so that we could quantify any changes.
We deployed surveys at the beginning and end of each day
to measure mood using the PANAS scale [55], a well-
validated 20-item inventory of mood which is comprised of
two scales to measure positive and negative affect. Items
included feelings such as interested, excited, distressed,
upset, and irritable. Participants were asked to rate to what
extent they felt that way at the present moment on a scale
ranging from very slightly/not at all to extremely.
Facebook interactions: Computer logging
Online interactions were logged with a custom-built
application that captured all activity in the Windows
Operating System. Captured activity includes beginning
and end times for the lifespan of every window, and the
beginning and end times for every instance of every
foreground window. Computer sleep mode, mouse and
keyboard activity were also logged, so that periods of time
in which an application was in the foreground window, but
the user was not actively using the computer, could be
ignored.
Note that we were unable to capture activities that occurred
within a window itself, e.g., capturing what a person was
looking at while on FB or any other application due to
privacy and technical limitations. While there are APIs that
can collect data on what one posts – statuses, comments,
photos, etc., there is no public API that allows one to track
actions to the second on Facebook. Even such APIs would
provide access to only a small fraction of what one does on
Facebook and would miss interactions that involve
scanning for awareness which can be only detected via
over-the-shoulder shadowing. We therefore focused on
measuring the total time one spends on FB as a holistic
measure of FB interaction, though we hope to be able to
capture more fine-grained interaction data in future
research.
Face-to-face interactions: SenseCam
F2F interactions were measured through SenseCam [23], a
lightweight wearable camera that participants wore around
their necks. The camera automatically takes a picture and
stores it locally, and as soon as the image is processed and
saved the next picture is taken. The average length of time
between photos is 15 seconds. SenseCam images were
processed by a face detection module, a publicly available
application produced by Microsoft Research
(http://research.microsoft.com/en-us/projects/facesdk/). It is
important to note that with this software, we cannot
distinguish whether the faces were the same person or not.
Therefore, the counts in our F2F variable should be
considered as a proxy for amount of F2F interaction over
the course of the day, and not necessarily distinct
interactions. It is thus a measure of how much interaction a
person experienced, and not precisely how many different
interactions one engaged in.
RESULTS
We first present an overview of results of the data collected,
the probe responses measuring mood, and then an overview
of FB and F2F interactions.
Data overview
Our 32 participants were observed for 5 days each, for a
total of 160 person-days, or a total of 1,509 hours of data
collection. Our computer logging software collected a total
of 91,409 computer window switches. We collected 2,809
experience sampling probes and analyzed 204,922
SenseCam photos.
Mood: probe responses
Our 32 participants averaged 17.56 probe responses per
day, for an average of 87.8 probe responses per participant.
The average Valence rating over participants was 38.83 (on
a scale of -200 to +200), showing a net positive affect.
Females (M=44.78, sd=74.98) reported significantly higher
Valence than males (M=32.75, sd=62.30), t(2807)=4.62,
p<.0001. The average Engaged rating over participants was
3.01 (sd=1.37), on a scale of 0-5 (high). Females (M=3.09,
sd=1.37) reported being significantly more Engaged than
males (M=2.93, sd=1.38), t(2808)=9.25, p<.004. The
average Challenged rating over participants was 1.82
(sd=1.42) on a scale of 0-5 (high). There was no significant
difference between females (M=1.78, sd=1.36) and males
(M=1.86, sd=1.47), t(2807)=1.43, p<.15, in feeling
Challenged. Thus, in our sample females reported higher
positive Valence, and reported to be more engaged in their
tasks than males, throughout their workday.
FB and F2F interactions
We next present an overview comparison of FB and F2F
interactions over the course of the day. Table 2 shows
average daily FB use in seconds and F2F interactions,
broken down by age and gender.
A 2 (Gender) x 3 (Age) ANOVA conducted on the
dependent variable of FB shows a significant Gender
difference (F(1, 161)=5.39, p<.02), and a trend showing
Age x Gender interaction (F (2, 161)=2.64, p<.08). Age is
not significant. Females use FB over twice as long on an
average day as males.
A 2 (Gender) x 3 (Age) ANOVA conducted on the
dependent variable of F2F interactions shows a significant
effect of Age (F(2, 162)=5.19, p<.007, but no Gender, or
Age x Gender interactions. In sum, people in the age range
of 30-40 had the most F2F encounters during the day on
average, with people under 30 having the least. In contrast,
people under 30 had the most FB usage, with ages 30-40
the least. Work roles could somewhat explain the
difference. Participants under 30 were all researchers
(researcher, intern, postdoc) whereas participants 30-40 had
more of a range of positions (researcher, intern, manager,
admin, designer, and consultant). Younger participants may
use social media more than their older counterparts, and this
might have an effect on their number of F2F interactions.
