Int.J.Human-Computer Studies 60 (2004) 771–797
Human–computer interaction issues for mobile
computing in a variable work context
NETCONN Solutions,Hagerstown,MD 21740,USA
School of Business Administration,Penn State Harrisburg,777 W.Harrisburg Pike,
Received 25 January 2003;accepted 19 July 2003
The current paper takes an introspective look at the human–computer interaction (HCI)
issues for mobile computing in a variable work context.We catalogue the current research in
four major categories.The major ﬁndings of our study are following.(1) A majority of HCI
issues,about 58%,fall under the category of computer systems and interface architecture
implications.(2) 23%of the articles focus on development and implementation issues.(3) 13%
of the articles focus on use and context of computer issues.(4) 6% of the articles focus on
human characteristics issues.Further,the literature indicates that the ﬁeld services is a main
application of mobile computing (46%) followed by sales force (21%),health care (17%),
ﬁeldwork (8%),insurance claims (4%) and journalism (4%).
r 2004 Elsevier Ltd.All rights reserved.
Researchers predict that mobile computing will be a signiﬁcant area of IT and
HCI research,development,and application (Hewett et al.,1996;Chandra et al.,
2000).Conservative estimates based on the 2000 Census report suggest that by 2006
10% of US workers will be completely mobile,with no permanent ofﬁce location.
This is estimated to be approximately 14 million workers who will work in the ﬁeld,
spending time with clients or providing remote services (Lucas,2001).Other
estimates put the number at about 70 million US workers who will perform all or
part of their jobs from remote locations in 2003 (Goth,1999).This trend has been
ARTICLE IN PRESS
E-mail address:firstname.lastname@example.org (P.C.Pendharkar).
1071-5819/$ - see front matter r2004 Elsevier Ltd.All rights reserved.
brought on partially by the maturation of wireless communications,which is
providing new possibilities for mobile work place scenarios.Not so long ago,
wireless communications were very limited with regards to functionality of devices
and speed of communications.Constraints such as screen size,memory,and storage
capabilities as well as data transfer rates averaging 14.4 KBPS limited the amount of
data that could be both displayed and accessed (Turisco,2000).With recent
advancements in wireless data transfer,such as GPRS and 3G services,data
transmission speeds mirror and will eventually exceed that of current landline dial-up
(Brodie and Perry,2001).However,some researchers believe that wireless network
capabilities will continue to be less advanced as the current state of the wired world
(Capra,2002).But certainly as networks increase their speed 10-fold,this allows
more opportunity and feasibility for mobile computing applications.Similarly,as the
costs of handheld computing devices fall,their use becomes more of a consideration
for the enterprise (Porn and Patrick,2002).As advances in technology increase
coverage,data speeds,and usability,greater user acceptance will fuel development of
new applications that are more feasible in the maturing mobile computing
environment (Crowley et al.,2000;Phamet al.,2000;Turisco,2000;Barbash,2001).
There are many issues that have slowed the adoption of mobile computing.One
reason is that mobile technology has lagged signiﬁcantly behind desktop computing
in terms of processing power,power availability,and network speed.Additionally,
the rapidly changing pace of mobile computing has reduced the ability for business
professionals to keep up with,understand,and utilize the available technology.
Another concern is whether or not future technology will meet customer demands
(Kentrick,2002).Mobile computing can,however,lead to a competitive advantage
in some industries (Varshney,1999).It can help improve workﬂow and efﬁciency as
well as reduce costs and risk management (Miah and Bashir,1997;Turisco,2000;
Porn and Patrick,2002).This is true,only if the mobile computing architecture
chosen compliments the work scenario and is understood by management so that it
can be fully exploited.
The objective of this paper is to provide a survey of literature regarding computing
in a mobile environment.We focus on the human–computer interaction (HCI)
considerations needed in mobile computing as well as possible design and
implementation solutions.These HCI issues include use and context of computers,
human characteristics,computer systems and interface architecture,and the
development process.By understanding the nature of the circumstances and
problems mobile workers face and the possible resources available to resolve their
issues,designers and developers are better able to respond to their needs
(Kristoffersen and Ljungberg,1999;Perry et al.,2001).Our research will be of
interest to business managers considering integrating computing with their mobile
workforce as well as developers and designers of future mobile computing products.
The rest of the paper is organized as follows.In Section 2,we summarize the
current literature on HCI issues for mobile computing in a variable work context.In
Section 3,we catalogue the current literature under different categories.In Section 4,
we summarize our ﬁndings,draw conclusions and provide directions for future
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797772
Mobile computing has generally been built on the concept of being able to connect
anytime and anywhere (Kalakota et al.,1996).Early research in wireless data access
and mobile computing has focused on the professional mobile worker who is
itinerant for a partial workweek and needs access to desktop applications and
services (Jones and Brown,2000;Pascoe et al.,2000;Brodie and Perry,2001;Perry
et al.,2001).Wireless communication tools allow these workers to stay in touch with
the ofﬁce,when remote,and access a subset of data from a portable device (Pascoe
et al.,2000).Concerns of the business traveler include planning for the
unpredictability of travel,effective use of down-time,communication and access to
ofﬁce information and colleagues (Perry et al.,2001).Pascoe et al.(2000),in their
research on ﬁeldwork,deﬁne this type of usage as ‘‘portable computing’’ as the user’s
intention is to access computing resources from a remote,but static environment
(e.g.during an off-site meeting,at an airport,or traveling on a train).In this
environment,the portable devices and applications are typically miniaturized
replications of the desktop environment (Bertelsen and Nielsen,2000;Turisco,2000).
According to Pascoe et al.(2000),true mobile computing occurs in a variable work
context and carries its own set of issues and technical implications.Mobile
computing,as Pascoe et al.(2000) deﬁne it,involves using while moving—not only is
the user and device not static,but the activity of the user involves a high degree of
mobility.This is the deﬁnition that will be used throughout this paper.In mobile
computing scenarios,it is the environment and the amount of mobility the user
requires that will inﬂuence the design of the appropriate architecture (Kristoffersen
and Ljungberg,1999;Pascoe et al.,2000;Perry et al.,2001).Special considerations
for this group of workers include dynamic user conﬁguration,limited attention
capacity,high-speed interaction,and context dependency (Kristoffersen and
Ljungberg,1999;Pascoe et al.,2000).It is important to note that this list is not
all-inclusive.The nature of workers using mobile computing,rather than those using
portable computing,is that they are a more diversiﬁed group.While users of
portable computing are typically professional white-collar workers using wireless
communication to remotely access ofﬁce documents,e-mail,and corporate knowl-
edge applications,users of mobile computing range from health care providers to
insurance claims adjusters to blue-collar service workers.For the most part,they are
accessing unique applications that speciﬁcally support their work functions.Some
examples of these applications include knowledge sharing,data collection,
equipment maintenance,inspection,and work order access.Therefore,each worker
group may have its own unique considerations depending on the environment they
work in and the task at hand.These considerations all have HCI implications for
mobile computing architecture that can affect design,worker productivity,and user
acceptance.These technical issues are above and beyond,but sometimes linked to,
the general wireless data concerns such as database access,security,and network
Weiser (1991) ﬁrst introduced the term ‘‘ubiquitous computing,’’ or ubicomp,in
1991 as a way to explain ‘‘machines that ﬁt the human environment instead of
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 773
forcing humans to enter theirs’’ (Weiser,1991).This concept has been applied to the
area of mobile computing in many studies and the understanding of it is very
appropriate to the purposes of the research presented in this paper (Hirakawa et al.,
1998;Crowley et al.,2000;Hinckley et al.,2000;Pham et al.,2000;Branco,2001).
Central to the concept of ubicomp is that technologies should disappear into the
background so that users can unconsciously apply them to the task at hand (Weiser,
1991;Abowd and Mynatt,2000).This is extremely relevant to mobile workers whose
use of computing power is generally secondary to the task at hand (e.g.an emergency
ﬁeld health care worker who needs to attend to a patient while collecting data for the
admitting hospital).The goals of ubicomp can be summarized as follows:(1) every
day human tasks must be understood and supported by an appropriate interaction
experience,(2) heterogeneous solutions should be available to offer differing forms
of interactive experience as situations warrant,and (3) these solutions,when
networked,should provide a holistic user experience (Abowd et al.,2002).Ubicomp
solutions typically include input and output (I/O) functionality that mirrors the way
humans communicate such as speaking,gesturing,and writing.This form of
communication has also been called multimodal interaction (Blattner,1994;Branco,
2001).Additionally,ubicomp solutions typically involve using context awareness in
a device as both an input function and to control output modes.Finally,device
ergonomics considerations are key in fostering the natural ‘‘ﬁt’’ of machines in the
human interaction.Invisible interaction with machines based on their contextual
perception and natural human communication forms is most likely to be adopted
rapidly in areas where traditional GUI interfaces are not appropriate (Crowley et al.,
2000).Mobile computing in a variable work context is such an area.As will be
discussed later in this paper,frequently a traditional GUI interface demands too
many visual and cognitive resources to be practical for mobile computing since
human interaction with the machine is generally secondary to the task at hand.
