Conference PaperPDF Available

Investigating The Role of Task Engagement in Mobile Interruptibility

Authors:

Abstract and Figures

Context-awareness of mobile phones is a cornerstone of recent efforts in automatic determination of user interrupt-ibility. Modalities such as a user's location, her physical activity , time of day, can be used in machine learning models to infer if a user is going to welcome an incoming notification or not. However, the success of context-aware inter-ruptibility systems questions the existing theory of interrupt-ibility, that is based on the internal state of the user, not her surroundings. In this work we examine the role of a user's internal context, defined by her engagement in the current task, on the sentiment towards an interrupting mobile notification. We collect and analyse real-world data on interrupt-ibility of twenty subjects over two weeks, and show that the internal state indeed impacts user interruptibility.
Content may be subject to copyright.
Investigating The Role of Task
Engagement in Mobile Interruptibility
Veljko Pejovic
Faculty of Computer and
Information Science,
University of Ljubljana
Slovenia
Veljko.Pejovic@fri.uni-lj.si
.
Abhinav Mehrotra
School of Computer Science,
University of Birmingham
UK
axm514@cs.bham.ac.uk
Mirco Musolesi
Department of Geography
University College London
UK
mirco.musolesi@ucl.ac.uk
Permission to make digital or hard copies of part or all 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 third-party components of this work must be honored.
For all other uses, contact the Owner/Author.
Copyright is held by the owner/author(s).
MobileHCI ’15 Adjunct, August 25–28, 2015, Copenhagen, Denmark.
ACM 978-1-4503-3653-6/15/08.
http://dx.doi.org/10.1145/2786567.2794336
Abstract
Context-awareness of mobile phones is a cornerstone of
recent efforts in automatic determination of user interrupt-
ibility. Modalities such as a user’s location, her physical ac-
tivity, time of day, can be used in machine learning models
to infer if a user is going to welcome an incoming notifica-
tion or not. However, the success of context-aware inter-
ruptibility systems questions the existing theory of interrupt-
ibility, that is based on the internal state of the user, not her
surroundings. In this work we examine the role of a user’s
internal context, defined by her engagement in the current
task, on the sentiment towards an interrupting mobile notifi-
cation. We collect and analyse real-world data on interrupt-
ibility of twenty subjects over two weeks, and show that the
internal state indeed impacts user interruptibility.
Author Keywords
Interruptibility; Notifications; Multitasking; Context-aware
computing.
ACM Classification Keywords
H.5.2. [Information Interfaces and Presentation (e.g. HCI)]:
User Interfaces; H.1.2. [Models and Principles]: User/Machine
Systems
Context-Aware Mobile Interruption Management
The mobile phone represents the most direct point of con-
tact with almost any individual on the planet, and, since
the smartphone revolution, also the most common means
of getting the information from the Internet. Being every
persons main communication link, the mobile often over-
whelms its user by information. Consequently, efforts have
been made to harness sensing and computing capabilities
of modern mobiles to design smarter interruption systems
that will not disturb users at times when they prefer to work
uninterrupted.
The role that the context may have in determining user in-
terruptibility has been recognised even before the smart-
phone era. In [3] Horwitz et al. build a Bayesian network
that describes user-defined cost of interruption from con-
textual audio-visual features captured by a camera and a
microphone in an office setting. Ho and Intille presented the
first study of mobile interruptibility, in which wearable ac-
celerometers were used to collect information about users’
physical activity, while mobile notifications were sent at the
same time [2]. Their study points out that the moments of
changing physical activity represent preferred moments for
interruption.
The smartphone has changed the way we reason about in-
terruptibility. Its sensing and computing capabilities enable
context-awareness previously explored only in so called
Wizard of Oz studies [1]. Naturally, efforts have been made
to infer user interruptibility from a richer physical context.
Ter Hofte builds a model of the interruptibility of a smart-
phone user, that is based on a user’s location, social cir-
cle and physical activity [9]. Yet, this work does not em-
ploy mobile sensing, but relies on self-reported information.