Comparing activities in 5-minute time units, time spent in
FB is weakly correlated with both App switching (r=.09,
p<.0001) and Doc switching (r=.11, p<.0001), N=37,788.
F2F is very weakly correlated with both App switching
(r=.05, p<.0001) and Doc switching (r=.08, p<.0001),
N=37,788.
Mood throughout the day: Valence, Engagement, and
Challenge
Our first research question addressed how FB use and F2F
interactions influence mood throughout the day. We report
the results of three mood states: Valence (positive and
negative affect), and feelings of Engagement and Challenge
in the current activity.
We collected data on the same participant for five days. To
account for the nested interdependence in our data, we ran
linear mixed models in SPSS using random and fixed
effects, and did this for all three mood measures. As we had
no a priori notion of what variables might be associated
with mood change, the ideal procedure would be an
automatic model fitting based on selecting those variables
that would result on the best fitting model. As SPSS does
not have an automatic model building procedure for linear
mixed models, we built the models by hand. We emulated a
backward elimination procedure as in stepwise regression,
where we started with all variables in the model and then by
hand gradually eliminated different combinations of
variables until we found the best fitting model based on the
BIC criterion1. We tested variable measures in different
time units (1, 5 and 10 minutes) prior to each probe
response and included the results that showed the strongest
correlation in the model.
An R2
statistic for linear mixed models must account for the
variance explained by both the fixed and random effects;
however, there is no standard method for specifying an R2
in these models [15]. To provide a sense of how much
variance the model explains, we ran a linear model
including only fixed effects to get an R2 value. By not
including random effects (participants), this of course will
underestimate the amount of variance explained but we feel
it is a reasonable estimate since the random effects are
small.
We tested models that included all the variables shown in
Table 1. As some variables were not normally distributed,
we did a log transformation on these variables: F2F, FB,
Email/Use, Email/Cal, App and Doc.
Influencing Valence throughout the day
We first report the results of the model for Valence change
over the course of the day. Table 3 shows the independent
variables in the model that best fit Valence as a dependent
variable. The model shows that the more F2F interaction in
the 5 minutes prior to the probe (i.e., F2F counts as
1 In linear mixed models, Schwarz’s Bayesian Criterian (BIC) is
the criterion used to find the best fitting model [49.]. The lower
the score, the better the fit of the model. The absolute number
itself is not meaningful—the BIC is used to compare between
models and is a well-established metric of model selection.
Valence model β t p
Intercept
-32.37
-1.06
.30
F2F prior 5 min. (counts)
4.90
2.44
.03
Email/Use prior 5 min.
(sec.)
-3.41
-2.85
.005
Big 5 Extroversion
2.69
2.32
.03
Table 3: Beta coefficients of variables for best fitting model
for Valence throughout the day. N=2809 cases.
N
FaceBook (sec.)
F2F (SenseCam
counts)
Females
17
715.14 (1662.65)*
78.51 (81.16)
Males
15
320.33 (827.64)*
74.62 (90.04)
Age
< 30
6
941.83 (2536.02)
37.33 (56.13)**
30-40
16
425.16 (830.71)
93.83 (95.62)**
> 40
10
454.58 (830.71)
72.18 (73.55)**
Overall
mean
32
529.92 (1348.19)
162 (76.69)
Table 2. Means (S.D.) of daily FB use and F2F interaction,
as measured by SenseCam counts, N=32 subjects, each
measured over five days, **=p<.001, *=p<.05.
measured in the SenseCam photos), the higher (i.e., more
positive) the Valence measure. Email/Use, on the other
hand, shows an inverse relationship with Valence: the lower
the seconds of email use in the 5 minutes prior to the probe
(i.e. reading or writing emails), the more positive the
Valence measure. There is also a significant positive
relationship with the Big 5 Extroversion personality score
and Valence: the more extroverted participants are, the
more positive is the Valence measure. There were no
significant interactions. FB was not found to impact
Valence. There was no significant difference in Valence if
one had a scheduled meeting or not before the probe. The
R2 of a linear model of the fixed effects alone is 14.4% (see
explanation above).
Influencing Engagement throughout the Day
We next focus on variables that might influence
Engagement over the course of the day. Our dependent
variable was Engagement, as measured by participants’
responses to the probe, “How Engaged Were You?” on a 6-
point scale ranging from 0=’not at all’ to 5=’extremely’.
Table 4 shows the best fitting model.
