Focused, Aroused, but so Distractible: A Temporal
Perspective on Multitasking and Communications
Gloria Mark
1
, Shamsi Iqbal
2
, Mary Czerwinski
2
, Paul Johns
2
1
Department of Informatics
University of California, Irvine
Irvine, CA 92697 USA
2
Microsoft Research
One Microsoft Way
Redmond, WA 98052 USA
{shamsi,marycz,Paul.Johns}@microsoft.com
ABSTRACT
A common assumption in studies of interruptions is that
one is focused in an activity and then distracted by other
stimuli. We take the reverse perspective and examine
whether one might first be in an attentional state that
makes one susceptible to communications typically
associated with distraction. We explore the confluence of
multitasking and workplace communications from three
temporal perspectives – prior to an interaction, when tasks
and communications are interleaved, and at the end of the
day. Using logging techniques and experience sampling,
we observed 32 employees in situ for five days. We found
that certain attentional states lead people to be more
susceptible to particular types of interaction. Rote work is
followed by more Facebook or face-to-face interaction.
Focused and aroused states are followed by more email.
The more time in email and face-fo-face interaction, and
the more total screen switches, the less productive people
feel at the day's end. We present the notion of emotional
homeostasis along with new directions for multitasking
research.
Author Keywords
Facebook; Email; Face-to-face interaction; multitasking;
productivity; interruptions
ACM Classification Keywords
H.5.3 [Information Interfaces and Presentation (e.g.,
HCI)]: Group and Organization Interfaces; K.4.m
[Computers and Society]: Miscellaneous.
General Terms
Human Factors
INTRODUCTION
While studies of multitasking and disruption have long
been a focus of the CSCW and CHI communities, the
emphasis has mostly been on understanding how
disruption occurs from an engaged state, either due to
external stimuli such as notifications or visits from
colleagues, or self-interruptions. However, there has been
little research investigating whether a person’s particular
mental state at the time could make one more susceptible
to being distracted.
Prior work has shown how online interactions, in
particular the use of social media and email, can be used
to infer what type of attentional state a person is in, such
as being focused or bored [22]. However, while such an
association was established, the direction of causality was
not clear: Does being in a particular attentional state make
one more susceptible of switching from their current task
to pursue certain online activities? Or rather does
switching attention from an ongoing task to certain
activities lead one to be in a particular attentional state?
In this paper, we explore the relationship of multitasking
and communications in the workplace. As this
relationship is a complex phenomenon, we choose to
examine this relationship through three temporal
perspectives, as multitasking occurs throughout the day:
what happens prior to switching activities that may lead to
workplace communications, how are communications
manifest during task switching, and how do the
cumulative effects of multitasking affect people's
assessment of work productivity at the end of the day?
Understanding the relationship of multitasking and
communications is important, as workplace
communications comprise a significant portion of the
workday [10]. Further, communications such as informal
face-to-face interactions or email are noted as a major
source of workplace distraction [10, 20, 26]. This study
builds on prior work [22], which examined how
attentional states in the workplace vary over the day, with
digital activities.
We conducted an in situ study in a large, U.S. global
organization where we tracked online activities of users
throughout the workday, and collected self-reports of their
engagement and feeling of being challenged using the
experience sampling method (ESM) [13]. Leveraging the
framework derived in [22], we associated different
attentional states with online activities. The results
suggest that particular attentional frames of mind lead
Permission to make digital or hard copies of all or part of this work fo
r
personal or classroom use is granted without fee provided that copies are no
t
made or distributed for profit or commercial advantage and that copies bea
r
this notice and the full citation on the first page. Copyrights for components o
f
this work owned by others than ACM must be honored. Abstracting wit
h
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. Reques
t
permissions from Permissions@acm.org.
CSCW '15, March 14–18, 2015, Vancouver, BC, Canada
Copyright is held by the owner/author(s). Publication rights licensed to ACM.
ACM 978-1-4503-2922-4/15/03…$15.00
http://dx.doi.org/10.1145/2675133.2675221
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
903
people to be susceptible to different types of task switches
to varying degrees. We also found a relationship between
the amount of task switching and workplace
communications. We discuss the impacts of our findings
on the current understanding of multitasking and
attentional states, and introduce the notion of emotional
homeostasis.
RELATED WORK
Multitasking and disruption
Existing work in the domain of multitasking and
disruption has primarily focused on how multitasking
impacts attention in the workplace, in particular, the
effects of an ongoing task being interrupted by another
activity. The underlying assumption is that external or
internal stimuli cause people to change their focus from
their current activity to another task. This may often be
characterized as distraction or disruption to the
interrupted task, even though the interruption may be
beneficial [6, 16, 20]. For example, Czerwinski et al. [4]
conducted a diary study to understand how interruptions
cause information workers to switch activities in the
workplace. Distractions as such can affect focus on
ongoing tasks. Mark et al. [20] found that it takes people
on average around 23 minutes to resume an interrupted
task. Iqbal and Horvitz studied how external interruptions
cause information workers to attempt to leave their
ongoing task at a stable state and then enter into a ‘chain
of distraction’, comprised of a series of activities
including checking email, instant messaging and
browsing [10]. O’Conaill and Frohlich [26] showed that
64% of workplace interruptions are beneficial, but the
recipient also does not resume work after 40% of the
interruptions. More recently, Mark et al. showed in a
study of multitasking among Millennials that the amount
of multitasking is positively associated with stress, but
social media usage coincides with lower stress [24].
Other work has focused on internal interruptions, where a
person may interrupt themselves during ongoing work
without external stimuli [1, 7, 17]. Jin and Dabbish
derived seven categories of self–interruption in the
workplace – adjustment, break, routines, waits, inquiry,
triggers and collection. However, research has suggested
that the organizational environment and individual
differences may determine how susceptible one may be to
different types of self interruption [7].
Workplace communications such as informal face-to-face
(F2F) encounters, as well as online interactions with
email, comprise a significant chunk of the workday, and
have been found to take up over a quarter of people's time
in the workplace [10]. Correspondingly, these
communications also comprise a significant amount of
interruptions. Email and informal F2F interactions alone
make up over 40% of interruptions [10]. Facebook (FB)
users in the workplace on average spend almost nine
minutes a day on FB. Typically this involves multiple
visits, which also interrupts workflow [23].
While most studies have looked at distraction due to
multitasking, no study to our knowledge has attempted to
study the phenomenon from the reverse perspective – that
people can be in particular frames of mind, or attentional
states, that make them more susceptible to being
distracted. Linking the attentional state to types of
activities can provide a better understanding of the nature
of distraction and the subsequent effects on productivity.
Attentional states in the workplace
Attentional states in the workplace are important to study
as they appear to be related to mood and possibly job
performance. Grandey et al. [11] found that positive affect
was related to job satisfaction, whereas negative affect
was related to negative emotional reaction, e.g.,
disappointment, depression and unhappiness. More
recently, Mark et al. [23] found that face to face
interactions at work were associated with positive mood
at the time they occur, whereas more Facebook use was
positively correlated with positive mood at the end of the
workday. More relevant to our study is that mood has
been shown to have an effect on distraction; e.g., Alder
and Benbunan-Fich [1] showed that negative feelings
trigger more self-interruptions than positive feelings.