Because of this phenomenon,it is important that we consider ubicomp solutions for
mobile computing in a variable work context.
Misunderstandings in the nature of mobile work can lead to problems in
implementation and technology misuse.This can cause a number of issues,the most
critical being loss of production,miscommunication,and increased costs (Engle and
Barnes,2000;Watad and Disanzo,2000;Erffmeyer and Johnson,2001;Perry et al.,
2001;Speier and Venkatesh,2002).While enhancements in wireless communication
technology have occurred recently,there are still factors limiting the broad appeal of
wireless computing.The biggest barriers are ergonomics and usability as well as the
lack of practical,personal,and timely interaction tools (Barbash,2001;Spriesters-
bach et al.,2001).Subsequently,workﬂows for mobile workers today are largely
paper-based and lack automation and back-end integration (Spriestersbach et al.,
2001).It is in this arena,as well as in knowledge management and communications,
that mobile computing can most readily serve the enterprise.
Most studies about computing for mobile workers in a variable work context
centers around HCI issues—how the ﬁeld worker interacts with the mobile
computing device.There is a drive to move away from the concept of a PDA as a
miniature workstation,which has been driven by the portable computing business
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797774
traveler segment.Studies have found that concepts in desktop computing are not
scaleable and do not ﬁt the mobile computing environment (Bertelsen and Nielsen,
2000).The new direction is a mobile computing interaction paradigmthat takes into
account the diverse environments and tasks mobile users face.The ﬁeld of HCI is
concerned with the ‘‘design,evaluation and implementation of interactive computing
systems for human use’’ (Hewett et al.,1996).Studies in HCI consider four areas of
research:use and context of computers,human characteristics,computer systems
and interface architecture,and the development process (Hewett et al.,1996).This
paper will focus on these four areas as current literature in the area of mobile
computing in a variable work context is reviewed.
2.1.Use and context of computers
The use and context for mobile computing is a combination of the task at hand
and the environment it will be performed in.This is very different from the HCI in
the remote ofﬁce environment found in portable computing applications.
Kristoffersen and Ljungberg (1999) identify four key elements that deﬁne mobile
work contexts and explain how they differ from the ofﬁce setting:
Tasks external to operating the mobile computer are the most important,as
opposed to tasks taking place ‘‘in the computer’’ (e.g.a spreadsheet for an ofﬁce
Users’ hands are often used to manipulate physical objects,as opposed to users in
the traditional ofﬁce setting,whose hands are safely and ergonomically placed on
Users may be involved in tasks (‘‘outside the computer’’) that demand a high level
of visual attention (to avoid danger as well as monitor progress),as opposed to
the traditional ofﬁce setting where a large degree of visual attention is usually
directed at the computer.
Users may be highly mobile during the task,as opposed to in the ofﬁce,where
doing and typing are often separated (Kristoffersen and Ljungberg,1999,p.276).
As described above,workers could face a heterogeneous environment within one
mobile session or task.For instance,mobility takes place within the context of time
and space (Wiberg and Ljungberg,1999).Context changes can vary in frequency,
speed,and predictability (Jones and Brown,2000).Therefore,the time in which a
task takes place can be just as critical as the location in some mobile applications and
could have implications for the ﬁt of a computing device (Baber et al.,1999).Typical
mobile computing tasks may include data collection and retrieval,knowledge
management,collaborative communication,and order management.This diversity
inherent in the mobile computing environment makes it difﬁcult to make
generalizations across work groups and functions (Perry et al.,2001).Not only
can each application of mobile computing dictate special considerations but,as users
change work contexts while using an application,so may their needs change.
Stressors in the worker’s environments can also affect the HCI by altering the
capacities and capabilities of the user (Baber et al.,1999;Hinckley et al.,2000).For
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 775
instance,high,variable noise levels can affect speech recognition performance (Baber
and Noyes,1996).Additionally,this and other environmental distractions can affect
the user’s cognitive ability (Baber and Noyes,1996).Variation in illumination and
glare can affect the readability of displays (Esposito,1997;Hinckley et al.,2000).
And,inclement weather conditions can affect the performance of both the user and
the device (Baber,1997).A rough ride in a vehicle can affect pen-based and touch-
screen input abilities (Holtzman,1999).Similarly,the tasks that a user is engaged in
can affect the user’s interaction with the computing device.For instance,a task can
cause the user to be physically constrained (e.g.a paramedic reaching through a car
door to aid a patient,a telecom ﬁeld service engineer on a telephone pole) or
challenged (e.g.a ﬁreﬁghter running through a burning building) which can impair
movement,cognitive reasoning,and the user’s ability to interact with the system
(Baber et al.,1999;Holtzman,1999).Therefore,it is important to design variable
I/O methods that can be conﬁgured by the user as situations and contexts change
(Holtzman,1999).A more automated solution would involve the use of context
sensors that could modify interaction based on the user’s environment (Hinckley
Collaborative work is another use context where HCI issues have to be addressed.
These environments require that participants interact with the shared data,so
adaptation of media to certain participants may hinder other participants’
comprehension and the efﬁciency of the communication (Branco,2001).Video
conferencing has been explored as a communications tool (Sprey,1996).In a mobile
environment,it is more difﬁcult to get quality images since computing power and
bandwidths are typically low and users must account for the unique errors in wireless
networks (Bassil,1996).Nevertheless,image capture and voice transmission have
proved to be key in perpetuating a collaborative experience (Smailagic,1999).Video
conferencing serves as a collaborative tool as ﬁeld workers can use wearable
computers to capture images and collaborate with a remote expert on repair steps
(Kraut et al.,1996).Another technique is to provide a written drawing or
annotations,rather than verbal output to the end-user,on the wearable visual
display (Cheng and Robinson,1998).
A concern in mobile computing studies is the amount of additional work that a
user has to engage in to accommodate IT resources to ﬁt the work context
(Kristoffersen and Ljungberg,1999).More accommodation makes it harder to focus
on the job at hand.Workers lose time and productivity when they have to
reconﬁgure their work setting and processes to allow mobile technology to support
their work.Human characteristic considerations include information-processing
abilities,interaction dynamics,communication,and physical and physiological
issues.Studies have focused on the direct manipulation interactive style (also know
as Windows,Icons,Menus,and Pointers or WIMP) found in traditional computing
devices and applications (Kristoffersen and Ljungberg,1999;Crowley et al.,2000).
While WIMP works well in a traditional desktop-computing environment,it is not
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797776
usually practical for mobile workers.Often,mobile workers need to focus their
visual attention on the task at hand or on safety issues and I/O operations are
secondary to the task at hand (Kristoffersen and Ljungberg,1999;Pascoe et al.,
2000;Pirhonen et al.,2002).Studies on the direct manipulation paradigm suggest
that it is exclusive (i.e.workers can only focus their visual attention on one object at
a time) (Kristoffersen and Ljungberg,1999).For instance,a driver cannot read a
map while maintaining safe driving practices.One option is to accommodate the
technology by modifying the task (e.g.the driver pulls off the road to read the map).
This however,can lead to worker frustration and loss of productivity (Kristoffersen
and Ljungberg,1999;Pascoe et al.,2000).Another,more suitable option is to
employ complementary design principles to address speciﬁc work contexts.This
involves moving out of the WIMP paradigm found in desktop computing to a user-
centric alternative (Baber et al.,1999).For instance,application and hardware can
be designed to shift HCI to other channels or senses appropriate for the task at hand
(Pascoe et al.,2000).A lot of work has focused on the use of touch and voice
recognition as input options,and auditory and tactile feedback as output
Other human characteristic studies have focused on the use of wearable
computers.These typically accept input from voice or gestures and provide audio
and visual output (Baber et al.,1999;Sawhney and Schmandt,1999;Laramee and
Ware,2002).While these can,at times,offer a more natural interaction with the
computing device,the size,weight,and position on the body can alter how the user
interacts with it (Baber et al.,1999).These devices are typically carried on the head,
neck,arm,trunk,or back (Baber et al.,1999;Sawhney and Schmandt,1999;
Laramee and Ware,2002).Depending on the weight,body placement,and activities
involved,the user may incur issues that can range from trivial feelings of
awkwardness to more long term musculoskeletal problems (Baber et al.,1999).