Pielot et al. examine a communication- and usage-oriented
context, by collecting and analysing a data set of text mes-
sages exchanged via smartphones together with the as-
sociated phone usage data [7]. Time since the screen was
on, time since the last notification, and similar features were
used in a classifier that infers if the users is going to attend
the message within a short time frame. InterruptMe [6], on
the other hand looks at the context in a more physical way.
The system senses user’s semantic location, physical ac-
tivity and time of day, and connects this context with one’s
interruptibility through an evolving classifier.
Mobile Phone and the Theory of Multitasking
Not all interruptions are equally disruptive, and the impor-
tance of improving our understanding of disruptiveness
was exemplified by the increasing amount of interruptions
brought by the latest waves of communication technologies.
In the traditional theory of interruptibility three main factors
that render an interruption disruptive have been identified:
the interrupting task complexity, task duration, and the mo-
ment of interruption. A unifying theory of interruptibility is
thoroughly explained in [8]. In a nutshell, our brain pro-
cesses the current task in the procedural memory, while
the declarative memory serves for the short- and long-
term storage of facts. Once interrupted, the primary tasks is
moved to the declarative memory, and the interrupting task
takes over the procedural memory. Reactivation of the pri-
mary task after the interrupting task is finished may require
recollection from the declarative memory. This recollection
tends to be less successful if the interrupting task was com-
plex and/or long. Time to rebuild the primary task’s problem
state also depends on the complexity of the primary tasks,
and the familiarity of the subject with the primary task –
the more practice one has with a certain task, the stronger
the activation bound in the declarative memory will be. The
moments of low cognitive workload, even the moments of
break within a more complex task, tend to be more suitable
for interruptions, as shown in experimental studies [5].
However, from the mobile device point of view the estima-
tion of the task complexity and the user’s skills pertaining to
the task, remain extremely challenging. Measuring the en-
gagement of a user in a task, in order to identify low work-
load moments, has been demonstrated, with a limited suc-
cess, via pupil dilation observation [5]. But the question we
pose in this paper is: Are task complexity and user engage-
ment still relevant determinants of a user interruptibility in
the mobile context?
All the above studies have been performed in tightly con-
trolled laboratory settings and without mobile devices. Fur-
thermore, the theory of multitasking has been constructed
on the data from unavoidable interruptions. In the mobile
realm, interruptions are delivered via notifications. Here,
in a so called bounded deferral manner [4], a user can de-
cide to postpone the reaction until a more suitable moment.
Given these differences, we decide to experimentally eval-
uate the relevance of task complexity and engagement on
mobile interruptibility.
Experimental Setup
To experimentally investigate the relationship between task
engagement and interruptibility, we develop and deploy an
Android-based experience sampling method (ESM) applica-
tion. The application occasionally delivers a notification on
a user’s phone; once the notification is clicked on, the user
is presented with a brief survey, containing questions about
the user’s current engagement in a task. The questions in-
clude:
Is your current activity interesting?
Is your current activity challenging?
Figure 1: Screenshot of the ESM
Android application. The survey
was presented once a user clicked
on a notification. Each notification
was announced by the user’s
default notification ringtone and
volume.
How well are you concentrated on it?
Is the activity important for you?
How skilled are you at it?
Each question is answered on a four-point Likert scale, and
possible answers include not at all,a little,somewhat, and
very much.
In addition, a question Is this a good moment to interrupt?,
answered on the same four-point Likert scale, is presented
to a user in order to gauge her sentiment towards the inter-
ruption moment.
The data collection was carried out for two weeks among
20 adult subjects, who received a modest monetary reward
for their participation, where “participation" was defined as
having the application running on their phones, and no re-
quirements were made on actually reacting to notifications.
We recruited subjects from both sexes and from the age
span 20 to 37 years old. Eight notifications were sent to
each participant per day at random times between 8 am
and 10 pm local time. In total 2334 notifications with sur-
veys were sent, out of which 859 were fully answered, and
our analysis concentrates on those answered messages.