We are interested in the relationship of attentional states
and people’s digital activity. To inform our study we draw
from a long legacy of related concepts in psychology,
which explain that people’s actions are motivated by a
desire to achieve a balanced emotional and psychological
state. As far back as 460 BC, Hipprocrates proposed that
health was related to a harmonious balance of elements in
the body. In the early twentieth century, Gestalt theorists
were interested in the idea of balance in terms of the
perceptual field using the concept of Pragnanz: in other
words, people try to reduce stress from the stimulus field
so as to achieve an internal equilibrium [18]. About the
same time, the physiologist Walter Cannon adopted the
term homeostasis and expanded its reach to include
emotional parameters as well as physical [5]. Kurt
Lewin’s field theory [19] built on the idea of balance and
introduced the basic principle of tension reduction as it
applies to internal states. According to Lewin, people
experience tension when needs are not satisfied, and
therefore they strive to attain these needs to reduce
tension in order to experience a state of equilibrium.
Lewin focuses on a person’s momentary needs; any
change of state is dependent on the situation and one’s
particular psychological state at the time. Other related
theories that grew out of Lewin’s field theory and that
similarly discuss notions of achieving a balanced mental
state include Heider’s balance theory [12], which
describes that people are motivated to maintain attitudes
that are consistent towards other people and objects over
time, so as to achieve a psychological balance. Cognitive
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
904
dissonance theory is also related, where discrepancies
between attitudes and behaviors introduce dissonance [9].
If a discrepancy exists, a person will aim to reduce the
dissonance.
More recently, internal balance, or homeostasis, has been
approached in terms of physiological responses. As we
are concerned with the reduction of tension, or stress, we
are interested in how people maintain emotional
homeostasis while multitasking. Emotion is typically
defined as a mental state that arises spontaneously rather
than through conscious effort and is often accompanied
by physiological changes [25]. Homeostasis is defined as
the ability or tendency of an organism or cell to maintain
internal equilibrium by adjusting its physiological
processes [25]. Responses to threats of homeostasis from
stressors occur by facilitating neural pathways which
mediate psychological functions such as arousal,
cognition and attention [5]. Stress can thus create an
imbalance and influence arousal and attentional state.
People are constantly challenged by stressors in the
environment and have developed adaptive responses in
order to preserve homeostasis. Therefore, emotional
homeostasis is a neural and physiological process that
maintains the equilibrium of mental states that would
enable a human to live and perform at normal levels [4].
It may be that, as workers get more stressed, bored or
frustrated in the workplace, they seek out homeostasis by
moving to another activity that brings them back to a
balanced state.
Thus, these various theories share the basic commonality
that people act in a way to seek out and experience a
balanced internal state. In the broader context of
multitasking in a digital environment, people may switch
to activities that will lead them to experience more of a
state of equilibrium if the current activity disrupts that
balance. Following Lewin [19], Heider [12], and
Festinger [9], people may choose actions that will lead
them towards reducing inner tension [21]. This is a novel
notion in the multitasking literature that we will explore
in this work.
INTERACTIONS AND MULTITASKING
As a starting point to examine more broadly how social
interactions influence multitasking behavior in the
workplace, we investigated the confluence of workplace
communications and multitasking in terms of three
temporal perspectives: prior to the interaction, throughout
the day (how communication acts interleave with
multitasking), and the cumulative effects at the end of the
day (the effect on assessing productivity). As a first step
we chose to focus on three prevalent types of
communications that would cover online and offline
interactions and work and social purposes. We therefore
selected three types of communications common in the
workplace: F2F, email, and FB. Our reasoning for
selecting these three types of communications is as
follows:
F2F: informal F2F interaction is common in the
workplace and is a significant source of distractions [8,
10]. F2F encounters are offline and are either work-
related or social.
Email: online email communication is also a significant
source of distractions, both due to self-interruption as well
as external interruption, e.g., due to notifications [10, 16].
Email in the workplace is likely to be mostly work
related, although obviously some personal email is also
carried out.
FB: Facebook is also online communication and is likely
a distraction mostly due to self-interruption. As [23]
found, it can function as a quick break when people are
engaged in work. FB in the workplace is generally social,
though in rare cases it may be work-related.
Thus, we focus on contrasting workplace
communications--differing in their online or offline nature
and in being social or work-related, to understand the role
they play in multitasking. While there are a number of
other workplace interactions that also occur with other
media (e.g., IM, LinkedIn, phone), as a starting point, we
examine these three types (FB, F2F and email).
We emphasize that these types of communications could
be work-related or social in nature, but our data does not
allow us to make this differentiation. Our approach is to
examine how engaging in these types of communications
interrupts the flow of activity, regardless of whether they
are social or work related. Similarly, when we consider a
task that involves digital activity, such as Internet
switching, we acknowledge that this may also be work-
related or not. Our focus is on interruptions of the flow of
activity irrespective of whether the communication type
or target activity is work-related or not. We next describe
Engagement‐High
Engagement‐Low
Challenge‐Low Challenge‐Hi
gh
Q1: Hi
ghly Engaged and
Chall
enged: “F
ocus”
Q2: Hig
hly Engaged, not
Challenged: “Ro
te”
Q3: Low
Engagement, not
Challenged: “Bored
”
Q4: Lo
w Engagement, High
Challenge: “F
rustrated”
Figure 1. A theoretical framework of different attentional
states [22]
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
905
the three temporal perspectives that we focus on in our
analysis.
Prior to the interaction: attentional states
Studies of multitasking and interruptions assume that
people are highly engaged in their work and then the
distraction pulls them away from their engagement. In
this paper we take the opposite perspective: we examine
whether people may first experience a particular
attentional state that makes them susceptible to
distractions such as checking FB, email, or chatting with a
colleague. For example, a person may become bored in
their current task, which makes them turn to FB as a
break. Recent work suggests that FB can be beneficial as
a quick break, leading to "grazing behavior", a low cost
interaction from a cognitive resources perspective,
potentially fun, and which enables the user to maintain
some control over the duration of the interaction [23].
To examine this notion of attentional states, we draw on
the theoretical framework used by [22], shown in Figure
1. This theoretical framework is based on considering
attentional states in terms of two dimensions: engagement
and challenge. Engagement is important to consider in the
workplace as it is a mental state of absorption in an
activity [28]. The second dimension that we consider is
challenge in work. Challenge refers to the amount of
mental effort that one exerts to perform a task and has
been associated with motivation in work [14].
Figure 1 describes four attentional states, shown in each
of the four quadrants. To be focused in work, engagement
in an activity is not sufficient. A person can be engaged in
an activity but it may not be challenging; a person can be
involved in "mindless" activities such as filling out forms.
Therefore, it is important to consider how challenging an
activity is along with how engaged one is in the activity.