A study of head-mounted displays showed that even a minimal weight could,over
time,impact user posture and performance (Baber et al.,1999).Additionally,the
display caused an increased range of head movements and reduced situational
awareness by competing with environmental visual demands,which created
additional interference (Baber et al.,1999).Tethered leads that connect various
components of the wearable systemmay also interfere with task performance (Baber
et al.,1999).Another study showed that a visual display unit led to faster
performance than the head-mounted display in a particular task as it took more time
and effort to focus on the head-mounted display unit (Baber et al.,1999).
Additionally,the study showed that auditory output provided the fastest
performance time,but with very high misinformation rate,and a graphic display
was the next fastest,followed by text (Baber et al.,1999).
Head-mounted displays are often used with visual I/O using both monocular and
binocular displays (Baber et al.,1999;Laramee and Ware,2002).They provide the
ability to have a high-resolution display without having to carry a large LCDdisplay
or be restricted to smaller PDA screens (Guerlain et al.,1999;Laramee and Ware,
2002).The video display units range from opaque and transparent monocular to
transparent binocular displays (Laramee and Ware,2002).Monocular transparent
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 777
displays are often preferred due to their lower cost,power consumption,and weight.
One eye can view the surrounding environment,while the other eye is ﬁxed on a
virtual image that seems to ﬂoat over the other view.This allows the free eye to
continue viewing the surrounding environment which can be critical to many ﬁeld
work functions (Laramee and Ware,2002).While they do have their beneﬁts,these
wearable displays can cause perception issues,which include binocular rivalry,visual
interference,depth of focus,phoria changes,unnatural eye movements,and eye
dominance (Laramee and Ware,2002).Binocular rivalry occurs when dissimilar
images are presented to each eye.Humans have no conscious control over reaction
to this.Typically,the eyes alter between a ﬁght for dominance and a merging of the
two images into a fragmented mixture (Laramee and Ware,2002).Visual
interference occurs when two images are not clearly distinguishable.This often
occurs when patterns are similar between the two views.Depth of focus issues refer
to interference that can occur when the virtual image on the transparent monocular
is at the same focal point as the surrounding background causing them to blend
(Laramee and Ware,2002).Prolonged use of a monocular can result in changes in
phoria,which is the direction of the gaze of the eye when at rest (Laramee and Ware,
2002).Typically,this can take a few minutes to correct,but can lead to headaches.
Head-mounted displays also do not allow for natural coordinated movement of the
head and eyes when scanning an object.With these displays,only eye movement can
be used for scanning,which can cause eyestrain (Laramee and Ware,2002).These
devices also can affect equilibriumas well as cause disorientation (Baber et al.,1999;
Laramee and Ware,2002).A study on these issues found that binocular rivalry and
visual interference negatively affected task performance.However,movable elements
in the visual display help to increase eye dominance,allowing for less impact on user
perception abilities (Laramee and Ware,2002).
Direct manipulation used mostly in desktop applications does offer beneﬁts that
can be lost when HCI is shifted to other forms of interaction.Without direct
manipulation techniques,workers have to do more memorization.For instance
when using a series of thumb touches to represent various computing functions
(e.g.much like a single and double mouse click perform different operations) there
are inherent limitations.Humans have a hard time remembering past three or four
touches (Pascoe et al.,2000).Therefore interfaces have to be designed to address this
limitation.Similarly,navigation though unstructured or poorly designed screens can
impede a user’s cognitive ability (Miah and Bashir,1997).Designers can use form
ﬂow diagrams to map the interface and better understand the user experience (Miah
For the most part,when a lot of visual attention is needed for a particular task
outside the computing environment,minimal user attention interfaces are used on
the mobile computing device.The screen is typically divided into quadrants that
represent functions (Kristoffersen and Ljungberg,1999;Pascoe et al.,2000).But,for
one-handed operations,which are sometimes required,this becomes a challenge.
Typically,when a mobile computing device,such as a Palm device,is held in one
hand,the thumb can serve as an input mechanism.However,due to the thumb’s
limited physical movement capabilities,it is much easier to stack the touch areas on
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797778
top of each other,rather than align them side-by-side or in a matrix arrangement
(Pascoe et al.,2000).Stacking the touch areas gives the thumb more range of
movement on the screen.
Other human considerations include different attitudes about technology that can
hinder implementation and user acceptance (Engle and Barnes,2000;Watad and
Disanzo,2000;Erffmeyer and Johnson,2001;Speier and Venkatesh,2002).
Additionally,unions can add an extra layer of complexity during implementation
efforts (Guerlain et al.,1999).Worker attitudes and strategies to address them need
to be understood in order to successfully design,develop,and implement mobile
computing in the work environment.Finally,mobile computing can hinder normal
communication patterns.For instance when a mobile computing device was
introduced into a work site to replace a paper-based documentation system,it
actually made it more difﬁcult to access and share information since face-time with
co-workers was reduced (Luff and Heath,1998).
2.3.Computer systems and interface architecture implications
The use context of computers and human characteristics have large implications
for the design and development of computer systems and interface architectures.For
greater user acceptance and to realize the beneﬁts of increased productivity,mobile
computing devices and applications need to be designed in a way so that they meld
with the work ﬂow patterns and do not hinder them.This section will explore
wireless network technologies,hardware,I/O tactics,and knowledge management
solutions that can aid the mobile worker.However,since mobile workers are such a
heterogeneous group,a critical technical consideration is dynamic user conﬁgura-
tion,which will also be explored.This allows users to adapt the hardware and
application to the variable work context so that their interaction with the computing
device can be customized as their context changes.
Mobile computing users link to the network in many different ways including
Wireless WANs and LANs,Wireless Personal Area Networks (PANs),ad hoc
networks,cellular,satellite,and infrared (Wickelgren,1996;Miah and Bashir,1997;
Zimmerman,1999;Fagrell et al.,2000;Varshney and Vetter,2000;Kentrick,2002).
In some cases where costs or coverage are a limiting factor,the handheld stores data
and then synchronizes with a laptop via a serial cable or cradle (Guerlain et al.,1999;
Fagrell et al.,2000;Pascoe et al.,2000;Porn and Patrick,2002).However,most
often it is a wireless link that is used because it allows for faster interaction with the
network and it helps to mitigate the limited computing power of hand held devices
(Guerlain et al.,1999).Through wireless data transfer,the server can store data
instead of the client,which frees up computing resources in the mobile device.
Wireless connections also allow asynchronous real-time integration and update with
the database as well as real-time data mining from multiple sources (Barbash,2001).
Additionally,there is value with the integration of the web at the application level for
knowledge management solutions (Barbash,2001).Wireless data architecture,
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 779
however,still comes with its own set of issues.On the network side these include
limited bandwidth,unreliable connections,questionable security,and a lack of a
single connectivity standard.On the hardware side,these issues include lack of
storage capacity in devices,small screen sizes,and limited memory (Varshney,1999;
Turisco,2000;Barbash,2001).Advances in technology have already addressed some
of these issues,and will continue to do so in the future.
There are no truly ubiquitous,single source wireless services.While users can
typically rely on one wireless networking solution when the level of mobility and
network reliability needed are low,users who need constant coverage over dispersed
areas rely on a variety of communication mediums (e.g.satellite,WWANs,and
WLANs).Therefore,applications must be built so that the communication sub-
systemis isolated fromthe application layer (Miah and Bashir,1997).This will allow
the application to communicate over a wide variety of media and protocols.
Additionally,since connectivity is not completely ubiquitous,many users ﬁnd the
need to keep a copy of vital data on the client device (Mahajan et al.,1998).