Here, the model shows that the fewer seconds of FB use
prior to the probe, the more Engaged one reports. In
contrast, the more F2F interaction prior to the probe, the
more Engaged one reports to be. Contrary to Valence, the
more time spent in Email/Use, the higher the reported
engagement. The more task switching (as measured by
App), the more engaged one reports to be. Email/Cal use is
inversely related to Engagement. Interestingly,
Conscientiousness, from the Big 5 personality inventory, is
positively associated with Engagement. No interactions
were significant. Though linear mixed models in SPSS do
not report multi-collinearity, a regression analysis of these
factors shows all variance inflation factors (VIF) to be <1.1,
indicating that multi-collinearity is not a problem. The R2
of a linear model of the fixed effects alone is 30.6%.
Influencing Challenge throughout the Day
What leads people to feel challenged throughout the day?
Our dependent variable was Challenge, as measured by the
probe question: 'How Challenged were you?' using the 6-
point Likert scale. Table 5 shows the independent variables
that produce the best fitting model.
The model shows that FB use is inversely related to
Challenge: the more one uses FB, the less challenged one
feels. Email/Use is positively related to Challenge: the more
time spent email reading/writing, the more challenged one
self-reports. Task switching (as measured by App switches)
is positively related to feeling challenged. However,
Email/Cal use is inversely related to feeling challenged.
One’s negative mood at the beginning of the day (as
measured by the PANAS NA BEG), is positively associated
with feeling challenged throughout the day.
We also find a significant Age x Gender interaction but no
other interactions. In Table 5, the beta coefficients show
that females 30-40 feel positively challenged; all other Age
x Gender levels show that participants feel negatively
challenged. As Table 6 shows, for those under 30, males
Engagement model
β
t
p
Intercept
1.97
3.32
.002
FB use prior 10 min. (sec.)
-.10
-3.29
.004
F2F prior 5 min. (counts)
.08
1.93
.05
Email/Use prior 5 min. (sec.)
.16
5.63
.0001
App prior 10 min. (counts)
.10
3.03
.002
Email/Cal prior 10 min. (sec.)
-.08
-3.60
.0001
Big 5 Conscientiousness
.04
2.05
.05
Table 4: Beta coefficients of variables for best fitting
model for Engagement throughout the day. N=2809 cases.
Challenge model
β
t
p
Intercept
1.07
2.70
.007
FB use prior 10 min.
(sec.)
-.18
-3.99
.001
Email/Use prior 5 min.
(sec.)
.11
2.98
.005
App prior 10 min.
(counts)
.08
2.48
.013
Email/Cal prior 10
min. (sec.)
-.06
-2.89
.004
PANAS NA BEG
.86
2.88
.004
Age X Gender
<30, F: -.71
1.65
.04
<30, M: -.18
30-40, F: .07
30-40, M: -.68
>40, F: -.80
>40, M: 0*
Table 5: Beta coefficients for best fitting model for Challenge
throughout the day. N= 2809 cases. *=This parameter is set to
0 because it is redundant.
Age
Gender
Mean (SE)
<30
F
1.38 (.27)
M
1.91 (.40)
30-40
F
2.16 (.16)
M
1.41 (.24)
>40
F
1.29 (.37)
M
2.09 (.21)
Table 6: Mean (SE) of Age x Gender levels that influence
How Challenged one is throughout the day.
report being more challenged than females; for those 30-40,
females are more challenged than males; and for those over
40, males are more challenged than females. The variance
inflation factor is < 1.8 for all variables indicating that
multi-collinearity is not a problem. The R2 of a linear model
of the fixed effects alone is 39.6%.
We summarize the results of our first research question.
The more F2F interaction one has, the more positive is
one’s Valence rating and the more one feels Engaged. The
longer one spends in FB, the less Engaged and Challenged
one feels. Length of time of Email/Use is consistently a
strong influence on all three types of mood: negatively
related to Valence, and positively related to Engagement
and Challenge. Task switching is positively related to both
Engagement and Challenge. If one starts out the day with a
negative affect, then this shows a spillover effect in making
people feel more challenged throughout the day.
Personality traits are also associated with mood: the higher
the Extroversion score, the more positive the Valence, and
the higher the Conscientiousness score, the higher the
Engagement rating.
Changes in mood at the end of day
Our second research question asked what factors would be
associated with a person’s mood at the end of the day.
Would cumulative time in FB and amount of F2F
interactions have an effect on one’s mood at the end of the
day? What other types of user actions throughout the
workday make one feel more positive or negative by the
end of the day?