Experimental Results
The goal of our study is not to build a fully contained predic-
tive model of a mobile user’s interruptibility and base it on
the information about the user’s current engagement in a
task. Indeed, relying on the explicit user-provided informa-
tion about task engagement, such a model would be highly
impractical, while background sensing of this information
may not be feasible with the current smartphone hardware.
Instead, aim to investigate whether task engagement plays
a role in mobile interruptibility, and if so, in which way differ-
ent aspects of task engagement determine interruptibility.
To quantify the relationship between the five aspects of task
engagement (here labelled as Interesting,Challenging,
Concentrated,Important and Skilled) and the sentiment
that the user expressed towards the moment of interruption,
we fit a linear regression model with the sentiment as a de-
pendent variable, and the five task engagement aspects as
independent variables.
Table 1: Dep = Sentiment
Variable Pearson Coefficient t
(Std. Err.) (Sig.)
Interesting -.03 -.82
(.04) (.410)
Challenging -.12 -3.87∗∗
(.03) (.000)
Concentrated -.17 -3.67∗∗
(.05) (.000)
Important -.05 -1.36
(.04) (.173)
Skilled .17 4.33∗∗
(.04) (.000)
(Regression Constant) 1.40 15.21∗∗
(.09) (.000)
N 859
R20.10
F(5,854) 18.76 (p=.000)∗∗
The parameters of the fitted linear regression are shown
in Table 1. The regression statistically significantly explains
the sentiment towards interruptibility (F(5,854) = 18.76, p =
.000). However, the proportion of variation in the sentiment
that can be explained by the engagement variables is not
high (R2= 0.10), indicating that other factors that we have
not considered in this analysis may influence the interrupt-
ibility.
Since all the survey questions are answered on the same
Likert scale, we encode all the answers with the follow-
ing values not at all=0,a little=1,somewhat=2, and very
much=3. A positive correlation coefficient next to an inde-
pendent variable indicates that a higher rating of this vari-
able (i.e. from a little to very much) leads to a higher rating
of the moment as a good moment to interrupt. In Table 1
only Skilled positively correlates with the sentiment. Thus,
if a person is highly skilled in their current task, they are
less likely to be irritated by an incoming notification. The
other variables are negatively correlated with the sentiment.
However, of those, only two – Challenging and Concen-
trated – are correlated at the significance level of 1% or
less (denoted by **).
Discussion
Despite the change in the way we interruptions are com-
municated and negotiated with the user, the results from
Table 1 are in accordance with the standard theoretical ap-
proach explaining interruptibility in a static context. For ex-
ample, a user skilled in a task is less likely to be disturbed
by an incoming mobile notification. A likely explanation is
that such a user needs less time to reconstruct the primary
task state after the interruption, due to stronger activation
links in the declarative memory. On the other hand, the data
shows that a user who is highly concentrated on a challeng-
ing task is more vulnerable to disruptions.
It is important to note that our metric of interruptibility is
subjective – it is up to our users to declare their sentiment
towards interruptibility. Since the study was done in an un-
controlled real-world setting, there was no way for us to
measure the actual impact of an interruption on a primary
task. It is also interesting that despite the possibility of de-
ferring their actions for a more appropriate moment, users
still feel annoyance when a notification arrives at times of
high engagement. We suspect that the smartphone culture
and social expectations around mobile usage require a user
to keep information about a lingering notification in mind,
even if she postpones her action and continues working on
the original task.
Conclusion
Brain processes are in the foundations of models explain-
ing human interruptibility. With the advent of personalised
sensing devices, the opportunity opened to study interrupt-
ibility by sensing only the external context. Recent progress
in that direction calls for re-examination of the need for in-
ferring one’s cognitive load in order to describe their in-
terruptibility. Our study shows that although notifications
allow users to defer interruptions for a later moment, the
engagement with the current task still plays a significant
role in determining users’ interruptibility. For future work we
leave automated recognition of task engagement via mobile
sensors. In addition, we believe that understanding cogni-
tive processes involved in multitasking can lead to further
improvement of context-aware task management systems.