However,mobile devices typically have limited storage and processing capacity
(Moray,1996).Therefore,with mobile computing,there has to be a balance between
data stored on the device and data accessed or relayed via networking (Moray,
1996).For instance,when making a mapping functionality available for a group of
mobile utility ﬁeld service engineers,the data storage requirements ranged from 9 to
60 gigabytes of data.Mass storage on the mobile device allowed the workers to be
totally autonomous,so that they had access to the data whenever network services
were unavailable,however,this systemlacks ﬂexibility in terms of updates and speed
of use (Moray,1996).On the other hand,delivery of the data via communications
allows for more ﬂexibility,but suffers from potentially lower availability (Moray,
1996).Additionally,large ﬁles can become costly to transfer (Moray,1996).The data
access strategy most appropriate was found to be a hybrid between both
communication and device storage (Moray,1996).In one approach,high-level
maps were stored on the device,and detailed locations of utility lines and pipes were
wirelessly transferred when needed (Moray,1996).Another approach may be to
store the maps on the device,and update via wireless transfer when necessary
(Moray,1996).This same hybrid strategy can also be used to give ﬁeld workers the
ability to enter data and store it on the mobile device until a network connection is
present (Kalakota et al.,1996).Additionally,the speed of input is greatly enhanced if
the device stores the data and then sends it to a remote host when computing
resources and network bandwidth are optimal.However,this does not allow for real-
time data updates.
Some techniques have been employed to facilitate the mobile worker’s interaction
with the wireless network.For instance,one technique is to conﬁgure the server
engine to detect the connection speed and,based on this and the device capabilities,
vary the amount and type of information sent.This can allow the user to receive less
information if the connection is slow.Additionally,information can be sent in the
optimal conﬁguration to render on the device (Fagrell et al.,2000).Another
technique relies on the use of an ultra thin client.This allows the data storage to be
done at the server level,and frees up computing resources for the mobile worker
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797780
(Kristoffersen and Ljungberg,1999).Thin client computing enables the use of
smaller,light-weight,and lower cost hardware with better power management and
battery life,but is at the mercy of network availability (Guerlain et al.,1999).
Mobile computing hardware must be chosen so that it is suitable for the task at
hand (Pascoe et al.,2000;Porn and Patrick,2002).However,is it important to note
that fully customized hardware can be an expensive option and should only be used
if the beneﬁts are justiﬁed (Branco,2001).When possible,it is more cost-effective to
use existing low-cost hardware that has some user conﬁguration options built in
(Pascoe et al.,2000).While it is hard to generalize across all mobile work situations,
Pascoe et al.(2000) provide a list of hardware criteria for ﬁeldworkers that could be
applied to other industries:(1) Pen user interfaces provide a nature substitute for
paper-based systems.They are also more suitable to the mobile environment than
miniature keyboards.(2) The device should have a small and unencumbering form
factor that does not adversely affect the user’s body or senses.(3) Battery life should
be substantial enough to last for an entire workday without requiring a replacement.
(4) Devices should be robust enough to withstand dropping and environmental
conditions of the workplace.There are options for ruggedized equipment,but it is
often available at a considerable price.(5) The device should support connectivity to
multiple sensor devices,equipment,and networks (Pascoe et al.,2000).Similarly,
Moray (1996) suggests that mobile ﬁeld service workers need devices that can work
in various light levels,can withstand being dropped,are of appropriate size and
weight to be carried,and have sufﬁcient portable power.
Often issues arise when there is no place to put the device.Field workers will often
transfer information to paper,which reduces productivity and interrupts the
workﬂow (Kristoffersen and Ljungberg,1999).Additionally,one-handed operation
is often necessary due to the nature of the task,so hardware needs to be designed
with this in mind (Kristoffersen and Ljungberg,1999;Pascoe et al.,2000).Another
possible solution is to deploy the use of wearable computers (Guerlain et al.,1999;
Kristoffersen and Ljungberg,1999).However,these may sometimes get in the way as
in a head-mounted option that did not appeal to users because it had the potential of
obstructing their vision and added weight to their head (Guerlain et al.,1999).
Additionally,depending on the computing power needed,the mobile device can be
so large that it requires a backpack to carry it around.This is not always practical
and does not usually meet user acceptance and safety standards (Guerlain et al.,
1999).In one example of a wearable computer system,a handheld monocular with
an integrate mouse,called an Eyewand,was deployed.This served as an input (in the
form of mouse navigation) and output device (in the form of an interface visible
through the monocular) that supported a rich 800 600 pixel (SVGA) (Guerlain
et al.,1999).This device allowed users to visualize more complicated screen elements
and navigate via mouse,which worked well for the work context involved.Heads up
displays are another option used in auto-mounted units as well as in aircraft.
However,reports show that they can cause disorientation,trouble with focusing,and
confusion when operating equipment (Laramee and Ware,2002).Some of these
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 781
units displace the line of sight and can cause problems with hand-eye coordination
(Rolland et al.,1995).
Another important consideration when using minimal-attention user interfaces is
the presence of tactile input devices.These include raised push buttons that may be
of different shapes and sizes and as well as linked to distinct sounds when pressed.
For instance,devices with a chording keyboard would allow for more tactile
feedback of input (Pascoe et al.,2000).Push buttons can also be modiﬁed so that
they feel different or require a variable amount of force based on the application
design (Poupyrev et al.,2002).Some designers have considered pads instead of
buttons that have physical scale,texture,shape,and orientation relative to the
magnitude of directional changes to make the buttons easier to recognize and press
without needing full visual attention (Kristoffersen and Ljungberg,1999;Pirhonen
et al.,2002).Other development in this arena has been focused on tilt sensors as well
as force impulse and tactile push-button simulation based on touch-screen input
(Poupyrev et al.,2002).
Hardware can also be developed to allow for input gestures to reﬂect the user’s
natural interaction with the computing device.For instance,as natural input
mechanisms,eye ﬁxation can be used for selection and head movements for scrolling
an application (Crowley et al.,2000).In one study,a device was equipped with an IR
proximity sensor,touch sensitivity on the outer casing,and a tilt sensor (Hinckley
et al.,2000).By combining these tools,the device was able to capture natural,
informative gestures and react accordingly.For instance,picking the unit up
triggered the power switch to turn on.When walking,the unit behaved differently
than if the user were standing still.When the device was held like a cell phone or
microphone,the unit began recording a voice memo.When the unit was tilted froma
vertical to a horizontal position or upside down,the image on the display would
follow accordingly.Scrolling on the display was accomplished by a simple tilt in the
direction the user wanted to view (Hinckley et al.,2000).A series of audio beeps is
used at output so the user can verify operation (Hinckley et al.,2000).These types of
gesture inputs require little visual or cognitive attention.Additionally,studies by
Hinckley et al.(2000) show that sensor fusion (i.e.the combining of multiple context
sensing inputs) is key in expanding the capability of context awareness.Devices can
also employ the use of a light sensor to adjust a display quality without demanding
attention from the user (Hinckley et al.,2000).
Sometimes many devices are used by one individual depending on the task at hand
(Holtzman,1999;Pascoe et al.,2000;Pham et al.,2000;Smailagic,1999).For
instance,in one application for ﬁeld service engineers,devices used by each worker
include a base unit the size of a laptop which is wirelessly connected to a remote
server located in the home ofﬁce,a cellular phone tethered to the base unit,a smaller
removable satellite unit that replicates portions of the base unit display,and a
wearable microphone that links to the cellular phone (Smailagic,1999).This allows
the ﬁeld worker the ﬂexibility of various input methods such as pen-based,voice,and
keyboard (Smailagic,1999).Another example is an application in the ﬁeld of
emergency health care workers,where three different devices were used in the ﬁeld
environment to collect and display data (Holtzman,1999).They were all wirelessly
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797782
networked together to allow for real-time data exchange.The ﬁrst device,the
Personal Data Monitor,was attached to a patient to record vital signs and collected
input via biological monitors (e.g.heart rate,blood pressure,and temperature).The
second device,called the Field Medic Associate,was in the form of a wearable
headset that collected speech input on care done at the initial point of patient
contact.It gave feedback to the user,whose hands and visual attention were involved
with the patient,through speech synthesis and signal tones.Finally,the third device,
called the Field Medic Coordinator,was designed for use after the patient had
received urgent care,or by an on-site supervising medic.This device contained a
graphic display of the input fromthe other two devices,and allowed both speech and
pen-based input.The Field Medic Coordinator also had the ability to use a WAN
connection to relay patient data back to the admitting hospital,so that they were
prepared for the impending patient arrival.As shown in these scenarios,some ﬁeld
work situations employ a multitude of networked devices to capture data as the
2.3.3.Input and output techniques
I/O techniques are critical issues.Not only does the mobile nature of the work
need to be considered,but the size limitations of mobile computing devices are also
an important factor.For instance,text entry speed on mobile computing devices
with keypads ranging from three to over 26 buttons vary from 7 to 10 words-per-
minute (wpm) (Butts and Cockburn,2001;MacKenzie,2002).This speed typically
doubles with the use of linguistically enhanced input techniques such as T9 and
LetterWise which compare text input to a database to anticipate the intended word
(MacKenzie,2002;Silfverberg et al.,2000).Nevertheless,these input rates do not
compare to the average desktop keyboard typing speed for transcription of 33 wpm
(Karat et al.,1999).On the other hand,computer speech,touch-screen keyboard,
and handwriting studies show an average between 15 and 25 wpm,after
error correction,with speech being the highest (Karat et al.,1999;Varile and
Mobile workers are typically both producers and users of corporate data as they
access systems for information,and input observation or customer data from the
ﬁeld (Kalakota et al.,1996).This can take various forms including dispatch services,
work order management,and enterprise communications.Methods to ease the
burden of input and spread the requirements of processing output over all the
human senses while still maintaining data integrity are of importance.One method is
to deploy a minimal attention user interface.This may mean dividing a touch-screen
into three or four large regions to allow for visually simpliﬁed I/O operations
(Kristoffersen and Ljungberg,1999;Pascoe et al.,2000).Other techniques focus on
shifting the HCI to unused channels or senses appropriate for the task at hand
(Pascoe et al.,2000).For instance,ubicomp researchers strive to make I/O
interactions more like natural modes of communication.These solutions tend to lean
towards implicit input functionality when possible to mirror natural interactions
with the environment,focusing on handwriting,speech,and gestures,rather than
push-button input (Abowd et al.,2002;Pirhonen et al.,2002).In this section,we will
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 783
consider various I/O techniques and explore the idea of context awareness with
regards to I/O.
