Social Media and Health
#chi4good, CHI 2016, San Jose, CA, USA
Sleep Debt in Student Life:
Online Attention Focus, Facebook, and Mood
1
1
2
2
Gloria Mark , Yiran Wang , Melissa Niiya , Stephanie Reich
1
2
Department of Informatics
School of Education
University of California, Irvine
University of California, Irvine
{gmark,yiranw2@uci.edu}
{mniiya,smreich@uci.edu}
This reflection on sleep by the poet Keats was written
almost 200 years ago. Yet in our current digital age, the
topic of sleep (or lack thereof) is as important as ever,
widely discussed in the academic community as well as in
the popular press. One dominant theme has been how
people rarely get enough sleep--the term "sleep debt" is
actively discussed in the public sphere. Sleep deprivation
can have serious consequences, leading to errors in the
workplace [23], accidents [2], and even to falling asleep
involuntarily at critical times, such as while driving [2].
However, how sleep deprivation might affect how people
use information technology (IT) has yet to be explored.
ABSTRACT
The amount of sleep college students receive has become a
pressing societal concern. While studies show that
information technology (IT) use affects sleep, here we
examine the converse: how sleep duration might affect IT
use. We conducted an in situ study, and logged computer
and phone use and collected sleep diaries and daily surveys
of 76 college students for seven days, all waking hours. We
examined effects of sleep duration and sleep debt. Our
results show that with less sleep, people report higher
perceived work pressure and productivity. Also, computer
focus duration is significantly shorter suggesting higher
multitasking. The more sleep debt, the more Facebook use
and the higher the negative mood. With less sleep, people
may seek out activities requiring less attentional resources
such as social media use. Our results have theoretical
implications for multitasking: physiological and cognitive
reasons could explain more computer activity switches:
related to less sleep.
Considering the current tendency of information workers
and students to be online and accessible for interruptions
much of the day (and night), the relationship of IT use and
sleep is an important issue. Over the last decade, there has
been discussion of how constant connectivity is associated
with multitasking [32], interruptions [12], and checking
email [32], during what might normally be sleeping hours,
disrupting or shortening sleep.
Author Keywords
Multitasking; sleep; productivity; computer logging; in situ
study; sensors; social media; Facebook; mobile phone;
mood
In human-computer interaction (HCI), attention has been
given to prototypes and apps that provide novel ways to
measure sleep [3, 11, 24], including sensing social data
[38]. In other fields, although studies suggest that IT use
might interfere with sleep, cf [1], there have not yet been
studies showing the converse: how sleep duration might
affect IT use. Why is it important to consider how sleep
might affect IT use? First, if a connection between sleep
duration and IT use is indeed found, then this would point
to the notion of sleep as a variable that should be
considered in HCI research, particularly as it relates to user
performance with digital media. Second, this research could
identify specific computer behaviors that are present when
people are lacking sleep, which could prevent human error.
Third, such research could inform the design of
technological interventions that might lead people to
improve their sleep habits, e.g. by recognizing and
informing about digital media behaviors that might be
impacted by sleep deprivation.
ACM Classification Keywords
H.5.3 [Information Interfaces and Presentation]: Group and
Organization Interfaces; K.4.m [Computers and Society]:
Miscellaneous.
INTRODUCTION
"O
magic
sleep!
O
comfortable
That broodest o’er the troubled sea of the mind...."
bird,
John Keats, in Endymion, 1818
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. Copyrights for
components of this work owned by others than the author(s) must be
honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee. Request permissions from
Permissions@acm.org.
CHI'16, May 07 - 12, 2016, San Jose, CA, USA
Copyright is held by the owner/author(s). Publication rights licensed to
ACM.
ACM 978-1-4503-3362-7/16/05 $15.00
DOI: http://dx.doi.org/10.1145/2858036.2858437
The purpose of this study is to examine how sleep might
affect IT use. Rather than rely on laboratory settings, we
conducted an in situ study to capture real world
contingencies that could play a role in the relationship
between sleep and IT use. Sleep deprivation in young
5517
Social Media and Health
#chi4good, CHI 2016, San Jose, CA, USA
people, especially college students, has been given much
attention, particularly on how sleep is affected by
technology use (cf 1, for a review). As a first step to
studying this converse relationship, we examine how
technology use is affected by sleep duration with college
students. Our results show that with less sleep, students
perceive higher work pressure and productivity, spend
shorter durations of time focusing their attention online,
increase their time on Facebook, and experience more
negative mood. These results contribute to the growing
body of research examining the relationship of sleep and IT
use.
evening with negative effects on sleep as well as on
academic performance [9]. Thus, while there is mounting
evidence that IT use consequently affects sleep, no one has
looked at the opposite relationship, which we explore in this
study.