For example, stimulating primary task recollection from the
declarative memory via context-relevant links.
Acknowledgements
The analysis in this paper relies on the data collected as
a part of the UK EPSRC-funded project “UBhave: ubiqui-
tous and social computing for positive behaviour change"
(EP/I032673/1).
REFERENCES
1. James Fogarty, Scott E. Hudson, Christopher G.
Atkeson, Daniel Avrahami, Jodi Forlizzi, Sara Kiesler,
Johnny C. Lee, and Jie Yang. 2005. Predicting Human
Interruptibility with Sensors. ACM Transactions on
Computer-Human Interaction 12, 1 (March 2005),
119–146.
2. Joyce Ho and Stephen S. Intille. 2005. Using
Context-Aware Computing to Reduce the Perceived
Burden of Interruptions from Mobile Devices. In CHI’05.
Portland, OR, USA.
3. Eric Horvitz and Johnson Apacible. 2003. Learning and
Reasoning about Interruption. In ICMI’03. Vancouver,
Canada.
4. Eric Horvitz, Johnson Apacible, and Muru Subramani.
2005. Balancing awareness and interruption:
Investigation of notification deferral policies. In User
Modeling 2005. Springer, 433–437.
5. Shamsi T Iqbal and Brian P Bailey. 2006. Leveraging
characteristics of task structure to predict the cost of
interruption. In CHI’06. Montreal, Canada.
6. Veljko Pejovic and Mirco Musolesi. 2014. InterruptMe:
designing intelligent prompting mechanisms for
pervasive applications. In UbiComp’14. Seattle, WA,
USA.
7. Martin Pielot, Rodrigo de Oliveira, Haewoon Kwak, and
Nuria Oliver. 2014. Didn’t You See My Message?
Predicting Attentiveness to Mobile Instant Messages. In
CHI’14. Toronto, ON, Canada.
8. Dario D Salvucci and Niels A Taatgen. 2010. The
multitasking mind. Oxford University Press.
9. G. H. Ter Hofte. 2007. Xensible interruptions from your
mobile phone. In Mobile HCI’07. Singapore.
... To mitigate the disruption caused by the interruptions, deferring them to opportune moments, such as task breakpoints [34,80] or rough boundaries during task execution [4] could largely mitigate the cost of interruption. In addition to fnding factors which contribute to the impact of interruption, many recent studies have sought to predict opportune and interruptible moments for receiving notifcations that leverage features including sensor data [39,76] and contextual information such as the current activity [65,95] and engagement [20,64] of users. As VR is not yet prevalent in people's daily lives, only a few attempts to investigate the sending of real-life notifcations in VR have been made (e.g. ...
... However, not all interruptions are equally unacceptable. Studies have shown that many factors, such as the type of primary task, the level of task engagement, and the timing of an interruption, could afect the disruptiveness of interruptions [34,56,64]. Various other studies have been conducted to reduce the impacts of interruptions; for example, scheduling interruptions at task breakpoints or transitions has been found to mitigate interruption cost [1,4,34,80]. ...
... As a result, given that the engagement label data were obtained from the participants, later, in training the model, we used summarized engagement information (i.e., considering the majority of the engagement annotation for every moment; for example, if 16 participants considered the ffth second of a roller coaster 360° video as high engagement and four participants considered it as low engagement, we considered that moment high engagement). Given that prior research has found that interruptibility is related to contextual information such as the current activity [51,65,80,90,95,100] or engagement [20,43,57,63,64,100] of users, we assumed that these features could improve the prediction. ...
... That is, prior research attempts have sought to identify moments at which users are receptive for responding to messages [26], reading news articles [42], reading advertisements [55], responding to interventions [25,44,47], responding to questionnaires [24,41], and so on. By and large, such moments have been found to relate to users' prevailing contexts, including their personal condition (e.g., emotional state [24]), the characteristics of whatever activity they are involved in [24,28], their level of engagement with that activity [28,39], and their social context [17]. Others suggested that receptivity is also dependent upon notifcation content [10,16]. ...