Quadrant 1 indicates that when one is highly challenged
and highly engaged then the attentional state can be
characterized as 'Focus'. Quadrant 2 indicates that an
attentional state where one is highly engaged but not
challenged describes mechanical type of thinking. This is
referred to as 'Rote' to indicate attentional states
associated with rote type of work. Quadrant 3 shows that
when one is not at all engaged and not challenged, then it
indicates a state of being 'Bored'. Quadrant 4 indicates
that when one is challenged but not engaged, then one is
frustrated. An example of a frustrated attentional state
could be when a software developer is working on solving
a difficult bug in a program. We stress that these labels
are merely referents; what is more important is to
consider the underlying dimensions of being challenged
and engaged.
Therefore, we apply the framework in Figure 1 to
examine a person's attentional state prior to an interaction.
Our goal is to examine whether we can find relationships
of particular prior attentional states with subsequent
initiations of workplace communications. If so, this may
suggest that a person may already be in a particular
attentional state that makes them susceptible to certain
types of interactions (or distractions).
During multitasking: how communications interleave
with other activities
A second temporal perspective that we examine concerns
how workplace communications interleave with other
multitasking behavior. This analysis could shed light on
the role workplace communications play in switching
between different tasks. Certain communications could
serve different functions in multitasking. They can be a
break when a person is highly engaged in work, as [23]
found with FB. They can also serve to provide
information needed to perform tasks, such as when using
email or F2F for task completion. They can also be
habitual, as when one surfs the Internet, and then one
becomes accustomed to also checking FB or email,
starting off a chain of distraction [16].
The number of projects a person has can also affect how
communications interleave with multitasking. We expect
that the more projects one has, the more task switching
occurs, which in turn could present more opportunities for
initiating further communications
. For example, if one
switches projects, they may turn to email to retrieve
newly pertinent information or they may seek a colleague
to consult with for updates. We therefore look at task
switches and project count in conjunction with our three
main interaction types (FB, F2F and email).
Another common practice of activity switching could be
due to switching websites on the Internet. When one
switches between projects, email may play a more
important role since it is generally work-related; one may
need to use email to seek information for projects. On the
other hand, when one uses the Internet – even if it is still
related to the ongoing task- one may check FB, since it is
a relatively low cost switch within the same application.
However, if FB is a break from work as [23] suggests,
then FB use might be prevalent while one is working and
switching between projects. This relates to Lewin’s idea
of tension reduction. On the other hand, one might also
choose to have F2F interaction while switching Internet
sites if one is bored.
End of day: cumulative assessment of productivity
Our third temporal perspective focuses on how people
feel at the end of the day. A longstanding question about
multitasking is the effect that switching tasks has on
productivity (cf [2]). It is unclear what effect the
relationship of workplace communications and
multitasking might have on productivity. We are
particularly interested in the effects of FB, F2F and email
interactions throughout the day on the cumulative
assessment of productivity at the end of the day. The
underlying reasoning behind this exploration is that these
interactions could have both positive and negative effects
on productivity. For example, while FB is generally
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
906
considered a distraction from ongoing activities, it could
also serve as a break during a productive session, and
therefore be considered with feeling engaged [23].
Likewise, F2F and email could be both work related and
social.
Research foci
In sum, we study three types of workplace
communications: FB, email, and F2F, from three different
temporal perspectives:
Attentional states prior to the communication
Throughout the day: how communications interleave
with multitasking
End of the day: cumulative assessment on feelings of
productivity
METHODOLOGY
We conducted an in situ study in the fall of 2012 at a
large U.S. corporation. We used a mixed-methods
approach where we logged people's digital activity along
with using experience sampling (ESM). The logging
allowed us to track a wide range of digital activities with
detailed precision. As stated earlier, ESM was used to
collect user perceptions of engagement and challenge, as
well as other self-report measures at frequent intervals
throughout the day. We also deployed surveys for other
subjective and demographic measures. Further details of
these, and other, measures not reported in this paper can
be found in [23].
Participants were recruited through advertising,
convenience sampling and other participants’
recommendations. Thirty-two people (17 females, 15
males) participated. Participants included researchers,
managers, administrators, an engineer, a department
director, a designer, and a consultant.
Methodology. Each participant's digital activity was
logged for a period of five work days, typically Monday
through Friday. When participants traveled or missed a
day, they made up the missed day the following week (in
most cases). The computer logging software and ESM
software were installed on participants' computers the
Friday before the study began. Participants were assured
of anonymity in their data and it was protected via
encryption.
We logged online interactions with custom-built software
that captured all activity in the Windows 7.0 Operating
System. This included beginning and end times for the
lifespan of every window, and the beginning and end
times for each instance of every foreground window.
Logging is done only when a window is moved to the
foreground, i.e., if an email client is open, its use will not
be logged unless it becomes active. Changing tabs in a
browser was counted as separate switches. Mouse and
keyboard activity were captured, as was computer sleep
mode, so that we could ignore periods of time when a
window was open but was not being used in the
foreground. Capturing what email was being read or any
other application interaction was not collected due to
privacy and technical limitations. All participants used
Outlook for email.
Email and FB interaction were measured through the
logging program. F2F interaction was measured through
the use of SenseCams [15], a lightweight wearable
camera worn around the neck. The camera takes pictures
approximately every 15 seconds. The images were then
processed with face detection software, a publicly
available application (http://research.microsoft.com/en-
us/projects/facesdk/). The software does not recognize
faces; it only provides information about whether a person
was present or not in the photo. The counts in our F2F
variable therefore do not measure distinct interactions,
just the amount of interaction.
We used ESM, in the form of a small pop-up window that
appeared on the computer screen to capture the
participants' perspective in situ. Experience sampling has
been shown to have both internal and external validity
[13]. Experience sampling has been used extensively in
studies to capture the experience of flow, an immersive
state in an activity [13]. We used a hybrid interval-
contingent and event-contingent sampling approach [13].
The sampling was done: 1) whenever a user left email
after uninterrupted active use in that application for at
least three consecutive minutes or when in Facebook after
a full minute, and 2) whenever a user logged into
Windows or unlocked the screen saver (event-contingent).
If 15 minutes passed without a sampling, then a probe
was triggered (interval-contingent).
Participants were instructed to go about their usual
workday activities and were told to answer the ESM
probes when the probe windows popped up on their
computer screens. We emphasized that they should
answer the probe questions as accurately as possible but
they could cancel the probe window at any time. Subjects
were given the following verbal and written instructions:
"
Sometimestheratingscalewillpopupandmayannoyyou,
especially if you were in the middle of doing something. If
youfeelannoyed,donotrateyourmoodbasedonthe
annoyance of the pop‐up window. Instead, rate your
experiencebasedonthetaskorinteractionyouweredoing
atthetimeofthepop‐upwindow.Ifyoufeelthatyoucannot
rate your mood fairly due to the annoyance of the pop‐up
window,thenhit‘cancel’andthewindowwilldisappear.