The most widely used input mechanisms on mobile devices beyond push buttons
are touch screens.Pen-base computing is one form of touch-screen input.This type
of computing was originally unsuccessful when it was ﬁrst launched in the early
1990s,but improvements in handwriting recognition have made it a more viable
option (Abowd et al.,2002).However,while pen-based computing works in some
situations when a user has two free hands,it is not always practical.Other options
include serialized thumb taps (similar to single and double mouse clicks) and thumb
stroke gestures in various natural directions to represent a variety of functions
(e.g.stroke right to count an observance in one application or to play audio in
another) (Pascoe et al.,2000;Pirhonen et al.,2002).When using thumb-touch input,
it is important to design the target on the touch-pad so that there is little room for
user error in situations when minimal visual attention is available.Therefore,in
order to ease navigation,in most cases the screen is divided into three or four regions
to represent different functions (Kristoffersen and Ljungberg,1999;Pascoe et al.,
2000).More recent technological studies involve the use of tilting of the mobile unit
as an input mechanism (Hinckley et al.,2000;Poupyrev et al.,2002).
Speech and handwriting recognition are two growing forms of input.The beneﬁts
of speech recognition include minimal user attention input,direct system entry,
remote microphone capabilities,and faster speed of operation compared to other
competing input methods (Noyes,1995;Karat et al.,1999).In one mobile computing
application for emergency medical workers,speech recognition was chosen as an
input option after it was identiﬁed that workers naturally called out their
observations of patients and treatment steps as they were being performed
(Holtzman,1999).Speech recognition,however,is a tricky implementation in the
mobile computing environment.It is only feasible in a low-noise environment or
where noise will not negatively affect the task at hand (e.g.in observational
ﬁeldwork studies) (Guerlain et al.,1999;Pascoe et al.,2000).Additionally,the
processing power needed for speech recognition is not typically found in a compact,
portable device.Therefore,wireless network connections are critical for this
application so that transformation of speech can take place in a server engine.
The server can then return a text document as output on the mobile device (Barbash,
2001).On the other hand,processing power for handwriting recognition can reside
on the client.This is now becoming more of an option as handwriting applications
are better able to serve users (Barbash,2001).But again,this is only practical in some
contexts where the user has two hands free with which to write (Kristoffersen and
Ljungberg,1999).Finally,both of these input methods suffer from misrecognition
issues which can sometimes be resolved with a limited vocabulary range built into the
application or the use of integrated dictionaries (Noyes,1995).In most applications,
these misrecognition situations can only be identiﬁed through audio or visual
feedback and often need to be resolved via manual manipulation or mediation
(Noyes,1995;Dey et al.,2002).
Another formof input utilized in mobile applications is auto-population based on
context awareness.Context awareness is the ability for a computing device to
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797784
identify the task a user is focused on,the environment the task is occurring in,and
the status of the device.Asimple application is for a device to auto-populate the time
a task takes place based on a clock internal to the device.A more complicated
example could be used by ﬁeldworkers engaged in observational studies (Pascoe
et al.,2000).For instance,a mobile device,connected to other devices such as a GPS
tracking device and a digital thermometer,could auto-populate the time and
location that an observation takes place as well as the temperature.If this
application were integrated into a knowledge management system,a pre-conﬁgured
user interface could be pushed to the user that was customized for that location
(e.g.plants or animals that may be in the area).Additionally,sensors could identify
an animal by a worn radio collar and data on that subject could be pushed to the
user for further editing.While context awareness can aid in the input operation,it
may be very difﬁcult to view all of this on a small screen.Therefore,is it important to
have a complementary desktop application that can provide a visualization tool for
the context data (Pascoe et al.,2000).Context data can be entered in other ways as
well.One method is machine vision,which provides detection,identiﬁcation,and
tracking (Crowley et al.,2000;Branco,2001).
Other context aware solutions are designed to auto-conﬁgure applications based on
changes in the work context.Applications that a user may need could be pushed to
the mobile device as they approach a particular object or location (Kristoffersen and
Ljungberg,1999).Other studies have looked at the auto-population of customer or
site data as a mobile worker changes contexts (Spriestersbach et al.,2001).This type
of application has considerable positive impacts on the amount of input a user would
be required to enter.However,one limitation with this approach is the accuracy of
context aware applications.For instance,time clocks imbedded in computing devices
are not always accurate.Additionally,GPS tracking is only accurate up to a limited
distance.Spriestersbach et al.(2001) considered an approach that would deploy
sensors to key locations (e.g.customer sites) that would trigger context awareness in
the mobile application.This,however,does have a high cost of deployment and equal
or greater beneﬁts must be realized to make it a practical application.
One issue with using context awareness as input arises when it is used to sense
ambiguous data.Location and temperature are very factual and can be measured by
a variety of tested devices.Speech,touch,proximity,and facial expressions,however,
can be somewhat harder to measure accurately.Using a variety of input mechanisms
can improve reliability (Hinckley et al.,2000).For instance,a device in a shirt pocket
may trigger a close proximity reading,but since it is not being held (an act read by
touch sensors) it does not turn on (Hinckley et al.,2000).Artiﬁcial Intelligence (AI)
techniques are often used to address this challenge.However,since there is ambiguity
in the sensed data,the system programmer must often infer how to deal with this
data in the application (Branco,2001;Dey et al.,2002).Another tool is the use of
mediation to remove any remaining ambiguity (Dey et al.,2002).In mediation,
dialogue between the user and the system allows the user to resolve questions about
how ambiguous input should be interpreted.
Output techniques in mobile computing need to be concerned about two factors:
small screen size and minimal user visual attention availability.Output often serves
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 785
as feedback to an input action.For example,when a user drags and drops an icon
on their desktop,the input is entered by the mouse device and the output is the
visual feedback of seeing the icon move.To replace visual feedback mechanisms,
designers have employed tactile and audio feedback (Pascoe et al.,2000).Tactile
feedback includes the use of variable vibration and impulses when an area is
touched on the screen (Pascoe et al.,2000;Poupyrev et al.,2002).Audio feedback
includes voice synthesis and tones that occur when actions are performed
(Kristoffersen and Ljungberg,1999;Pascoe et al.,2000;Pirhonen et al.,2002).
Audio output has been effective at improving interaction and presenting non-visual
information on mobile devices (Brewster and Cryer,1999;Sawhney and Schmandt,
2000).Astudy of a bottling factory showed that repetitive streams of sounds allowed
people to keep track of activity,rate,and functioning of machines (Gaver et al.,
1991).Additionally,auditory awareness through the use of water sounds was found
to be an unobtrusive output mechanism.The rush of the water increased from a
splash to rapid water ﬂow to indicate various application states (Sawhney and
In some cases tactile and audio output are not practical or meaningful and the user
still needs to rely on visual forms of output.In these cases,simplicity is key so that
the user needs to apply only minimal visual attention (Pascoe et al.,2000).However,
some tasks are more complicated and require a more complex output screen.In these
cases small monocular output devices,such as the Eyewand,can be used to provide a
rich user interface (Guerlain et al.,1999).Additionally,applications can apply
scrolling for access to large amounts of information (Bertelsen and Nielsen,2000).In
some cases,proper output rendering can only be realized through a complementary
desktop application,which can allow for easier manipulation and interpretation of
results (Pascoe et al.,2000).