... Such interruptions can arrive at times users feel are inappropriate, and thus have a range of negative impacts [15]. The negative impact of interruptions has been revealed in many studies [3,24,26,28,29,31,39,50], and users have used strategies to control phone alerts to manage whether they want to be aware of such interruptions (e.g. [6].) ...
... Lee et al. [26] found that activity engagement was negatively correlated with receptivity to messaging notifcations. Similarly, Pejovic et al.'s [39] found that task engagement played a signifcant role in determining interruptibility. And Kushlev et al.'s [24] analysis of whether people's feelings afected their choice of what kinds of content to engage with suggested that they are less attentive to distractions when they feel energetic. ...
... Nudges delivered at the wrong time can lead to decreased satisfaction, negative emotions, hyperactivity, and distraction (Adamczyk and Bailey 2004;Kushlev et al. 2016;Mark et al. 2008). Researchers have identified the opportune moments to deliver digital nudges in the form of notifications that can be non-interruptive (Mehrotra et al. 2016;Okoshi et al. 2015;Pejovic et al. 2015). These findings indicate that identifying an adequate nudging moment is an essential aspect of the digital nudge design. ...
Conference Paper
Full-text available
In 2020, over a billion people spent at least three hours a day on social media, primarily engaging in what is described as mindlessly scrolling through their newsfeed. This illustrates the growing societal concern of digital wellbeing and social media addiction. Reducing the time spent on these platforms is challenging since they are designed to be addictive. This paper presents the design and evaluation of a digital nudging intervention that unhooks users from their mindless social media use by making them more mindful. We evaluated the intervention through a two-week single case experimental design (N =20). The findings show that weekly digital consumption was significantly reduced by over 20.58%. The evaluation of the intervention's usability and potential revealed that the intervention made participants mindful of their digital behavior and scored high on usability. Our findings advance how ethical nudges could be self-designed, considering privacy to mitigate social media addiction.
... Kushlev et al. [13] reported that external interruptions from notifications could increase the signs and symptoms of inattention and hyperactivity. Other negative effects of being overwhelmed by notifications include decreased productivity and slower and more error-prone performance [11,25,31]. ...
... As communication activity increasingly moves to mobile and ubiquitous platforms [21,41], measuring individuals' receptivity across constantly changing contexts is becoming more challenging. Prior research used various contextual factors in its attempts to estimate and predict various notions of receptivity, including attentiveness [22,57], responsiveness [38,43], interruptibility [48,50,52,53,55,73,77], and opportune moment [25,31,37,39,56,59,78,79]. For example, Dingler and Pielot [22] predicted mobile users' attentiveness to incoming messages using logs of their phone usage, and achieved an accuracy rate of close to 80%. Lee et al. [43] likewise predicted users' responsiveness to their IM contacts based on IM chat logs, with up to 71% accuracy (AUROC). Pielot et al. [58], on the other hand, predicted mobile users' attentiveness to notifcations via their interactions with the notifcation center, their screen activity and ringer mode, and other phone-sensor data, again with an accuracy of slightly over 70%. ...
Article
Notifications are an indispensable feature of mobile devices, but their delivery can interrupt and distract users. Prior work has examined interventions, such as deferring notification delivery to opportune moments, but has not systematically studied how users might prefer an intelligent system to manage their notifications. Hence, we directly probed Android smartphone users’ notification preferences via a one-week experience-sampling study ( N = 35). We found that users prefer mitigating undesired interruptions by suppressing alerts over deferring them and referred to notification content factors more frequently than contextual factors for explaining their preferences. Then we demonstrated the challenges and potentials of leveraging user actions to help predict notification preferences. Specifically, we showed that a model personalized using user actions achieved a performance gain of 39% than a generic model. This improvement is similar to the 42% performance gain using labels solicited from the user while using observable user actions causes no extra disruption.