"
We used rating scales used in other ESM approaches [21]
to measure the following: for Engagement, participants
were asked 'In the task/interaction you were just doing:
How Engaged Were You?' using a 6-point Likert scale
(0=Not at All; 5=Extremely). To measure Challenge,
participants were asked the same question as above, but
instead: "How Challenged Were You?' using the same
Likert scale: (0=Not at All; 5=Extremely). We also
measured Valence (positive and negative affect, not
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
907
reported here) and Arousal, (on a vertical axis that crossed
a horizontal Valence axis) using a range of -200 (low
arousal) to +200 (high arousal), based on the Circumplex
model for Valence/Arousal (for a review, including its
validity, see [27]). Subjects were asked to click with their
cursor on that point on the scale that best expressed their
feeling "right now." The timestamp when participants
submitted the probe was recorded. All data from ESM
was normalized within participants.
Measures
Table 1 summarizes the measures from our collected data
reported in this paper.
RESULTS
Our dataset consisted of 1,509 hours of participant
observation (32 participants, five days each). Our
experience sampling yielded 2,809 probes, averaging 87.8
responses per participant, and averaging 17.56 probe
responses per person per day. Our SenseCam photo
capture yielded 204,922 photos. The ESM results were
normalized and we used the top and bottom third of
responses for the engagement and challenge dimensions
(i.e. eliminating the middle responses). Since only seven
responses occurred in the 'Frustrated' category, we do not
consider this category for the rest of the analysis. The
responses fell in the quadrants shown in Fig. 1 as follows:
Focused (45.4%), Rote (18.9%), Bored (35.6%).
We begin with an overview of our results. Table 2 shows
the average daily duration, frequency, and duration per
visit for Facebook, email, and F2F interactions. Because
the SenseCam does not compute duration, we can only
provide counts of faces detected in photos as a proxy for
amount of F2F interaction. The average duration per visit
for FB is about 18 seconds and for email is about 32
seconds. Participants averaged about 87 SenseCam counts
per day (note that SenseCam counts are a proxy for
amount of F2F interaction).
We discussed earlier that checking FB and Email might
be a frequent and habitual behavior. We found that
participants visited FB on average 21 times per day, with
a maximum of 264 visits per day. If we examine only
those days in our data where people visited FB at least
once a day (i.e., of those who used FB on any day), the
average visits per day climbs to 38 unique visits per day
(SE=4.92). Participants visited email much more daily:
averaging 74 times per day, with a maximum of checking
373 times per day.
Based on the work roles participants reported in a survey,
work roles were coded into three categories: concerning
Administration and technical support (5 people), Research
(19 people), and Management (8 people). For Work Role,
we found that for average daily duration of FB use,
Researchers spend significantly less time on FB
(M=309.59, SE=76.06) than Managers (M=952.15,
SE=359.62) or Admins (M=669.15, SE=159.14): F(2,
161)=3.53, p<.03. There is no significant effect of Work
Role on Email duration. With F2F, Admins average
significantly more F2F time per day (M=101.58,
SE=16.37), followed by Managers (M=98.46, SE=13.04),
and then Researchers (M=60.47, SE=8.57), F(2,
161)=4.34, p<.02.
The average Project count per person is 6.7 (sd=3.7), with
a range of 1-18. There is a significant difference of
Project Count according to Work Role: F(2,31)=3.68,
p<.04. Researchers (M=5.3, SE=.22) have fewer projects
on average than Admins (M=8.7, SE=.93) or Managers
(M=8.6, SE=.67). The average number of times that a
Measure Description
FB interaction FB seconds 1, 5, 10 min. prior and
after the probe
FB counts of unique visits
Email
interaction:
Email seconds 1, 5, 10 min. prior
and after the probe
Email counts of unique visits
F2F interaction: SenseCam counts 1, 5, 10 min.
prior and after the probe
Application
Switching
Number of switches between
applications (e.g. Word, Excel)
Internet
switching
Number of switches on the Internet
Attentional
states
Counts of ESM responses that fell
into each of the quadrants: Focus,
Rote, Bored, Frustrated states.
Arousal low (-200) to high (+200)
Productivity self-
report
daily end of day survey: 7-point
Likert scale
Project count general survey
Positive/negative
mood
PANAS mood scale, from daily
beginning and end of day survey)
[29]
Table 1. Summar
y
of measures used.
FB Email F2F
Number visits/day
(counts)
20.98 (5.04) 74.05 (10.54) --
Duration per day
(sec.)
583.48
(206.75)
(9 min. 43
sec.)
2071.16
(294.46)
(34 min. 31
sec.)
86.88
(10.65)
1
Duration per visit
(sec)
17.75 (2.81) 32.06 (2.80) N/A
Table 2. Average daily values of number of visits, duration,
duration per visit, for FB, Email use.
For F2F, only counts
are available. Mean (SE).
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
908
person switches applications daily is 566.0 (SE=41.89),
ranging from 183 to 1035 times per day.
Overall, the usage data suggests that both FB and email
usage are characterized by short bursts: checking many
times per day with a short duration for each visit. Amount
of F2F and FB interaction, as well as number of projects,
differ by work role.
Attentional state: susceptibility to workplace
communications
In our first research question, we are interested to see
whether a person's current attentional state is associated
with their subsequent workplace communications. Is it
possible that a particular attentional state, such as feeling
bored or feeling that work is rote, is a precursor to using
FB, checking email or having a F2F interaction? Based on
the ESM probe data, we can determine what type of
attentional state the participant reported experiencing
(using the framework in Fig. 1). Our probes asked people
to rate how they felt "right now". Since we have
timestamps of the logged computer data, we can therefore
examine what activity people did after reporting their
attentional state.
However, it is possible that if the person is already
performing a certain workplace communication before the
ESM probe, they could continue that communication after
the probe. In that case it would be difficult to determine
whether it was the attentional state that led to the
communication or if it was the prior communication that
was simply being continued. For example, if FB causes
people to be bored, then a probe after FB use would show
a bored state. If a person then subsequently continued FB
use, we would not be able to determine if it was the bored
state that led to FB use or if rather prior FB use continued
after the probe. Therefore, we controlled for the same
type of communication before the probe as the target
communication we are investigating after the probe. By
controlling for the same type of communication prior to
the attentional state, we can control that this prior
behavior is not influencing the analysis of the association
of attentional state and behavior after the probe. It is
important to keep in mind that there could be any number
of activities that could influence a person's attentional
state.
We used a GLM to examine attentional states prior to the
three workplace communication types that we are
investigating: FB use, email use, and F2F interaction, to
see if there are significant patterns of prior attentional
states. As we do not have any a priori knowledge of what
window of time to test, we selected the timeframe of a 10-
minute window of time prior to, and after, the probe. We
also compared 5 and 1-minute windows, with similar
results, but as the 10-minute window showed the
strongest effects, we will report this analysis throughout
the rest of the paper.
Facebook
We found significant differences in FB use duration that
followed the different attentional states (see Table 3). FB
duration following a Rote state was longer than FB
duration following a Bored state, followed by a Focused
state; a Bonferroni post-hoc test showed a significant
difference between Rote and Focused.
We next checked whether the same results also apply for
the number of unique FB visits. In other words, is there a
particular attentional state associated with FB checking
behavior, as well as FB duration of use? A GLM using the
number of FB visits in the next 10 minutes, controlling
for FB visits in the prior 10 minutes, shows significant
results consistent with the FB duration results in Table 3.