Context aware forms of I/O allow computing to become more ‘‘invisible’’ to the
user (Abowd et al.,2002).On the input side,this invisibility comes to fruition when
the system is able to infer data about the user’s identity,location,effect,or activity
by their presence and interactions in the environment (Abowd et al.,2002).On the
output side,this invisibility comes to light by designing output that a user can
monitor through their peripheral perception rather than conscious attention (Abowd
et al.,2002).Context awareness makes interaction with the computing machine less
distracting.This section will explore context aware input and control of output
Sawhney and Schmandt (1999,2000) have experimented with developing context
sensitive solutions for messaging in mobile environments.These typically employ
auditory user interfaces that accept speech input and provide audio output with
variable intrusiveness levels based on the context sensed by the device,message
priority,and usage level.For instance,if the device senses that the user is in a
meeting or is in noisy surrounding where an audio cue might not be recognized,it
will either not notify the user,or use an alternative means of output such as visual or
tactical (Sawhney and Schmandt,1999,2000).While these solutions have been
typically designed for e-mail and other messaging applications,they can be applied
to areas of dispatch and knowledge management.
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797786
2.3.4.Knowledge management solutions
Another technical HCI concern that is critical for mobile workers is knowledge
management.Since ﬁeld workers are usually alone most of the day in remote
locations,they have issues with staying abreast of ongoing development in other
projects as well as their enterprise in general (Wiberg and Ljungberg,1999).
Additionally,knowledge management is key for areas such as ﬁeld service and
ambulatory health care (Sugimatsu et al.,1994;Holtzman,1999).The study by
Fagrell et al.(2000) suggests the following main requirements for knowledge
1.Support evolving tasks and notify users of interdependencies.
2.Offer and overview of records,including annotations.
3.Suggest available expertise.
4.Filter information based on task and long-term interest.
5.Enable adaptation to user preferences and mobile device capabilities
(Fagrell et al.,2000,p.213).
It is important,however,for the server engine to identify the user connection and
device type,so that information sent can be customized to the computing
environment.Another key component of knowledge management is Just-In-Time
(JIT) knowledge delivery.For instance,knowledge could be pushed to a device using
context awareness and AI to assign contextual properties to ﬁles (Hirakawa et al.,
1998;Jones and Brown,2000;Pascoe et al.,2000).This could allow customer
records,maintenance manuals,and maps to become readily available as the user’s
context changes.Another example of JIT knowledge is in an equipment service
context.For instance,as a service worker disassembles a motor for repair,each part
removed can be scanned via a bar code into the mobile device.As the device receives
the bar code,it can display the next step in the disassembly process (Bertelsen and
Nielsen,2000).This method could also be used for work order management,as it is
often difﬁcult to display a full,complicated work order on a mobile device screen
(Kristoffersen and Ljungberg,1999).So,as the ﬁeld worker completes various steps
in the work order,the next task appears.Context-sensitive knowledge management
allows workers to be focused on the job rather than the computer-speciﬁc
Field service support systems are part of knowledge management solutions.
A system called Doctor was designed to support air-conditioning service engineers
(Sugimatsu et al.,1994).These ﬁeld engineers are challenged by the variety
(e.g.approximately 1000 types) of room air-conditioners on the market that they
need to support.Additionally,after troubleshooting the unit,they typically need
parts speciﬁc to the unit needing repair.Without the parts in hand,the ﬁeld service
engineer may need to make multiple trips to the customer’s site.Doctor aids the ﬁeld
worker by diagnosing the customer’s issue,before the ﬁeld worker is notiﬁed,by
comparing the case to similar repair cases stored in a database.Based on this
comparison,it sends a ‘‘top ten’’ list of appropriate parts to the ﬁeld worker along
with the call.The ﬁeld worker can then collect the appropriate parts before
responding to the call and can reduce time in searching through various repair
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 787
manuals.Experimentation of this system revealed that 90% of the customer repair
cases can be resolved by the top three parts on the list (Sugimatsu et al.,1994).The
process of dispatching the ﬁeld worker could be further automated by tying this
system to an inventory control database and a location-aware system so that the
closest worker with the right parts in hand could be dispatched,thus increasing
Given these technical considerations,the most important aspect for the mobile
worker is control.Users should be able to customize their application and device
easily to ﬁt their changing work environments (Barbash,2001).There are situations
where they may want to select from a set of interaction styles as they move from
contexts with high visual attention needs to those with low (Kristoffersen and
Ljungberg,1999;Barbash,2001).For instance,ﬁeldworkers doing observational
studies require high visual interaction with subjects during certain time periods,but
once the observation is over,the ﬁeldworker would want the opportunity to input
data in a more visual manner via traditional text editing (Pascoe et al.,2000).
Additionally,users may want to interact with data on a larger scale not available
within the mobile device (Kristoffersen and Ljungberg,1999;Pascoe et al.,2000;
Porn and Patrick,2002).So,some applications may require desktop interfaces for
more complex data manipulation,visualization,and retrieval.
Context awareness can also be used for dynamic conﬁguration as it can recognize
the capability of a unit and adapt media to the device capabilities (Phamet al.,2000).
These techniques may include splitting (e.g.to spilt a audio/video feed into its
individual output components and direct them to a PC and neighboring telephone
when no sound card is installed in the PC),conversion (e.g.to convert text to speech
when no text output method is available),and ﬁltering (e.g.delivery of sub-content
which can be rendered on the device such as delivery of only the audio part of a video
message) (Pham et al.,2000).Applications can also employ context awareness to
adapt to changes such as exhausting battery power or availability of networked
services.Middleware can be used as a solution to store context awareness,collect
context data from the application,and then react accordingly (Capra,2002).For
instance,a physical copy of the data may be preferred if the storage space on the
device permits (Capra,2002).Otherwise,users can view the data on-line.In another
example,the application may request that the middleware disconnect when
bandwidth is ﬂuctuating or battery power is low (Capra,2002).
A ﬁnal conﬁguration consideration is that in many cases ﬁeld workers may share
mobile devices (Guerlain et al.,1999).This is especially true in occupations with
shift-work.Devices and applications should be able to store user preferences and
customization that will re-appear upon log-in.
2.4.The development process
Work contexts,human considerations,and the computer systems and interface
architectures to address these all need to be considered in the design and development
process.Studies on automation implementation mirror ﬁndings of customizability
and usability.Studies found that successful automation implementations consider the
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797788
work environment and user needs in the ﬁnal solution (Engle and Barnes,2000;
Watad and Disanzo,2000;Erffmeyer and Johnson,2001;Speier and Venkatesh,
2002).Additionally,they include evaluation efforts shortly after implementation to
identify and address areas of opportunity (Holtzman,1999;Kristoffersen and
Ljungberg,1999;Pascoe et al.,2000).Often,the full impacts of the implementation
effort are not known until after the user has had time to interact with the computing
system.The ultimate goal of the automation effort should be to ensure that the
technology aids the worker and does not take away fromor hinder themat any level.
With proper and timely evaluation,adjustments can be made to ensure efﬁciency and
productivity are maintained or improved.
Holtzman’s (1999) process to design a ﬁeld service system for emergency health
care workers can provide an example of development using HCI considerations.The
mobile computing solution was designed using a user-centered systems engineering
approach.For the process,the designers employed an extensive analysis of human
limitations,collected information on user data input and access needs,and reviewed
new computer technologies that could counteract the limitations.The steps of the
design process included:(1) analysis of tasks,users,and work contexts;(2) design of
solutions to the problems identiﬁed;(3) implementation of the solutions;and (4)
evaluation of the design.First,data was collected through observation,interviews,
industry literature and ﬁeld visits.A scenario-based engineering process was
employed where possible work situations were described and validated by experts,
shared with the development team to create a common vision,used to create
operationally deﬁned objectives for the system,and then developed into prototypes
for review by potential users.These prototypes went through the design process as
they were continually evaluated,revised,and presented to the sample user group for
many weeks.Once a viable prototype was chosen,it was implemented,evaluated,
and revised again.The result was a solution with high user acceptance that is
complete,fast,and easy to use (Holtzman,1999).