Sleep and Social Media Use
Increasingly, research has documented how social media
might contribute to sleep deficiencies, largely from people
staying awake to stream videos, peruse profiles, and chat
online [39, 50]. However, studies are lacking on how sleep
duration might affect social media and computer usage.
Such relationships are quite plausible. Social media is often
viewed as a lightweight activity [52] with socializing,
escape, and distractions being common reasons people
report for using social media [1, 47]. There is evidence of
an association between more time on social networking
sites and less effortful thinking [55]. Thus, sleep deprived
individuals who are online may likely engage in less
cognitively demanding and more distracting activities such
as social media use.
SLEEP AND INFORMATION TECHNOLOGY USE
Sleep Duration and Performance
Lack of sleep is linked to numerous negative cognitive and
health outcomes [23]. Lack of sleep can lead to reduced
memory and cognitive functioning [23, 30], poor academic
and work performance [23, 43], and safety risks such as
falling asleep while operating vehicles [2]. Sleep
deprivation and its effects remain prevalent among the US
adult population, with 37% of adults aged 20-39 years
reporting less sleep than the recommended 7-9 hours per
night [11]. Those people who stay up late or who are
categorized as “evening types” report attentional and
emotional problems [35].
Sleep and Multitasking
A number of studies have examined multitasking, or
activity switching, while working with digital media, [e.g.
12, 19]. Multitasking can occur by switching attention not
only among computer screens but also between devices
[22]. These studies have often pointed to the role of
interruptions (from external sources or oneself) [19], email
[32], or habitual behavior [19] in influencing multitasking.
Although not as thoroughly explored, sleep may also affect
technology use and multitasking. For instance, one study
found that those who ended their daily activity and went to
sleep earlier tended to multitask less the next day [33]. This
suggests that perhaps sleep patterns might be associated
with attention shifting.
Sleep debt
Sleep debt is the cumulative loss of, and consequent
pressure for sleep, that results from an inadequate amount
of physiologically normal sleep [51]. Sleep debt can accrue
in a variety of ways including going to bed later but
keeping wake time the same, having untreated sleep
disorders, or experiencing external sleep disturbances.
Since people vary in the amount of sleep they need, sleep
debt is based on the “cumulative hours of sleep loss with
respect to the subject-specific daily need for sleep" [51].
Numerous studies have found sleep debt to impact a host of
outcomes such as mood [18], weight gain [4], performance
[42], and health [29].
Studies of sleep deprivation and attention have found
insufficient sleep to result in decreased visual and auditory
attention [20]. Less sleep could thus contribute to less
cognitively taxing behaviors while on the computer such as
having less persistent attention with more switching of
focus between windows and applications. This alteration in
attention also applies to task switching [17]. It is quite
plausible that similar low-attention behaviors, due to less
sleep, could occur while on the computer as well. Thus, less
sleep could contribute to higher online multitasking in the
following ways. It could lead to more frequent switching of
attention to different computer screens. This switching
could be due to more self-interruptions, which comprise
almost half of all interruptions in the workplace and can
lead to task switches [12, 19]. Self-interruptions could be
due to higher distractibility, and lack of sleep could make
people more prone to distraction and thus more likely to
switch focus from their current task at-hand.
College Students and Sleep Duration
Young adults may be particularly at risk for sleep
disturbances and deprivation [6]. Lack of sleep is prevalent
among college students, especially women [8]. Poor
sleeping habits have negative impacts on undergraduate
physical and mental health as well as academic
performance such as lower test scores and grade point
averages [43, 49]. Poor sleep quantity and quality, which
include late nights, earlier wakeup times [49], and irregular
sleep-wake patterns [28] are associated with lower
academic performance, suggesting that routines around
sleep, in addition to sleep amount, affect college students.
Constant availability via mobile phones and the Internet
may lead to sleep disturbances. For example, among a
sample of young people, smartphones were often used
before bed and caused frequent sleep interferences [25].
Among adolescents, IT use tends to stretch late into the
RESEARCH QUESTIONS
In contrast to prior studies, which focused on how IT use
affects sleep, we look at the converse: how sleep duration
5518
Social Media and Health
#chi4good, CHI 2016, San Jose, CA, USA
might affect how people use IT. While numerous studies
point to the detrimental effects of having less sleep [23, 30,
43, 49], amount of sleep could also affect IT use. Further,
while we expect sleep debt to be related to sleep duration,
there are differences. One night of short sleep may affect
some performance measures the next day and not others.