Email
We also found significant differences in email duration
that followed the different attentional states (Table 3).
Participants were more likely to use email longer after
being in a Focused state, followed by a Rote state,
followed by a Bored state; a Bonferroni post-hoc test
showed a significant difference between Focused and
Bored states. Participants are therefore more likely to
spend more time in email when they are feeling focused,
compared to when they are feeling bored, or doing rote
work.
Checking the number of unique Email visits in the next 10
min. controlling for number of Email visits in the prior 10
min. shows no significant effect of Attentional state.
Face-to-face
For F2F, we used the SenseCam photo counts as a proxy
for amount of F2F interaction. We used a log transform
for F2F to improve normality. We found significant
differences in F2F interaction counts following different
attentional states (Table 3). Participants were more likely
to have more F2F interactions after being in a Rote state,
followed by a Bored state, followed by a Focused state. A
Bonferroni post-hoc test showed a significant difference
between Rote and Focused states. Thus, people are most
likely to engage in more F2F interactions when feeling
Rote (engaged but not challenged) in their current activity
A
ttn'l
State
FB (sec.) Email (sec.) F2F
(SenseCam
counts)
Focus 13.26 (3.0)* 86.56 (5.6)* 1.32 (.18)*
Rote 28.62(4.6)* 68.03 (8.7) 1.84 (.28)*
Bored 18.95 (3.5) 64.77 (6.6)* 1.73 (.21)
F(2,1070)=3.92,
p<.02, R
2
=.13
F(2,1070)=3.64,
p<.03, R
2
=.20.
F(2,1070)=4.30,
p<.02, R
2
=.27
Table 3. Mean (SE) of FB, Email, and F2F durations given
each prior attentional state of Focus, Rote, Bored. *Indicates
differences based on Bonferroni post hoc test, p<.05. Adjusted
R
2
is reported.
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
909
and least when they feel focused (engaged and
challenged).
Arousal and workplace communications
The result with email duration was contrary to that of FB
and F2F interactions, with the pattern showing that one
was first focused before doing email. We decided to
explore this more closely. A focused state of attention
involves the dimensions of high engagement and high
challenge. An underlying mechanism that we expect
would correlate with engagement and challenge is
arousal. In the ESM probes, we had also asked people to
rate their level of arousal. Indeed, we found that Arousal
is highly correlated with the dimensions of feeling
challenged (r=.42, p<.0001) and engaged (r=.57,
p<.0001).
We next looked at whether one's level of arousal is
associated with email use. A GLM analysis predicting
Email duration in the next 10 minutes, with Arousal and
Attentional state as independent variables, controlling for
Email duration in the prior 10 minutes, showed a
significant effect of Arousal: F(1,1070)=5.30, p<.02,
R
2
=.20. If Arousal is included in the model then there is a
significant Arousal x Attentional State interaction:
F(1,1070)=2.92, p<.05, and no main effect of Attentional
State. The variance inflation factors (measuring
multicollinearity) of Arousal and Attentional state were
each 1.4, which is an acceptable level as it is below 5.
Thus, Arousal is predictive of subsequent email use, and
Arousal and Attentional state (i.e. Focus) interact to
predict a higher level of email use.
Since FB and F2F were associated with following Rote
work, we did not expect Arousal to be a significant
predictor for these social communications (since Rote
work should not involve high effort). Indeed, when
Arousal was added to the GLM model previously
mentioned, and with Attentional state predicting FB use
in the next 10 minutes, controlling for FB use in the prior
10 minutes, it shows no significant effect of Arousal.
Examining F2F amount in the next 10 minutes, a GLM
also showed no significant effect of Arousal, controlling
for F2F amount in the prior 10 minutes.
Thus, we had hypothesized that particular attentional
states might be associated with particular communication
forms that follow. Our results show that when people are
in a rote state (feeling engaged but not challenged), they
subsequently spend more time on FB and check FB more,
as well as having more F2F interaction. However, when
people are in a focused state, they are more likely to then
spend a longer time on email. Email use is also associated
with arousal. Thus, when people are aroused and in a
focused state, they then spend a longer period of time
doing email.
Throughout the day: Multitasking and communication
We next looked at a second temporal perspective with
respect to multitasking: how workplace communications
(FB, F2F and email) interleave with multitasking
throughout the day. We used the variable of frequency of
switching applications and websites as a proxy for
multitasking. We also considered project count as we
expect that the more projects one is involved in, the more
opportunity one has to switch between projects. Activity
switches were in turn separated into two categories:
application switches and Internet switching. We reasoned
that focusing on application switches (App Switches)
might involve more different operations than Internet
switching, and thus would involve more cognitive shifts,
e.g., from a Word document (writing) to Excel
Application Switching
App PC WR
PC x
WR
A
dj.
R
2
FB
(sec.)
F 7.71 17.43 2.73 5.15
0.22
df 1,161 1,161 2,161 2,161
p 0.006 0.0001 0.07 0.007
Email
(sec.)
F 4.36 1.98 2.87 3.99
0.09 df 1,161 1,161 2,161 2,161
p 0.04 0.16 0.06 0.02
F2F
(counts)
F 2.51 1.97 1.9 1.95
N/A
df 1,161 1,161 2,161 2,161
p 0.12 0.16 0.15 0.15
Table 4. Model showing how average daily application
switches (App), project count (PC) and work role (WR)
affect avera
g
e dail
y
duration of FB
,
Email
,
and F2F counts.
Internet Switching
Int PC WR
PC x
WR
A
dj.
R
2
FB
(sec.)
F 12.85 19.06 2.51 4.77
0.24
df 1,161 1,161 2,161 2,161
p 0.0001 0.0001 0.08 0.01
Email
(sec)
F 5.71 2.25 3.72 5.99
0.10 df 1,161 1,161 2,161 2,161
p 0.02 0.14 0.03 0.003
F2F
(counts)
F 9.54 2.33 1.88 1.56
0.10 df 1,161 1,161 2,161 2,161
p 0.002 0.13 0.16 0.21
Table 5. Model showing how average internet switches
(Int), project count (PC) and work role (WR) affect average
daily duration of FB, Email, and F2F counts.
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
910
(calculations), to the Internet (search and reading) to
email (reading/writing). We also examined Internet
switching, as this could reveal insight into whether
checking email or FB might be frequent or habitual: when
one switches Internet sites one might also check email
and FB out of habit. We included Work Role in the model
since we found significant differences in FB and F2F
behavior with work role.
For all analyses we ran a separate GLM for FB, Email,
and F2F, with duration of each communication type as a
dependent variable, and with Project Count, App
Switches (or Internet switches), and Work Role as
independent variables. Since Project Count differs by
Work Role, we included an interaction term of Project
Count x Work Role.
Duration of FB, Email, F2F.
Table 4 shows a significant relationship of App Switches
to both FB and email duration. Thus, the more application
switches one does, the longer time one spends in FB and
email. There is also a significant Project Count by Work
Role interaction for FB and also for email duration. The
role of Researcher results in having the fewest projects
and spending the least amount of time in FB as well as
email, compared to other work roles. F2F showed no
significant effects, in contrast to email and FB which did
show significant effects.