3.Classiﬁcation of HCI issues for mobile computing
The research in this paper reviews literature focusing on recent studies of mobile
computing in a variable work context.Most of the literature was identiﬁed by
searching the customary library databases such as ABI/INFORM,Math Science,
ACM,IEEE,Business Periodicals Index (BPI),and World Cat.The descriptors such
as mobile computing,HCI,implementation issues were used to retrieve several
hundred abstracts.The abstracts were carefully reviewed and a decision was made
regarding the inclusion of a particular abstract/study in our research.The inclusion
criteria consisted of relevance,length and time of publication.Recently published
studies were given higher priority for inclusion.Small papers (less than 3 pages) and
papers with obscure/ambiguous abstracts were given lower priority for inclusion.
A total of 68 different studies were used and were catalogued according to the
main HCI issues dimensions in the literature review section.Table 1 illustrates the
number of articles and the authors for each HCI issues dimension.
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 789
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797790
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 791
The results in Table 1 indicate that HCI issue related to I/O techniques is of
paramount importance,and the HCI issues related to knowledge management and
human characteristics are the least addressed issues in the literature.
A few studies identiﬁed mobile computing applications in certain industries.We
catalogued these industries in Table 2.
Table 2 illustrates that of all the mobile computing applications for the workforce,
ﬁeld service was the main focus.It is hailed as the most important corporate activity
that can improve customer satisfaction and competitive stance (Rewari,1993).When
linking ﬁeld service with a corporate sales force through a CRM system,beneﬁts
become greatly improved.Health care realizes beneﬁts of mobile computing both in
the hospital setting where workers are ambulatory a majority of the time and in the
ﬁeld with emergency care teams.Likewise,mobile computing helps ﬁeld researchers
capture data as they attempt to meld into the environment during observational
studies.There are many more applications for mobile computing in a variable work
context beyond these examples.However,these examples bring to light the
heterogeneous character of this user audience,emphasizing that there is no one
solution to meet everyone’s needs.User acceptance and proﬁtability studies show
that individual user needs must be taken into account when designing mobile
4.Summary,conclusions and directions for future research
The primary objective of this paper was to summarize the ‘‘state-of-art’’ HCI
research and issues related to the mobile computing in a variable work context.The
results of our study indicate that the primary HCI issues related to mobile computing
in variable work context are issues related to I/Otechniques,and the issues related to
development and implementation.Mobile computing has applications varying from
journalism to ﬁeld service.Most of the applications of mobile computing in the
workforce are in the ﬁeld services.
The workers that use mobile computing are a heterogeneous group.Those
workers that are employed in a variable work context are even more diversiﬁed,
considering that their work context changes often—many times within one workday.
Their needs are also variable.It is important to understand and consider their work
context,human characteristics,and technical options available when designing
mobile computing resources.Additionally,it is important to move beyond
traditional modes of computing and discover alternative methods that could be
developed to fulﬁll these needs.Ubicomp is an area of research that provides many
non-traditional interaction solutions that mobile computing can leverage.The
integration of information systems into the work environment (whether mobile or
static) should not hinder the workﬂow or frustrate users.Rather,it should enhance
existing processes.This can only be done by understanding user needs,identifying
available resources,and then designing to those needs.
As we move toward a more context aware computing environment privacy
becomes a concern.A balance has to be found between the individual’s need for
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797792
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 793
privacy and corporate,government,and society’s need for information (Abowd and
Mynatt,2000;Hong,2002).Mechanisms need to be considered and built into mobile
computing systems to protect privacy by restricting certain context queries and
limiting storage of context data.
Additionally,much of the research in mobile computing in a variable work
context focuses on activities such as ﬁeld service,sales,and health care.These areas
may well be the earlier adopters of non-traditional computer interaction methods as
these techniques are more likely to bring quick beneﬁts in terms of worker
productivity,user acceptance,and return on investment.As the technology matures,
more studies need to be done in the use of mobile computing in other industries that
have workers performing jobs outside of the ofﬁce setting.
Finally,areas of ubicomp research are extremely applicable to mobile computing.
Since advances in this area are relatively new,they have a very high ‘‘cool factor’’ as
machines are being designed to interact with humans in more natural ways.This
phenomenon may bring early adopters who are fascinated by the functionality,who
later become frustrated by the limitations and error rates of these early appliances.It
is important that more research is done in this area so that researchers,practitioners,
and developers can begin to understand the ubicomp design space as they do the
GUI interface system that has been around for over 20 years (Hinckley et al.,2000).
There are now academic programs designed to explore these facets (Midkiff,2002).
By quantifying user experience and continued research of these tools in use,
designers and developers can deliver on solutions to help the user improve
productivity and the corporations improve proﬁts.
Abowd,G.D.,Mynatt,E.D.,2000.Charting past,present,and future research in ubiquitous computing.
ACMTransactions on Computer–Human Interaction (TOCHI) 7 (1),29–58.
Abowd,G.D.,Mynatt,E.D.,Rodden,T.,2002.The human experience.Pervasive Computing 1 (1),48–57.
Baber,C.,1997.Beyond the desktop:designing and using interaction devices.Academic Press,San Diego.
Baber,C.,Noyes,J.M.,1996.Automatic speech recognition in adverse environments.Human Factors 3,
Baber,C.,Knight,J.,Haniff,D.,Cooper,L.,1999.Ergonomics of wearable computers.Mobile Networks
and Applications 4 (1),15–21.
Barbash,A.,2001.Mobile computing for ambulatory health care:points of convergence.Journal of
Ambulatory Care Management 24 (4),54–60.
Bassil,J.,1996.Multimedia Over Mobile Networks Using the H.324 Family.The Institution of Electrical
Bertelsen,O.W.,Nielsen,V.,2000.Augmented reality as a design tool for mobile interfaces.Conference
Proceedings on Designing Interactive Systems:Processes,Practices,Methods,and Techniques,
Blattner,M.M.,1994.In Our Image:interface design in the 1990s.IEEE MultiMedia 1 (1),25–36.
Branco,P.,2001.Tools and techniques for interaction:challenges for multimodal interfaces towards
anyone anywhere accessibility:a position paper.Proceedings of the 2001 EC/NSF Workshop on
Universal Accessibility of Ubiquitous Computing:Providing for the Elderly,pp.26–27.
Brewster,S.A.,Cryer,P.G.,1999.Maximizing screen-space on mobile computing devices.Summary
Proceedings of ACMCHI ’99,pp.224–225.
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797794
Brodie,J.,Perry,M.,2001.Designing for mobility,collaboration and information use by blue-collar
workers.ACMSIGGROUP Bulletin 22 (3),22–27.
Butts,L.,Cockburn,A.,2001.An evaluation of mobile phone text input methods.Conferences in
Research and Practice in Information Technology 7,55–57.
Capra,L.,2002.Mobile computing middleware for context-aware applications.Proceedings of the 24th
International Conference on Software Engineering,pp.723–724.
Chandra,J.,March,S.T.,Mukherjee,S.,Pape,W.,2000.Information systems frontiers.Communications
of the ACM43 (1),71–79.
Cheng,L.,Robinson,J.,1998.Dealing with speed and robustness issues for video-based registration
on a wearable computing platform.Second International Symposium on Wearable Computers,
erard,F.,2000.Perceptual user interfaces:things that see.Communications of
the ACM 43 (3),54–64.
Dey,A.,Mankoff,J.,Abowd,G.,Carter,S.,2002.Speech and ambiguous input:distributed mediation of
ambiguous context in aware environments.Proceedings of the 15th Annual ACMSymposiumon User
Interface Software and Technology,Vol.4(2),pp.121–130.
Engle,R.L.,Barnes,M.L.,2000.Sales force automation usage,effectiveness,and cost-beneﬁt in Germany,
England,and the United States.The Journal of Business and Industrial Marketing 15 (4),
Erffmeyer,R.C.,Johnson,D.,2001.An exploratory study of sales force automation practices:
expectations and realities.The Journal of Personal Selling and Sales Management 21 (2),167–175.
Esposito,C.,1997.Wearable computers:ﬁeld-test results and system design guidelines.In:Howard,S.,
Hammond,J.,Lindgaard,G.(Eds.),Interact ’97.Chapman & Hall,London,pp.493–500.