Positive mood at the end of the day
To investigate positive mood at the end of the day, we
looked at the change in positive affect over the course of
the day, computing a dependent variable based on [PANAS
PA-END score – PANAS PA-BEG score]. We were
particularly interested to see if FB use or F2F interactions
were associated with a more positive mood at the end of the
day compared to the beginning of the day. We examined a
dataset consisting of each participant’s measures totaled for
each day. We ran a stepwise regression in SPSS entering
the following variables, controlling for individual
differences: FB, F2F, Email, Email/Cal App, Doc, App,
Age, Gender, Sociability, FB Import, Engaged, and
Challenged. Individual differences were not found to be
significant. The following model best fit the data, as shown
in Table 7: F(2,99)=9.99, p<.0001, R2=16.8. There is no
significant Engaged x FB interaction.
This model shows that the more one feels engaged in their
activity during the day, the more positive one feels at the
end of the day (compared to the beginning of the day). FB
also plays a role in influencing affect over the course of the
entire day: the longer one spends in FB over the course of
the day, the greater the increase in positive affect. Time in
FB contributed 5.3% of the R2, i.e. the variance explained,
of the change in affect.
Negative mood at the end of the day
To investigate what might be associated with negative
affect at the end of the day, we computed a dependent
variable based on [PANAS NA END score – PANAS NA
BEG score]. Using the same dataset as positive mood at the
end of the day (see above), and entering the same variables
into a stepwise regression analysis, we found none of our
measures to significantly influence negative mood at the
end of the day.
DISCUSSION
In this study we examined the open question of how FB use
and F2F interaction influence workplace mood. We studied
influences on mood from two perspectives: how mood
fluctuates throughout the day and how mood is experienced
at the end of the workday. Our study provides initial
evidence that FB and F2F both influence positive affect,
albeit in different ways. Moreover, workplace interactions,
as discussed, generally occur within a broader context of
other task activity and we found that email and task-
switching influence workplace mood as well. Demographic
and personality variables also contribute to explaining
workplace mood.
Our results suggest that online and offline interactions serve
different purposes in the workplace in terms of influencing
mood. F2F interactions are associated with positive affect
throughout the day whereas amount of FB use contributes
to an overall positive feeling at the end of the day. Our
results further show that throughout the day, F2F
interactions were positively associated with Engagement
whereas FB use was inversely associated with Engagement
and Challenge. F2F showed no relation to Challenge.
Our findings show that when people are engaged in F2F
interactions, it makes them feel more positive. F2F
interaction involves different stages: an opening phase (e.g.,
greeting, adjusting proximity), the interaction, and a closing
phase (parting rituals), which contributes to creating a
social commitment to some degree [28]. FB, on the other
hand is negatively associated with engagement and
challenge. FB is an online interaction that can be done
quickly, in a "grazing" fashion, without a greeting,
involvement, or closing stage. This can explain why FB is
not associated with high engagement or challenge, though
of course our users could have chosen to go to FB precisely
because they were in a state of low engagement in the first
place!
Positive End of Day Mood
β
t
p
Constant
-13.13
-4.31
.0001
How Engaged
3.76
3.89
.0001
FB Use (seconds)
.778
2.50
.01
Table 7: Beta coefficients to model the change in positive
affect at the end of the day.
To understand the FB results more holistically (i.e., that it
influences positive affect at the end of the day, and involves
low engagement and low challenge), we can be informed by
the results of the Engagement variable, also associated with
positive affect at the end of the day. Engagement was a
fairly strong influence of end of day positive affect,
explaining most of the variance. In our post study
interviews, nearly all participants reported that being
productive puts them in a good mood. Engagement in work
could be equated with feeling productive.
How are FB use and Engagement then tied together to
influence positive affect? FB use is weakly correlated with
both App switching and Doc switching which suggests it
may be used in a context of high task switching. Consistent
with the idea of grazing, people can quickly move in and
out of FB. As FB was not associated with high engagement
or challenge, together with its use during task switching, it
suggests that FB may be a "light" interaction experience. If
people are also highly engaged in their work for that day,
then FB could serve the purpose of offering a “break” from
other work. High engagement in work, along with light
breaks as FB affords, contribute then to people being in a
positive mood at the end of the day.
Negative affect at the beginning of the day, as measured by
the PANAS survey, also influenced feeling challenged
throughout the day. This spillover effect that we found
extends the work of Marco et al. [35] who, using an
experience sampling study, found that a person's negative
personality disposition leads to a feeling of distress in
handling events throughout the day.
What makes people feel happy at work? Our results thus
suggest that having F2F interactions and being engaged in
work influences positive affect. However, "light"
interactions of FB (as measured by negative challenge and
engagement) are also important in contributing to a positive
affect at work. Our result showing that F2F influences
positive affect extends the findings of Pea et al. [43] who
focused on moods of adolescent girls. Our results in an in
situ workplace environment also show that F2F encounters
impact positive affect.