Table 5 shows a significant relationship of frequency of
Internet switching to the duration of time spent in email,
FB, and F2F interaction. Thus, the more frequently one
switches Internet sites, the more time one spends in the
three communication types. A significant Work role by
Project count interaction for FB indicates that as Project
count increases, FB duration decreases more for
Researchers than for Admins and Managers. The
significant Work role by Project count interaction for
email means that as the number of projects increase, email
duration decreases for Admins, whereas for Researchers,
email duration increases.
Frequency of FB, Email
We next ran a separate GLM for FB and Email with
number of unique visits of each communication type as a
dependent variable, and with Project Count, App
Switches (or Internet Switches), and Work Role as
independent variables. Tables 6 and 7 show the
relationship of FB and email communications with
multitasking, from the perspective of number of visits
(note that F2F results, as measured by SenseCam counts,
are not included here, but rather are shown in Tables 4
and 5).
Table 6 shows that both average daily frequency of
visiting FB and email are strongly associated with
frequency of App Switches. Also, the more projects one
works on, the higher the frequency of checking FB and
email. A Project Count x Work role interaction exists only
for email: As the number of projects increase, Admins
check email more frequently than Researchers or
Managers.
Table 7 shows that Internet switching is highly associated
with frequency of visiting FB and email. We interpret the
significant project count effect for both FB and email as:
the more projects one has, the more switching between
projects one does, and the more opportunities there are for
checking FB and email while switching. Note that both
models for email have a very high value of R
2
(.49 and
.50) which indicates that project switching and Internet
site switching contribute quite a bit to explaining email
checking behavior.
In summary, both application switching and Internet site
switching are associated with higher FB and email use, in
terms of both duration and unique visits. Internet site
switching, but not application switching, is associated
with more F2F interaction.
End of day: Productivity and interaction.
Finally, we looked at the consequence of interactions and
multitasking on the end of the day self-reported
productivity. Here we addressed whether the amount of
Application Switching
App PC WR
PC x
WR
A
dj.
R
2
FB
F 25.1 22.53 0.45 1.06
0.24
df 1,161 1,161 2,161 2, 161
p 0.0001 0.0001 0.64 0.35
Email
F 56.23 7.21 29.34 20.41
0.49
df 1,161 1,161 2,161 2, 161
p 0.0001 0.008 .0001 .0001
Table 6. Model showing how average daily application
switches (App), project count (PC) and work role (WR)
affect frequency of FB and Email visits.
Internet Switching
Int PC WR
PCx
WR
A
dj.
R
2
FB
F 22.88 24.1 0.05 0.16
0.23
df 1,161 1,161 2,161 2, 161
p 0.0001 0.0001 0.95 0.86
Email
F 60.32 5.81 38.28 30.39
0.50
df 1,161 1,161 2,161 2, 161
p 0.0001 0.02 .0001 .0001
Table 7. Model showing how average daily Internet switches
(Int), project count (PC) and work role (WR) affect
fre
q
uenc
y
of FB and Email visits.
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
911
different types of workplace communications, along with
multitasking, were related to assessing productivity.
Remember that productivity self-reports were taken at the
end of each day. First, a GLM showed no significant
effect of Work Role. Next we looked at mood and
productivity. Based on the PANAS scale [29] (see Table
1), people who reported being more productive also
scored higher on their rating of positive mood at the end
of the day r=.30, p<.001, and scored lower on their rating
of negative mood at the end of the day: r=-.19, p<.04. In
terms of mood change over the course of the day,
(PANAS end of day - PANAS beginning of day), there
was a significant correlation of feeling productive and
developing a more positive mood over the course of the
day: r=.28, p<.004. Thus, the more productive people felt
their day was, the higher was their positive affect.
As productivity was measured as a Likert-scale item, we
conducted an ordinal regression with productivity as the
dependent measure. We created a variable of Total
Switches (i.e., all computer screen switches) by
combining App switches and Internet switches. We tested
a model using the independent variables of FB duration,
Email duration, F2F counts, and Total switches, and
included all 2-way interactions. We included Project
Count as a control variable.
Table 8 shows the significant variables that predict end-
of-day feeling of productivity, in the overall model. The
Cox and Snell pseudo R
2
is .18. We note that the
parameter estimates are not strong. There was a
significant negative relationship of email duration and
self-reported productivity: the more time spent on email
throughout the day, the less productive one feels. There
was also a negative relation of Total switches and
productivity: the more computer screen switches one does
over the course of the day, the lower the reported
productivity. F2F also shows a strong trend of a negative
relationship: the more F2F interaction one has over the
day, the less productive one feels.
We interpret the significant interaction of email and F2F
as follows. Though email duration and F2F duration alone
each had a negative relation with productivity, perhaps
F2F interaction that is work related is a catalyst to sending
emails that are work related, leading one to feel more
productive. F2F and FB also interact in a positive
relationship with productivity. One interpretation is that
though F2F alone is negatively related to productivity,
Model Parameter
estimate
Wald
statistic
(df=1)
P
Email duration -.007 4.83 .03
Total switches -.00001 4.56 .03
F2F -.009 3.27 .07
Email duration x F2F .004 4.30 .04
FB duration x F2F .001 6.18 .01
Table 8. Model of end-of-day self-report of productivity:
chi-square (6)=25.61, p<.0001.
Figure 2. A visual summary of the results of the three temporal perspectives. To simplify, only the results of the interactions are
presented. Red arrows indicate a negative association.
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
912
when people are engaged in work (along with higher FB
use [23]), then perhaps F2F interactions that are work
related lead to a higher feeling of productivity. Project
count as a control variable showed no significant effect;
the results hold irrespective of the number of projects one
works on.
Summary of interaction and multitasking
Based on our results, we present a visual summary of our
results (Fig. 2) showing the relationship of
communication and multitasking according to the three
temporal perspectives we examined.
DISCUSSION
In this paper we examined the relationship of multitasking
and communication from three different temporal
perspectives. First, we examined the attentional state of
people prior to a communication event. In HCI, it is
generally assumed that people work on tasks, are engaged
in them, and are then distracted by stimuli, such as email
or other people. Our results suggest a reverse perspective
for thinking about distractions. People may first be in a
certain attentional state which makes them susceptible to
distractions. We found that when people were in a rote
state (high engagement, low challenge), they were more
likely to do FB and have more F2F interaction. When
they were focused and aroused, they were likely to do
more email. We still found these relationships after
controlling for the same communication behavior prior to
the probe. In other words, our results suggest that
distractions could be explained by the current state of
one’s mind, i.e., one first experiences a particular
attentional state and then one is distracted by stimuli.
Our results in particular suggest that certain attentional
states are associated with different workplace
communication behaviors. FB interactions differ from
email interactions in that they are predominantly social
and casual. Therefore, when people are not challenged, it
makes sense that they might be susceptible to the
distractions of FB, which involve low effort. Email, on
the other hand, generally involves work, or as Barley [3]
claims, email can be regarded as a symbol of work (and
stress). It also makes sense then that when people are
focused and aroused, as we found, they are then in a state
where they are prepared to do email. We would like to
mention that we cannot claim causality; it is possible that
there are other underlying factors that could be associated
with activity and subsequent communication behaviors.