Article
Voice assistants, such as Amazon's Alexa and Google Home, increasingly find their way into consumer homes. Their functionality, however, is currently limited to being passive answer machines rather than proactively engaging users in conversations. Speakers' proactivity would open up a range of important application scenarios, including health services, such as checking in on patient states and triggering medication reminders. It remains unclear how passive speakers should implement proactivity. To better understand user perceptions, we ran a 3-week field study with 13 participants where we modified the off-the-shelf Google Home to become proactive. During the study, our speaker proactively triggered conversations that were essentially Experience Sampling probes allowing us to identify when to engage users. Applying machine-learning, we are able to predict user responsiveness with a 71.6% accuracy and find predictive features. We also identify self-reported factors, such as boredom and mood, that are significantly correlated with users' perceived availability. Our prototype and findings inform the design of proactive speakers that verbally engage users at opportune moments and contribute to the design of proactive application scenarios and voice-based experience sampling studies.
Chapter
Mobile ecological momentary assessments (mEMAs) require substantial user efforts to complete, resulting in low user compliance. One major source of incompliance is triggering mEMAs at inopportune moments. In this work, we propose a framework for implementing adaptive mEMAs using reinforcement learning (RL) to address the timing and context challenge, aiming to improve long term response compliance. To effectively model user state, we also propose a two-level user model with both momentary and routine state features. A novel k-routine mining algorithm is developed to extract routine state from passive sensing data. Using real mobile sensing data collected from 220 participants for over two weeks, we show that our proposed RL strategies consistently outperform the baseline methods including a random strategy and a supervised strategy in user compliance.
Article
Full-text available
In the last ten years, it was observed that the continuous upgrades in mobile device’s technology have increased and demonstrated their great potential in various learning environments. Besides, it has motivated researchers to apply more innovative computational techniques regarding context-aware and a set of variables that have been used at different virtual learning proposals. This paper presents a Systematic Mapping Review that focuses on context-aware analysis and its approach to learning processes in mobile learning (m-learning) and ubiquitous learning (u-learning). Furthermore, the study identifies variables that have been used in the past decade for context-awareness analytics and that have been applied to those learning processes. Especially at systems’ adaptations to learning styles and student-specific characteristics. A methodological process was applied to address problems through identification, critical evaluation, and integration of the most relevant works, where high-quality individual studies address one or several research questions. Results show that external variables, including location, time, and software, are the most used variables in the context-aware analysis, with 52.25% prevalence in research papers. Internal variables, including personal information, learning styles, and teaching styles, had a prevalence of 33.33%; variables less frequently used in the research included socioeconomic information and emotional content. The remaining 14.41% represents academic activities. These findings provide a basis for future research and development in the field of m-learning.
Conference Paper
Full-text available
Mobile instant messaging, e.g., via SMS or WhatsApp, often goes along with an expectation of high attentiveness, i.e., that the receiver will notice and read the message within a few minutes. Hence, existing instant messaging services for mobile phones share indicators of availability, such as the last time the user has been online. However, in this paper we not only provide evidence that these cues create social pressure, but that they are also weak predictors of attentiveness. As remedy, we propose to share a machine-computed prediction of whether the user will view a message within the next few minutes or not. For two weeks, we collected behavioral data from 24 users of mobile instant messaging services. By the means of machine-learning techniques, we identified that simple features extracted from the phone, such as the user's interaction with the notification center, the screen activity, the proximity sensor, and the ringer mode, are strong predictors of how quickly the user will attend to the messages. With seven automatically selected features our model predicts whether a phone user will view a message within a few minutes with 70.6% accuracy and a precision for fast attendance of 81.2%.
Conference Paper
Full-text available
We review experiments with bounded deferral, a method aimed at reducing the disruptiveness of incoming messages and alerts in return for bounded delays in receiving information. Bounded deferral provides users with a means for balancing awareness about potentially urgent information with the cost of interruption.