Email, task-switching and mood
While our main focus was on examining FB and F2F
interaction and mood, it is important to consider that these
activities are done within a context of digital media use in
the workplace. Email was a factor that surfaced as
significantly influencing all three mood measures
throughout the day. We found that the more time spent
reading and answering emails (Email/Use) throughout the
day, the lower was one’s positive affect. Studies of email
use have uncovered that it leads to stress [37, 38]. Our
study additionally shows that reading and responding to
emails is associated with feeling engaged and challenged.
Responding to email may disrupt ongoing work, making
resumption of tasks more challenging [26]. Email also
influences negative affect. Put simply: reading a lot of
email at work puts people into a "bad mood".
If we consider email as a communication tool we can
examine how it compares with F2F and FB as
communication mediums. F2F interaction and Email/Use
both elicit similar results of feeling engaged in the
workplace. F2F requires a degree of engagement to attend
to verbal and nonverbal information and also to respond to
the conversation partner [28]. Reading and responding to
emails also requires a certain degree of engagement as it
involves communicating with another.
On the other hand, FB and Email/Cal both showed similar
results of being inversely related to feeling engaged or
challenged. Checking one's Inbox and calendar (Email/Cal)
are aspects of task and time management which involves
gaining quick awareness. FB actions of reading status
updates can also be used to gain a quick awareness of
friends' status. Future research could distinguish FB
activities in a more fine-grained manner to examine
whether distinct activities are associated with different
aspects of mood.
Task switching, as measured by App switching, was
positively related to both Engagement and Challenge. We
might expect that when one is task switching one would be
engaged in this activity, as this involves constantly shifting
focus and it can be challenging to reorient back to an
interrupted task [17]. This result extends previous work on
multitasking which shows it is related to stress [37].
Personality and mood
We found that personality traits influence mood throughout
the day. Extraversion has been previously found to be
associated with positive affect in single self-reports [31, 54]
Our study demonstrates that Extraversion influenced
Valence using continual measurements in the context of a
dynamic workplace. We also found that Conscientiousness
was positively related to Engagement as it was measured
throughout the day. Though no personality study has ever
directly addressed Engagement, we would expect these
results. According to the Big 5 Inventory, self-discipline
and achievement-striving are facets of Conscientiousness
[7] which we expect are related to feelings of Engagement.
Therefore, our results contribute to studies of personality by
showing it influences mood as a person's context changes.
Gender and mood
Females use FB over twice as long on average per day as
males. Females report over the course of the day as having
significantly more positive Valence and they are
significantly more engaged in work than males. Our results
extend the findings of Kivran-Swaine et al. [29] who found
that females express more positive emotions than males
with Twitter use. Our results together suggest that females
in the workplace use FB more and are happier (though these
two results do not imply causality).
We found an Age x Gender interaction with the Challenge
measure: in both the younger and older age groups, males
report being more challenged than females; in our mid-
range age group, females report being more challenged than
males. We find this result important as it reveals that gender
and mood effects in the workplace are related to age. We
hope that this result can lead to more detailed examination
of gender differences and work.
Alternative explanations to mood rating
We consider alternative explanations to our mood results.
Demand characteristics are always a potential explanation
of the results, i.e., that participants rated their mood
according to how they believed the researcher wanted them
to behave. However, we do not believe that demand
characteristics played a major role. First, it was not clear to
participants what kind of mood they should expect to
report. Second, there was variability in mood assessments.
If people presumably believed that the researcher was
looking, for example, for positive (or negative) affect, we
would expect that mood would be rated consistently
throughout the day, and it was not.
Another possible explanation of our result showing no end
of day mood effect for F2F is that F2F interaction creates
positive feelings that dissipate very quickly or are not
remembered at the end of the day. FB, on the other hand,
could create feelings that are longer lasting throughout the
day. A mechanism that could lead to long lasting feelings is
that FB use constantly reminds people of close friends who
are remote. More research is needed to test the notion of the
temporal span of feelings with F2F interaction and FB use.
Sensors and human observation
A contribution of our study was to show that sensors are a
viable means for capturing in situ workplace behavior.
Sensors enable the continual capture of behavior with
minimal disruption for the participant. Another advantage
of automated capture of behavioral data is that it can be
scaled up enabling the investigation of more complex
phenomena such as workgroup or even organizational
behavior. Though our technology did not enable us to
capture fine-grained details of FB use or types of F2F
interaction, we hope that our study can spark research in
this direction.
Limitations
Our participants were all highly educated (at least having a
Bachelor’s degree) and about half were researchers. They
thus represent highly skilled knowledge workers. We
therefore can only generalize our results to similar types of
workers. The use of researchers as participants has been
used in other activity tracking studies, e.g. [10, 38].