However, sleep debt represents an accumulation of the
effects of loss of sleep. In other words, consistently
sleeping shorter than one's ideal amount could accumulate
and produce effects on different variables than sleep
duration. We thus examine sleep duration the night before
and sleep debt with different variables. The following are
our research questions and hypotheses.
lightweight, requiring little effort and often serving as a
quick break (see [52]). As such, it is not surprising that
youth report using social media for such things as
distraction, a way to fill time, and as a mechanism for
socializing [47]. Further, college students who use social
media, especially social networking sites like Facebook,
report feeling a need to check these sites regularly to stay
connected to others [48]. This is consistent with other
research that suggests that Facebook use is habitual [13].
We would not expect social media habits like Facebook use
to change so easily after one night of sleep loss but rather
would expect to see changes over longer term sleep loss
accumulation (i.e. sleep debt). For this research question we
examine sleep debt, as we are interested in whether we find
a relationship of accumulated sleep loss with the tendency
to increase lightweight activities.
RQ1: What are the impacts of sleep on college students’
perceptions of productivity and work pressure? Before
examining sleep effects on IT use, we begin by looking at
related experiences to sleep duration. Decades of research
have shown that sleep and health are related and that
deficits in sleep can lead to personal and fiscal costs [23].
Studies of worker productivity and sleep patterns have
found that decreased sleep leads to reduced memory,
cognitive functioning, and work performance [30]; and for
college students, sleep deficits result in lower academic
performance [43, 49]. Conversely, when students are well
rested, they tend to perform better academically and
exercise more [5]. Thus, we explore whether less sleep
duration the night before is associated with higher
subjective work pressure and lower perceived productivity.
We hypothesize that:
When students are more sleep deprived and using
technology, they might seek out activities involving fewer
cognitive resources [15]. When working online, they may
also be more likely to be distracted from work, and social
media is reported as a common form of distraction [47]. A
study of Facebook use and student engagement, using the
National Survey of Student Engagement (NSSE) showed a
negative relationship of Facebook use and engagement in
educational activities [21]. It has also been observed in the
workplace that when people are engaged in computer work
Facebook affords a light break from work [31]. The most
common social media site among adults ages 18-29 is
Facebook, with 87% reporting using it in 2014 [14]. Others
have also reported Facebook as the most popular social
media site among college students [21]. We therefore focus
on Facebook use and thus expect that when students have
higher sleep debt, when online they would be more likely to
spend more time doing Facebook as a lightweight activity
and also as a distraction.
H1: Students with shorter sleep duration the night before
will have higher perceived work pressure the next day.
H2: Students with shorter sleep duration the night before
will rate their productivity to be lower the next day.
RQ2: Is sleep related to college students’ multitasking on
the computer? Surveys of college students have found high
association between online computer use and sleep deficits
for that evening [37]. Other research shows that longer
duration of focus on the computer is related to less stress
[32]. Coupling these patterns with evidence that decreased
sleep reduces attention and focus [26], computer behavior
when people had less sleep might involve more switching
and less focused attention. A higher frequency of switching
on the computer has been considered as a measure of higher
multitasking [32]. In other words, from a fine-grained
perspective it is an indicator of switching attention between
different activities as screens and content change. Thus, we
expect that less sleep should be associated with higher
multitasking, as reflected by shorter attention duration on
any computer screen. Thus, we expect that:
H4: Students with more sleep debt will engage in more
Facebook use.
RQ4. How is sleep debt related to mood? A meta-analysis
of laboratory sleep studies found that sleep deprivation
significantly negatively impacts mood [42]. However, in
contrast, a study of college students who were sleep
deprived for 24 hours showed no significant changes in
mood, such as negative mood states of anger and anxiety
[43]. The authors explain this to the notion that 24 hours of
sleep deprivation is long enough to affect fatigue but not
long enough to impact mood. A large cross-sectional survey
study of college students though did find a relationship
between self-reported sleep disturbances and negative mood
[28]. Because sleep loss over 24 hours did not affect mood,
we expect that longer accumulated sleep could affect mood.
Therefore in this research question we examine sleep debt.
Based on these studies we expect that the more sleep debt
one has, the more it would negatively impact mood. Most
studies of the effects of sleep deprivation and mood have
been done in the laboratory or with surveys. We are not
H3: Students with shorter sleep duration the night before
will have shorter focus duration on computer windows the
next day.