Conference Paper
Full-text available
We present methods for inferring the cost of interrupting users based on multiple streams of events including information generated by interactions with computing devices, visual and acoustical analyses, and data drawn from online calendars. Following a review of prior work on techniques for deliberating about the cost of interruption associated with notifications, we introduce methods for learning models from data that can be used to compute the expected cost of interruption for a user. We describe the Interruption Workbench, a set of event-capture and modeling tools. Finally, we review experiments that characterize the accuracy of the models for predicting interruption cost and discuss research directions.
Article
Full-text available
A person seeking another person's attention is normally able to quickly assess how interruptible the other person currently is. Such assessments allow behavior that we consider natural, socially appropriate, or simply polite. This is in sharp contrast to current computer and communication systems, which are largely unaware of the social situations surrounding their usage and the impact that their actions have on these situations. If systems could model human interruptibility, they could use this information to negotiate interruptions at appropriate times, thus improving human computer interaction.This article presents a series of studies that quantitatively demonstrate that simple sensors can support the construction of models that estimate human interruptibility as well as people do. These models can be constructed without using complex sensors, such as vision-based techniques, and therefore their use in everyday office environments is both practical and affordable. Although currently based on a demographically limited sample, our results indicate a substantial opportunity for future research to validate these results over larger groups of office workers. Our results also motivate the development of systems that use these models to negotiate interruptions at socially appropriate times.
Article
Full-text available
A challenge in building interruption reasoning systems is to compute an accurate cost of interruption (COI). Prior work has used non task-specific cues to predict COI, but ignore characteristics related to the /structure/ of a task. This work investigates how well characteristics of task structure can predict COI, as objectively measured by resumption lag. In an experiment, users were interrupted during task execution at various subtask boundaries to collect a large sample of resumption lag values. Statistical methods were employed to create a parsimonious model that uses characteristics of task structure to predict COI. A subsequent experiment with different tasks showed that the model can predict COI with reasonably high accuracy. Our model can be expediently applied to many goal-directed tasks, allowing systems to make more effective decisions about when to interrupt.
Article
The mobile phone represents a unique platform for interactive applications that can harness the opportunity of an immediate contact with a user in order to increase the impact of the delivered information. However, this accessibility does not necessarily translate to reachability, as recipients might refuse an initiated contact or disfavor a message that comes in an inappropriate moment. In this paper we seek to answer whether, and how, suitable moments for interruption can be identified and utilized in a mobile system. We gather and analyze a real-world smartphone data trace and show that users' broader context, including their activity, location, time of day, emotions and engagement, determine different aspects of interruptibility. We then design and implement InterruptMe, an interruption management library for Android smartphones. An extensive experiment shows that, compared to a context-unaware approach, interruptions elicited through our library result in increased user satisfaction and shorter response times.
Conference Paper
The potential for sensor-enabled mobile devices to proactively present information when and where users need it ranks among the greatest promises of ubiquitous computing. Unfortunately, mobile phones, PDAs, and other computing devices that compete for the user's attention can contribute to interruption irritability and feelings of information overload. Designers of mobile computing interfaces, therefore, require strategies for minimizing the perceived interruption burden of proactively delivered messages. In this work, a context-aware mobile computing device was developed that automatically detects postural and ambulatory activity transitions in real time using wireless accelerometers. This device was used to experimentally measure the receptivity to interruptions delivered at activity transitions relative to those delivered at random times. Messages delivered at activity transitions were found to be better received, thereby suggesting a viable strategy for context-aware message delivery in sensor-enabled mobile computing devices.
Predicting Human Interruptibility with Sensors
  • Daniel Atkeson
  • Jodi Avrahami
  • Sara Forlizzi
  • Johnny C Kiesler
  • Jie Lee
  • Yang
Atkeson, Daniel Avrahami, Jodi Forlizzi, Sara Kiesler, Johnny C. Lee, and Jie Yang. 2005. Predicting Human Interruptibility with Sensors. ACM Transactions on Computer-Human Interaction 12, 1 (March 2005), 119–146.