However, we believe that task characteristics of our
participants are very similar to what is found in many kinds
of information work: dealing with multiple tasks, deadlines,
and heavy reliance on information technology at work.
A known issue with the Experience Sampling method is
that the probe can be a source of interruption [21].
However, in the post-study interviews, participants did not
report this as a problem. We believe that the reason is
because the probe was able to be answered in seconds.
Another potential limitation is that raised by Columbetti [5]
who argues that a positive-negative valence scale misses
complexity as a measure of affect. However, studies such as
Steptoe et al. [51] have carefully validated this dichotomous
valence measure and we feel that the level of discrimination
was adequate for the goal of our study.
Another limitation was our use of the SenseCam to measure
face-to-face interaction. As explained, the SenseCam can
only be considered a proxy for amount of F2F interaction;
the face detection software that we used cannot distinguish
unique faces. Therefore, the F2F interaction measure should
be regarded as “amount” of interaction rather than distinct
interactions. This is similar to measuring FB use, as we
could not distinguish whether people were reading one post,
or scanning over many posts. The SenseCam also would
have failed to photograph faces if people were standing
sideways when they spoke to someone. It also does not
capture interactions that occurred within the 15 second
window of time between SenseCam photo shots. Therefore,
our measure of F2F interaction could have underestimated
interaction counts.
CONCLUSION
Our results showed that both F2F interactions and FB use
do influence positive affect in the workplace though
differently. F2F involves more engagement in interactions,
whereas the low engagement and challenge associated with
FB use is consistent with a lightweight interaction that
contributes, together with engagement in work, to making
people feel good by the end of the day.
Our findings provide initial evidence of how social
interactions affect mood at the workplace, suggesting the
usefulness of incorporating social media platforms in the
workplace as well as in promoting informal workplace
interactions. This has important implications for decision-
makers in the workplace who wish to balance interactions
and task engagement with positive affect of employees.
ACKNOWLEDGMENTS
This material is based upon work supported by the National
Science Foundation under Grant Number 1218705. We
thank Munmun de Choudhury and Scott Counts for their
valuable comments.
REFERENCES
1. Adamczyk, P.D. and Bailey, B.P., If Not Now When?
The Effects of Interruptions at Different Moments Within
Task Execution. in Proceedings of the ACM Conference on
Human Factors in Computing Systems, (2004), 271-278.
2. Bailey, B.P. and Konstan, J.A. On the Need for
Attention Aware Systems: Measuring Effects of
Interruption on Task Performance, Error Rate, and
Affective State. Journal of Computers in Human Behavior,
2006, 22 (4). 709-732.
3. Brief, A. and Weiss, H. Organizational behavior: Affect
in the workplace. Annual Review of Psychology, 2002, 53.
279-307.
4. Choudhury, M.D. and Counts, S. Understanding Affect
in the Workplace via Social Media Proceedings of
Computer Supported Cooperative Work, 2013, 303-315.
5. Columbetti, G. Appraising valence. Journal of
Consciousness Studies, 2005, 12 ((8-10)). 103-126.
6. Correa, T., Hinsley, A. and Zúñiga, H.G.d. Who
interacts on the Web?: The intersection of users’ personality
and social media use. Computers in Human Behavior, 2010,
26. 247-253.
7. Costa, P.T. and McCrae, R.R. Revised Personality
Inventory (NEO-PI-R) and NEO Five-Factor Inventory
(NEO-FFI) manual. Psychological Assessment Resources,
1992.
8. Csikszentmihalyi, M. and Larson, R. Validity and
reliability of the experience sampling method. . Journal of
Nervous and Mental Disease, 1987, 175. 526-536.
9. Czerwinski, M., Cutrell, E. and Horvitz, E., Instant
Messaging: Effects of Relevance and Timing. in People
and Computers XIV: Proceedings of HCI, (2000), British
Computer Society, 71-76.
10. Czerwinski, M., Horvitz, E. and Wilhite, S., A diary
study of task switching and interruptions. in Proceedings of
the ACM Conference on Human Factors in Computing
Systems, (2004), 175-182.
11. Dabbish, L. and Kraut, R.E., Controlling interruptions:
awareness displays and social motivation for coordination.
in Proceedings of the ACM Conference on Computer
Supported Cooperative Work, (2004), 182-191.
12. Dey, A.K. and Abowd, G.D., CybreMinder: A Context-
Aware System for Supporting Reminders. in Proceedings of
2nd International Symposium on Handheld and Ubiquitous
Computing, (2000), 172-186.
13. Digman, J.M. Personality structure: Emergence of the
five-factor model. Annual Review of Psychology, 1990, 41.
417-440.