When people are in a rote state, as with FB use, they seem
to be more susceptible to F2F interactions. F2F
interactions can certainly involve either work or social
communications. As we did not record the content of F2F
meetings, we cannot distinguish the proportions of
communications of our participants that involved work.
However, the fact that a rote state was likely to precede
F2F interactions, and the fact that arousal showed no
significant predictive effect (unlike email), suggests that
most F2F communications of our sample may have been
social.
Our second temporal perspective examined how
workplace communications interleaved with multitasking
throughout the day. It is also possible that the practice of
Application and Internet site switching can also make
people susceptible to distractions (in our case, concerning
communications). Here we found that our three
communication types were involved in multitasking in
slightly different ways. Though imperfect, we used the
measure of application switching as a proxy for task
switching. The fact that both email and FB are associated
with application switching suggests that in the course of
switching between different applications, people take time
to check their email or FB. We expect that when one is
engaged in work-related projects, then they might seek
email for information or as [23] found, use FB as a work
break. It is possible then that application switching also
makes people susceptible to checking email, or FB; if one
is switching anyway, why not check to see if there are
recent emails related to work or FB postings that can offer
a break?
We offer possible explanations for the relationship of App
switching and Internet switching and communications.
First, App switching and Internet switching can introduce
opportunities for people to self-interrupt. As people are
exposed to a variety of different types of information
while switching applications, it is likely that some
information could trigger reminders that involve
communications with people. A second explanation is that
this behavior is habitual. People may simply check email,
FB and have F2F interaction because they have developed
patterns of such behavior over time. Habits can be
triggered by context and become deeply ingrained [30]. A
third explanation concerns emotional homeostasis which
we shall discuss shortly.
Project count was associated with FB use, and with
frequency of checking email. We would expect that the
more projects one works on, the more social connections
and dependencies one would have in the organization, and
this could influence email interaction. Also, following the
finding that FB might be a break from work [23], the
more projects one has, the more one might turn to FB as a
short relief from work. The fact that email duration was
associated only with application switches and not project
count, suggests that time spent on email is compelling
irrespective of how many projects one has.
Though our results on productivity are not strong, they do
suggest that computer screen switching, and amount of
time spent on email and in F2F interaction is detrimental
to feeling productive at the end of the day. Opportunity
cost could be an explanation: the more time spent in
workplace communications, the less time is available for
doi
ng
other types of work. Also, switching tasks involves
a cognitive cost of having to reorient [16] which could
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
913
result in a lower feeling of productivity. One might feel
that they have wasted their time on email or F2F
interaction, as people have a limited amount of time
during the day. The more time one spends in one activity
(e.g., email), the more time it takes away from another
activity perhaps more directly associated with feeling
productive (e.g., writing a research paper). This result
suggests that the time spent on workplace
communications via email, and F2F may not be
considered by people to necessarily be productive,
especially if the time spent on these communications was
for social reasons. Such interactions, though, may well be
associated with other productivity measures, and this calls
for further research.
Though our measure of productivity involved subjective
reports, this productivity measure actually can capture
many other underlying attitudes. We found, for example,
that productivity is highly correlated with positive affect.
Thus, subjective productivity could be a barometer for
happiness in the workplace.
Multitasking and homeostasis
Our results thus suggest that one's attentional state may
make one susceptible to certain kinds of distractions.
Further, application switching and Internet site switching
are also associated with a higher use of some
communication media. Our results lead us to return to our
earlier discussion about emotional homeostasis. We put
forth the following notion: people might move toward
online or offline communications that lead them to be in a
state where they are more balanced psychologically [4]
[9, 12, 19]. Perhaps people prefer to continue those
particular behaviors that will maintain their current
psychological state. For example, as we found, if people
are switching Internet sites (in a rote state, engaged but
not challenged) then perhaps they seek to continue
communication behavior in a similar type of attentional
state (a casual F2F interaction or FB use). Similarly, it
may be possible that, if people are already in a state of
focus, they may want to continue doing work that requires
a degree of focus. By continuing to experience this same
attentional state, they are trying to attain a psychological
balance, or emotional homeostasis. As claimed in prior
work [9, 12, 19], people desire to reduce tension and
maintain equilibrium.
However, switching activities could have varying
consequences. In some cases, people may switch tasks
and communication in order to try to attain emotional
homeostasis. On the other hand, there may also be a
cognitive cost to switching, as it may increase tension.
The cognitive cost may not only be in switching contexts
but also in switching attentional states. These different
potential outcomes lead us to distinguish between external
and internal interruptions and activity switching. External
switches are triggered by sources outside of a person, e.g.,
another person, an email notification, or a telephone
ringing. Internal switches are triggered by oneself, i.e., a
person chooses to switch their activity due to their current
needs and motivations. Some research suggests that
external interruptions are associated with stress [21]. We
propose that it is the internal switches due to one's own
volition that could be explained by emotional
homeostasis.
Internal switches could be geared toward achieving
equilibrium. We see two ways that the notion of
homeostasis applies. First, perhaps people self-interrupt to
reduce tension from their current activity, as reducing
tension is geared towards achieving a balance, per
Lewin's field theory notion [19]. Some support for this is
suggested by [23] which found that Facebook use is
associated with a more positive mood. People may turn to
social media such as Facebook as a social break to thus
reduce current tension that is experienced. A second way
that emotional homeostasis could apply is that people may
self-interrupt to a particular activity that enables them to
continue their current attentional state. For example, if
people are already in a focused state, then they are already
in a state conducive to email use (see [23]). It may require
effort to move from a state of low challenge to one of
high challenge (a characteristic of focus, see Table 1).
Thus, if people are already in a state of high challenge and
high engagement, it is less effort to simply continue to do
activities that maintain this attentional state. This would
enable one to maintain a balanced emotional state.
We did not collect contextual information to enable us to
distinguish self-interruptions from external interruptions.
However, we believe that most FB use was due to self-
interruption. We also polled our participants on their
email behavior, and of the 15 who replied, 67% reported
that they self-interrupt half or more of the time to check
email, as opposed to being externally interrupted by an
email notification. F2F interaction could be due to either a
self-interruption or external interruption. Therefore, we
believe that our results could be consistent with the idea
that people self-interrupt to try to move towards a state of
equilibrium.
We therefore propose the following hypothesis: people
switch tasks (and consequently, attentional states) so as to
attain an emotional equilibrium. Perhaps people seek out
those communication media and those communication
partners to attain a balanced state. Thus, perhaps we turn
to our colleagues and friends for emotional homeostasis?
This is a hypothesis that can be empirically tested in
future research.
Implications for multitasking research
Our study can be used to inform other multitasking
research. Our results could lead to further investigations
about what other subsequent behaviors could be
associated with attentional states, including the use of
other types of digital and communication media.