RQ3: Is sleep debt associated with Facebook use? Social
media consumption (as opposed to production) is often
5519
Social Media and Health
#chi4good, CHI 2016, San Jose, CA, USA
aware of studies that have examined the relationship of
sleep deprivation and mood where people have been
tracked in an in situ environment. Therefore, we
hypothesize:
user opened a new window or switched between already
open windows. A window was an application or a web
browser tab. Each log record included the start time and
duration of the active window, the name of the application,
a URL if the window was a web browser tab, and idle time.
Timestamps were recorded to the second. Only time spent
in the current window was measured. For example, if a
webpage was open in the background when the user used
Word in the foreground, Kidlogger only counted the time
spent in Word. Computer idle time was not included. Phone
logging over seven days was collected using the AWARE
Framework application (www.awareframework.com). A
log record was generated each time a user opened an
application, web browser tab or switched between apps.
The log captured activity to the millisecond. Data from
computer and phone logging was partitioned into daily logs
and numbered, generating at least seven days of
computer/phone use per person. Due to scheduling issues,
for some participants more than 7 days of data was
collected. The distribution of full days of data collection
was as follows: 58 students have 5 full days (7 days in
total); 9 students have 6 full days (8 days in total); 1 student
has 8 full days (10 days in total); 3 students have 7 full days
(9 days in total); and 1 student has 8 full days (10 days in
total).
H5: Students with more sleep debt will experience more
negative mood.
METHODOLOGY
We conducted an in situ observational study at a large
public university on the U.S. west coast in Jan.-May of
2014. We used a mixed-methods approach with automatic
computer and phone logging as the primary method of data
collection. In order to capture the context of students’ life at
school, their sleep schedule, and mood, we also used daily
surveys and a one-time general survey about attitudes and
demographic information. Computer and phone logging and
other data collection occurred over seven days.
Participants and Procedure
Participants were recruited from undergraduate courses,
student residences, and snowball sampling. In total, we
collected data from 76 undergraduates: 34 males and 42
females. Due to the availability of monitoring software and
its limitations, the study was restricted to students who used
both Windows computers and Android phones. Students’
ages ranged from 18 to 23 years (mean=19.3) and the
median college year was sophomore.
Facebook (FB). Measures of FB usage from both the
computer and phone logs were used to calculate frequency
and duration of visits. The duration of time on FB from
computer and phone logs were added together to produce a
total daily time on Facebook.
On Day 1 of the study, participants visited a campus
laboratory where the computer and phone logging software
were installed on their devices. Students who also had
desktop computers were given software installation
instructions. Participants were sent a link to an online
general survey on Day 1 and instructed to complete it
before the end of the study (by Day 7). Two daily surveys
were sent each day: one at 7 a.m. (a sleep diary for the night
before), asking participants to complete it once awake and
another at 9 p.m. (an end-of-day survey), asking them to
complete it before going to bed. On Day 7, semi-structured
exit interviews (for about 30 minutes) were conducted.
Participants were asked about their experiences during the
study, their technology habits, and their beliefs about how
technology could affect stress, productivity, and mood.
Descriptive coding was used to identify key themes that
emerged from these interviews [46]. Participants were paid
$100 for their participation.
Focus duration. Multitasking can be considered from
different perspectives: at a broad level examining task
switching, or at finer-granularity, examining attention shifts
among different activities which could be within the same
task. We adopt this fine-granularity perspective to
investigate the relationship of sleep and its effects on how
people shift their attention when working on the computer.
We feel that measuring duration of focus on computer
screens is a reasonable proxy for attention duration with
computer work, and this has been used as a measure of
attention in multitasking, cf [32]. By computing its inverse,
average duration of focus can alternatively be converted
into a measure of average switching among different
computer windows per time unit. Focus duration is
measured as the average duration spent on any active
window, based on the logging software.
Measures
A total of 71 people were used in the analysis. Five people
were excluded due to missing computer data (logging
software problems, software getting uninstalled by antivirus
software, and one participant uninstalling the software). For
these 71 participants, we collected computer and phone
logs, sleep duration, social media use and daily surveys, and
single general surveys.
Sleep duration. Daily sleep duration was measured via a
self-report sleep diary, modified from [36]. Starting from
Day 2 of their study, participants were asked to enter the
time they went to bed the previous day (hour and minute)
and the time they woke up (hour and minute). Thus, data
were collected for at least six nights (sleep on Days 1-6).