14. Duggan, M. and Brenner, J. The demographics of social
media users - 2012. Pew Internet and American Life
Project, 2012,
http://www.pewinternet.org/Reports/2013/Social-media-
users/The-State-of-Social-Media-Users.aspx.
15. Edwards, L., Muller, K., Wolfinger, R., Qaqish, B. and
Schabennerger, O. An R2 statistic for Fixed Effects in the
Linear Mixed Model. Stat Med. , 2008 (27(29)). 6137.
16. Gonzales, A.L. and Hancock, J.T. Mirror, mirror on my
facebook wall: Effects of facebook exposure on self-
esteem. Cyberpsychology, Behavior and Social networking,
2011, 14. 79-83.
17. Gonzalez, V.M. and Mark, G., "Constant, Constant,
Multi-tasking Craziness”: Managing Multiple Working
Spheres. in Proceedings of the ACM Conference on Human
Factors in Computing Systems, (Vienna, Austria, 2004),
113-120.
18. Hampton, K., Goulet, L.S., Rainie, L. and Purcell, K.
Social networking sites and our lives. Pew Internet and
American Life Project, 2011,
http://pewinternet.org/Reports/2011/Technology-and-
social-networks.aspx.
19. Hampton, K. and Wellman, B. Neighboring in Netville:
How the internet supports community and social capital in a
wired suburb. City and Community, 2003, 2 (4). 277-311.
20. Hancock, J., Gee, K., Ciaccio, K. and Lin, J.M.-H. I'm
sad you're sad: emotinoal contagion in CMC. Proceedings
of Computer Supported Cooperative Work, 2008, 295-298.
21. Hektner, J., Schmidt, J. and Csikszentmihalyi, M.
Experience Sampling Method: Measuring the quality of
everyday life. Sage, Thousand Oaks, CA, 2007.
22. Herzberg, F. The Motivation to Work. Transaction
Publishers, 1993.
23. Hodges, S., Williams, L., Berry, E., Izadi, S.,
Srinivasan, J., Butler, A., Smyth, G., Kapur, N. and Wood,
K. SenseCam: A Retrospective Memory Aid Ubicomp,
2006, 177-193.
24. Iqbal, S.T. and Bailey, B.P., Investigating the
Effectiveness of Mental Workload as a Predictor of
Opportune Moments for Interruption. in Proceedings of the
ACM Conference on Human Factors in Computing
Systems, (Portland, Oregon, USA, 2005), 1489-1492.
25. Iqbal, S.T. and Horvitz, E., Conversation Amidst
Computing: A Study of Interruptions and Recovery of Task
Activity. in The 11th International Conference on User
Modeling, (Corfu, Greece, 2007), Berlin: Springer.
26. Iqbal, S.T. and Horvitz, E., Disruption and Recovery of
Computing Tasks: Field Study, Analysis and Directions. in
Proceedings of the ACM Conference on Human Factors in
Computing Systems, (San Jose, California, USA, 2007),
677-686.
27. Jin, J. and Dabbish, L.A. Self-interruption on the
computer: a typology of discretionary task interleaving
Proceedings of the SIGCHI Conference on Human Factors
in Computing Systems, ACM, Boston, MA, USA, 2009,
1799-1808.
28. Kendon, A., Harris, R.M. and Key, M.R. Organization
of Behavior in Face-To-Face Interaction. Walter de
Gruyter, 1975.
29. Kivran-Swaine, F., Brody, S., Kiakopoulos, N. and
Naaman, M. Of joy and gender: emotional expression in
online social networks ACM 2012 Conference on Computer
Supported Cooperative Work, 2012, 139-142.
30. Kramer, A.D. The spread of emotion via facebook
Conference on Human Factors in Computing Systems
(CHI), 2012, 767-770.
31. Larsen, R. and Ketelaar, T. Personality and
susceptibility to positive and negative emotional states.
Journal of Personality and Social Psychology, 1991, 61 (1).
132-140.
32. Latorella, K.A. Investigating Interruptions: Implications
for Flightdeck Performance. 1999,
http://www.interruptions.net/literature/Latorella-NASA-99-
tm209707.pdf.
33. Maes, P. Agents that Reduce Work and Information
Overload. Communications of the ACM, 1994, 37 (7). 30-
40.
34. Maglio, P. and Campbell, C.S., Tradeoffs in Displaying
Peripheral Information. in Proceedings of the ACM
Conference on Human Factors in Computing Systems,
(2000), 241-248.
35. Marco, C.A. and Suls, J. Daily Stress and the Trajectory
of Mood: Spillover, Response Assimilation,Contrast, and
Chronic Negative Affectivity. Journal of Personality and
Social Psychology, 1993, 64 (6). 1053-1063.