Fagrell,H.,Forsberg,K.,Sanneblad,J.,2000.FieldWise:a mobile knowledge management architecture.
Proceeding of the ACM2000 Conference on Computer Supported cooperative Work,pp.211–220.
Gaver,W.W.,Smith,R.B.,O’Shea,T.,1991.Effective sounds in complex systems:the ARKola
simulation.Proceedings of the CHI ’91,April 28–May 2,1991.
Goth,G.,1999.Mobile Devices Present Integration Challenges.IT Pro,11–15.
collaborative ﬁeld operation with personal information processing systems.Mobile Networks and
Applications 4 (1),37–48.
Hinckley,K.,Pierce,J.,Sinclair,M.,Horvitz,E.,2000.Sensing techniques for mobile interaction.
Proceedings of the 13th annual ACM Symposium on User Interface Software and Technology,Vol.
Hirakawa,M.,Mizumoto,S.,Yoshitaka,A.,Ichikawa,T.,1998.A situation-sensitive interface for the
management of personal documents.Proceedings of the International Workshop on Multimedia
W.,1996.Curricula for Human–Computer Interaction.ACM Press,New York.
Holtzman,T.G.,1999.Computer–human interface solutions for emergency medical care.In Interactions 6
Hong,J.I.,2002.The context fabric:an infrastructure for context-aware computing.Conference Extended
Abstracts on Human Factors in Computer Systems,pp.554–555.
Jones,G.J.F.,Brown,P.J.,2000.Information access for context-aware appliances.Proceedings of the 23rd
annual international ACM SIGIR Conference on Research and Development in Information
Kalakota,R.,Stallaert,J.,Whinston,A.,1996.Mobile agents and mobile workers.Proceedings of the
29th Annual Hawaii International Conference on System Sciences,pp.354–365.
Karat,C.M.,Halverson,C.,Horn,D.,Karat,J.,1999.Patterns of entry and correction in large
vocabulary continuous speech recognition systems.CHI 99 Conference Proceedings,pp.568–575.
Kentrick,C.,2002.The transfer of wireless data:meeting future demand.Futurics 26 (1/2),92–107.
Kraut,R.E.,Miller,M.D.,Siegel,J.,1996.Collaboration in performance of physical tasks:effects on
outcomes and communication.Proceedings of CSCW96,November 16–20,pp.57–66.
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 795
Kristoffersen,S.,Ljungberg,F.,1999.Making place to make IT Work:empirical explorations of HCI for
mobile CSCW.Proceedings of the International Conference on Supporting Group Work (GROUP
Laramee,R.S.,Ware,C.,2002.Rivalry and interference with a head-mounted display.ACMTransactions
on Computer–Human Interaction (TOCHI) 9,238–251.
Lucas.,Jr,H.C.,2001.Information technology and physical space.Communications of the ACM44 (11),
Luff,P.,Heath,C.,1998.Mobility in collaboration.Proceedings of the CSCW ’98 Conference on
Computer-Supported Cooperative Work,pp.305–314.
MacKenzie,S.,2002.Papers:mobile text entry using three keys.Proceedings of the Second Nordic
Conference on Human–Computer Interaction,pp.27–34.
Mahajan,S.,Donahoo,M.J.,Navathe,S.B.,Ammar,M.H.,Malik,S.,1998.Grouping techniques for
update propagation in intermittently connected databases.Proceedings of the ICDE ’98,pp.1–8.
Miah,T.,Bashir,O.,1997.Mobile workers:access to information on the move.Computing and Control
Midkiff,S.F.,2002.Rapid prototyping of computing systems.Pervasive Computing 1,14–17.
Moray,M.E.,1996.Mass storage for delivery of maps and documents to the ﬁeld.IEE Colloquium on
Mass Storage Systems,Vol.206,pp.2/1–2/6.
Noyes,J.M.,1995.Enhancing mobility through speech recognition technology.IEE Colloquium on
Developments in Personal Systems,pp.4/1–4/3.
Pascoe,J.,Ryan,N.,Morse,D.,2000.Using while moving:HCI issues in ﬁeldwork environments.ACM
Transactions on Computer–Human Interaction (TOCHI) 7 (3),417–437.
Perry,M.,O’Hara,K.,Sellen,A.,Brown,B.,Harper,R.,2001.Dealing with mobility:understanding
access anytime,anywhere.ACM Transactions on Computing–Human Interaction (TOCHI) 8 (4),
Pham,T.,Schneider,G.,Goose,S.,2000.A situated computing framework for mobile and ubiquitous
multimedia access using small screen and composite devices.Proceedings of the Eighth ACM
International Conference on Multimedia,pp.323–331.
Pirhonen,A.,Brewster,S.,Holguin,C.,2002.Speech,Audio,Gesture:gestural and audio metaphors as a
means of control for mobile devices.Proceedings of the SIGCHI Conference on Human Factors in
Computing Systems:Changing Our World,Changing Ourselves,Vol.4(1),pp.291–298.
Porn,L.M.,Patrick,K.,2002.Mobile computing acceptance grows as applications evolve.Healthcare
Financial Management 56 (1),66–70.
Poupyrev,I.,Maruyama,S.,Rekimoto,J.,2002.TouchEngine:a tactile display for handheld devices.HCI
for Mobility 9,644–645.
Rewari,A.,1993.AI in Corporate Service and Support.IEEE Expert 8 (6),5–38.
Rolland,J.P.,Biocca,F.A.,Barlow,T.,Kancherla,A.,1995.Quantiﬁcation of adaptation to virtual eye
location in see-thru head-mounted displays.Proceedings of the IEEE Virtual Reality Annual
Sawhney,N.,Schmandt,C.,1999.Nomadic radio:scaleable and contextual notiﬁcation for wearable
audio messaging.Proceeding of the CHI 99 Conference on Human Factors in Computing Systems:the
CHI is the Limit,May 1999,pp.15–20.
Sawhney,N.,Schmandt,C.,2000.Nomadic radio:speech and audio interaction for contextual messaging
in nomadic environments.ACM Transactions on Computer–Human Interaction (TOCHI) 7 (3),
Silfverberg,M.,MacKenzie,I.S.,Korhonen,P.,2000.Predicting text entry speed on mobile phones.
Proceedings of the CHI 2000 Conference on Human Factors in Computing Systems,Vol.2(1),pp.9–13.
Smailagic,A.,1999.An evaluation of audio-centric CMU wearable computers.Mobile Networks and
Applications 4 (1),59–68.
Speier,C.,Venkatesh,V.,2002.The hidden mineﬁelds in the adoption of sales force automation.Journal
of Marketing 66 (3),98–111.
Sprey,J.A.,1996.Videoconferencing as a Communication Tool.IEEE Transactions on Professional
Communication 40 (1),41–47.
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797796
Spriestersbach,A.,Vogler,H.,Lehmann,F.,Ziegert,T.,2001.Integrating context information into
enterprise applications for the mobile workforce—a case study.Proceedings of the First International
Workshop on Mobile Commerce,pp.55–59.
Sugimatsu,H.,Hori,S.,Sawada,A.,Azuma,K.,1994.A CBR application:service productivity
improvement by sharing experience.1994 IEEE Symposium on Emerging Technologies and Factory
Turisco,F.,2000.Mobile computing is next technology frontier for healthcare providers.Health Care
Financial Management 54 (11),78–80.
Varile,G.,Zampolli,A.,1997.Survey of the State of the Art in Human Language Technology.Cambridge
Varshney,U.,1999.Networking support for mobile computing.Communications of AIS 1 (1),1–30.
Varshney,U.,Vetter,R.,2000.Emerging mobile and wireless networks.Communications of the ACM43
Watad,M.M.,Disanzo,F.J.,2000.Case study:the syngerism of telecommuting and ofﬁce automation.
Sloan Management Review 41 (2),85–96.
Weiser,M.,1991.The computer of the 21st century.Scientiﬁc American 265 (3),66–75.
Wiberg,M.,Ljungberg,F.,1999.Exploring the vision of anytime,anywhere in the context of mobile
work.Knowledge Management and Virtual Organization.
Wickelgren,I.J.,1996.Local-area networks go wireless.IEEE Spectrum 33,34–40.
Zimmerman,T.G.,1999.Wireless networked digital devices:a new paradigm for computing and
communication.IBMSystems Journal 38 (4),566–574.
ARTICLE IN PRESS
J.York,P.C.Pendharkar/Int.J.Human-Computer Studies 60 (2004) 771–797 797