Five days is considered sufficient to produce reliable
measures of sleep [45]. Sleep diaries have been found to be
very accurate [7], are commonly used in sleep studies [5, 8,
Computer and phone logging. Computer activity over seven
days was captured with Kidlogger (kidlogger.net). This
Windows freeware generated one log record each time a
5520
Social Media and Health
#chi4good, CHI 2016, San Jose, CA, USA
9, 41] and have been used as a gold standard when testing
sleep-monitoring devices, e.g. [34]. Self-reported sleep has
also been validated with actigraphs [54]. Sleep diary reports
were included in the end-of-day surveys described below.
productive do you feel you were today? (using a Likert
scale, 1= Not at all; 5=Extremely).
Sleep debt. People can accumulate sleep debt on days when
they work and they generally compensate for loss of sleep
on free days. Sleep debt is operationalized as follows. The
larger the difference between mid-sleep time on free days
compared to mid-sleep time on work days, the larger is
considered the sleep debt [44]. The mid-sleep time is
calculated as the mid-point between sleep onset and
awakening. The analysis by Roenneberg and colleagues
[44] shows that most people compensate for sleep loss
during the week (work days) by sleeping longer on
weekends (free days). Therefore, the ideal mid-sleep time
on free days is calculated from observed mid-sleep time on
free days, corrected by the weekend and weekday sleep
durations, by the following formula recommended by [44]:
•
Sleep Duration: (Sleep diary) the amount of sleep selfreported daily based on time to sleep the previous night
and time one woke up the next morning. Greater values
for sleep duration reflect a greater amount of sleep.
•
Sleep Debt: calculated by the formula above, based on
self-reported daily sleep. Greater values represent a
greater amount of sleep debt.
•
Total Computer Duration: the total daily active time on
the computer, based on the logging software.
•
Focus Duration: the average duration spent on any
active computer window, based on the logging
software. One outlier was removed.
•
FB Duration on Computer/Phone: the total daily FB
use on the computer and phone combined, based on the
logging software. One outlier was removed. A square
root transformation was done to improve normality.
•
Affect balance: the positive daily PANAS score minus
the negative daily PANAS score (End-of-day survey)
•
Interviews: participant responses to semi-structured
interviews of their technology experiences.
MSFSC = MSF - .05*(SDF-(5*SDW + 2*SDF)/7)
where MSF=mid-sleep time on free days, SDF=sleep
duration on free days, and SDW is sleep duration on work
days. (5*SDW + 2*SDF)/7 is the average weekly sleep
need.
Control variables:
Based on students' schedules, we used weekday as a
"workday" and weekends were "free days". We took the
difference between the mid-sleep time of each workday and
the ideal mid-sleep time, as the daily sleep debt. We took
the absolute value, since the theory of sleep debt is that it is
the absolute departure from the ideal mid-sleep time that is
representative of sleep debt. It is the disruption from ideal
sleep that matters. As sleep debt has a long tail, we used a
log transformation to make the distribution normal.
•
Age, gender (General survey)
•
Workload: number of course credit units taken at the
time of the study (General survey)
•
Deadlines: (From the end-of-day survey): How much
did deadlines influence you today? (using a Likert
scale, 1= Not at all; 5=Extremely) (End of day survey).
RESULTS
We collected more than 1,720 hours of computer logs from
71 participants, excluding computer idle time. Most
participants reported in exit interviews that the week of the
study was representative of a typical school week.
Mood. Mood was measured using the PANAS scale at the
end of every day. The PANAS [53] is a well-validated scale
that measures two dimensions of mood: positive and
negative affect. Following [27], we created a measure of
Affect Balance, which refers to a balance between positive
and negative affect, where the negative score is subtracted
from the positive score.
Overview of results
Survey measures. Participants filled out daily end-of-day
surveys where they answered questions on work pressure
and perceived productivity. Other daily measures were also
asked but are not covered in this paper. A general survey
asked students for demographic information, technology
use habits, course load, and class standing.
We begin by reporting an overview of the results. Table 1
shows descriptive statistics of the measures of interest. The
average sleep duration in our sample was 7.9 hours, with a
wide range of 2.6 to 14.0 hours. Students showed a positive
sleep debt, meaning that on average they accumulated sleep
loss. Focus duration on any computer window averaged 1
minute, 38 seconds. Sleep duration and sleep debt are
weakly negatively correlated: r= -.11, p
Purchase answer to see full
attachment