Moreover, an understanding of how state transitions occur
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
914
and what the associated activities are could be an
interesting avenue to explore. Also, a wider range of
attentional, as well as emotional states could also be
examined. The idea of emotional homeostasis, a new idea
in multitasking research, should be further studied as well.
Our results contrasting application switching and Internet
switching could lead to other investigations around why
Internet switching might lead people to check and spend
more time on FB and Email, as well as engage in more
F2F interactions. Is this behavior habitual or are there
perhaps other factors involved, such as that Internet
switching provides cues which lead one to switch more to
these tasks? All of these questions invite further scrutiny.
Our goal is to understand the workplace experience so as
to provide insight on how people can improve their
experiences. Our results on productivity self-reports,
coupled with our multitasking results, suggest that people
could feel more productive (and consequently, happier) in
the workplace if they could have a better understanding of
their own workplace communication behavior. We feel
that our results can be used to inform the design of
workplace tools that could provide people with better
feedback on their communication patterns and activity
switching behaviors.
Limitations
Our participants all had at least a Bachelor’s degree and
half were researchers. Therefore, we can only generalize
our results to highly educated information workers. It is
very possible that the ESM probes could have led to
switching behaviors, as they interrupted participants.
Interruptions have been recognized as an issue with ESM
[13]. However, the probe could be answered in a few
seconds and the participants were instructed to cancel any
probe that they could not answer due to it interrupting
them. When asked after the study, some participants
reported that this interruption side effect was a hindrance.
We used SenseCam photos as a proxy for face-to-face
interaction. Our software could not distinguish unique
faces--we can only use the counts as an estimate of
amount of F2F interaction, not number of different
interactants. Therefore, the F2F interaction measure
should be regarded as a rough proxy of how much F2F
interaction one had, not number of people. The SenseCam
likely underestimated interaction counts since a photo was
taken about every 15 seconds; an interaction that occurred
between shots would not have been captured.
CONCLUSIONS
Our study raises new issues concerning multitasking and
distractions. Our results suggest that, contrary to what has
been assumed for some time in multitasking research,
people may first be in a particular attentional state that
makes them susceptible to being distracted. Our results of
window switching due to attentional and interaction
baselines also suggest that this type of multitasking
behavior could encourage further communication
behavior, some of which could be distractions from work.
The relationship of multitasking and communications (as
potential distractors or attractors) is very complex and we
hope that our study can lead to new research directions to
gain a deeper understanding of the topic.
ACKNOWLEDGMENTS
This material is based upon work supported by the NSF
under grant #1218705.
REFERENCES
1. Adler, R.F. and Benbunan-Fich, R. Self-interruptions
in discretionary multitasking. Computers in Human
Behavior, 2013, 29 (4). 1441-1449.
2. Aral, S., Brynjolfsson, E. and Alstyne, M.W.V.
Information Technology and Information Worker
Productivity. Information Systems Research, 2011.
3. Barley, S., Meyerson, D. and Grodal, S. Email as a
Source and Symbol of Stress. Organization Science,
2011, 22 (4). 887-906.
4. Buck, R. Social and emotional functions in facial
expression and communication: The readout
hypothesis. Biological Psychology, 1994, 38. 95-115.
5. Chrousos, G. and Gold, P. The concepts of stress and
stress symptom disorders: Overview of physical and
behavioral homeostatis. JAMA, 1992, 267 (9). 1244-
1252
6. 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.
7. Dabbish, L., Mark, G., #237, Gonz, c.M., #225 and
lez. Why do i keep interrupting myself?:
environment, habit and self-interruption Proceedings
of the SIGCHI Conference on Human Factors in
Computing Systems, ACM, Vancouver, BC, Canada,
2011, 3127-3130.
8. Ean, L.C. Face-to-face Versus Computer-mediated
Communicaton: Exploring Employee's Preference of
Effective Employee Communication Channel.
International Journal for the Advancement of Science
and Arts, 2010, 1 (2). 38-47.
9. Festinger, L. A Theory of Cognitive Dissonance.
Stanford University Press, Stanford, CA, 1957.
10. 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.
11. Grandey, A.A., Tam, A.P. and Brauburger, A.L.
Affective States and Traits in the Workplace: Diary
and Survey Data from Young Workers. Motivation
and Emotion, 2002, 26 (1). 31-55.
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
915
12. Heider, F. Social perception and phenomenal
causality. Psychological Review, 1944, 51 (6). 358-
374.
13. Hektner, J., Schmidt, J. and Csikszentmihalyi, M.
Experience Sampling Method: Measuring the quality
of everyday life. Sage, Thousand Oaks, CA, 2007.
14. Herzberg, F. The Motivation to Work. Transaction
Publishers, 1993.
15. 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.
16. 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.
17. 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.
18. Kohler, W. Dynamics in Psychology. Liverright, New
York, 1942.
19. Lewin, K. Defining the "Field at a Given Time.".
Psychological Review, 1943, 50. 292-310.
20. Mark, G., Gonzalez, V.M. and Harris, J. No task left
behind?: examining the nature of fragmented work in
Proceeding of the SIGCHI conference on Human
factors in computing systems ACM Press, Portland,
Oregon, USA 2005, 321-330
21. Mark, G., Hausstein, D. and Klocke, U. The cost of
interrupted work: more speed and stress Proceedings
of the SIGCHI Conference on Human Factors in
Computing Systems, ACM, Florence, Italy, 2008,
107-110.
22. Mark, G., Iqbal, S.T., Czerwinski, M. and Johns, P.
Bored mondays and focused afternoons: the rhythm
of attention and online activity in the workplace
Proceedings of the 32nd annual ACM conference on
Human factors in computing systems, ACM, Toronto,
Ontario, Canada, 2014, 3025-3034.
23. Mark, G., Iqbal, S.T., Czerwinski, M. and Johns, P.
Capturing the Mood: Facebook and Face-to-Face
Encounters in the Workplace ACM Conference on
Computer Supported Cooperative Work, 2014, (to
appear).
24. Mark, G., Wang, Y. and Niiya, M. Stress and
multitasking in everyday college life: an empirical
study of online activity Proceedings of the SIGCHI
Conference on Human Factors in Computing
Systems, ACM, Toronto, Ontario, Canada, 2014, 41-
50.
25. Neuberg, S.L., Kenrick, D.T. and Schaller, M.
Human Threat Management Systems: Self-Protection
and Disease Avoidance. Neuroscience and
Biobehavioral Reviews, 2011, 35. 1042-1051.
26. O'Conaill, B. and Frohlich, D., Timespace in the
workplace: dealing with interruptions. in CHI,
(1995), 262-263.
27. Posner, J., Russell, J.A. and Peterson, B.S. The
circumplex model of affect: An integrative approach
to affective neuroscience, cognitive development, and
psychopathology. Development and
Psychopathology, 20
05
, 17. 715-734.
28. 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.
29. 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.
30. Wood, W. and Neal, D.T. A new look at habits and
the habit-goal interface. Psychological Review, 2007,
114 (4). 843-863.
Technologies in the Workplace
CSCW 2015, March 14-18, 2015, Vancouver, BC, Canada
916