36. Marder, B., Joinson, A. and Shankar, A., Every Post
You Make, Every Pic You Take, I'll Be Watching You:
Behind Social SPheres on Facebook. in 45th Hawaii
International Conference on System Sciences (HICSS),
(2012), 859-868.
37. Mark, G., Gudith, D. and Klocke, U. The cost of
interrupted work: more speed and stress Proceedings of
CHI'08, ACM Press, Florence, Italy, 2008, 107-110.
38. Mark, G., Voida, S. and Cardello, A. "A Pace Not
Dictated by Electrons": An Empirical Study of Work
Without Email. Proceedings of CHI'12. ACM Press, 555-
564.
39. McCrae, R. and Costa, P. The five factor theory of
personality. in Handbook of Personality: Theory and
Research, L.A. Pervin, O.P. Johns, NY: Guilford, 1999,
139-153.
40. McDuff, D., Karlson, A.K., Kapoor, A., Roseway, A.
and Czerwinski, M. AffectAura: an intelligent system for
emotional memory Proceedings of CHI'12, 849-858.
41. McFarlane, D.C. and Latorella, K.A. The Scope and
Importance of Human Interruption in HCI Design. Human-
Computer Interaction, 2002, 17 (1). 1-61.
42. Paul, C. and Komlodi, A. Emotion as an indicator for
future interruptive notification experiences Proceedings of
CHI '12 Extended Abstracts, ACM Press, 2012, 2003-2008.
43. Pea, R., Nass, C., Meheulla, L., Rance, M., Kumar, A.,
Bamford, H., Nass, M., Simha, A., Stillerman, B., Yang, S.
and Zhou, M. Media use, face-to-face communication,
media multitasking, and social well-being among 8- to 12-
year-old girls. Developmental Psych, 2012, 48 (2). 327-336.
44. Rhoades, J. and O'Connor, K. Affect in computer-
mediated and face-to-face work groups: The construction
and testing of a general model. Computer Supported
Cooperative Work, 1996, 4. 203-228.
45. Rich, C. and Sidner, C.L. COLLAGEN: A
Collaboration Manager for Software Interface Agents. User
Modeling and User-Adapted Interaction, 1998, 8 (3/4).
315-350.
46. Russell, J. Core affect and the psychological
construction of emotion. Psychological Review, 2003, 110
(1). 145-172.
47. Ryan, T. and Xenos, S. Who uses Facebook? An
investigation into the relationship between the Big Five,
shyness, narcissism, loneliness, and Facebook usage.
Computers in Human Behavior, 2011, 27. 1658-1664.
48. Schaufeli, W., Salanova, M., Gonzalez-Rom, V. and
Bakker, A. The measurement of engagement and burnout:
A two sample confirmatory factor analytic approach.
Journal of Happiness studies, 2002, 3. 71-92.
49. Schwartz, G. Estimating the dimension of a model.
Annals of Statistics, 1978, 6 (2). 461-464.
50. Skeels, M. and Grudin, J. When social networks cross
boundaries: A case study of workplace use of facebook and
linkedin ACM Group '09, 2009. 95-104.
51. Steptoe, A., Wardle, J. and Marmot, M. Positive affect
and health-related neuroendocrine cardiovascular and
inflammatory processes. Proceedings of the National
Academy of Sciences, 2008, 102 (18). 6508-6512.
52. Teuchmann, K., Totterdell, P. and Parker, S.K. Rushed,
unhappy, and drained: an experience sampling study of
relations between time pressure, perceived control, mood,
and emotional exhaustion in a group of accountants.
Journal of Occup Health Psychology, 1999, 4 (1). 37-54.
53. Tosun, L.P. and Lajunen, T. Does Internet use reflect
your personality? Relationship between Eysenck's
personality dimensions and Internet use. Computers in
Human Behavior, 2010, 26 (2). 162-167.
54. Watson, D. and Clark, A.L. On traits and temperament:
General and specific factors of emotional experience and
their relation to the five-factor model. Journal of
Personality, 1992, 60 (2). 441-476.
55. Watson, D., Clark, A.L. and Tellegen, A. Development
and validation of brief measures of positive and negative
affect: The PANAS scales. Journal of Personality and
Social Psychology, 1988, 56 (6). 1063-1070.
56. Williams, K. and Alliger, G. Role stressors, mood
spillovers and perceptions of work-family conflict in
employed parents. . The Acad of Mgt J, 1994, 37 (4).
57. Zijlstra, F.R.H., Roe, R.A., Leonora, A.B. and Krediet,
I. Temporal Factors in Mental Work: Effects of Interrupted
Activities. Journal of Occupational and Organizational
Psychology, 1999, 72. 163-185.
