The Ambient Calendar
Owen Phelan, Lorcan Coyle, Graeme Stevenson, and Steve Neely
School of Computer Science and Informatics,
UCD Dublin, Ireland
owen.phelan@ucd.ie
Abstract. It is becoming dicult to convey information from an ever-
increasing number of digital sources to users in a condensed and meaning-
ful way. This growth has particularly occurred with peripheral informa-
tion sources. These are of general interest to users, but do no require
or typically command constant focus or attention. Examples include
weather, stock data, blogs, and calendars. Ambient Displays present in-
formation unobtrusively in an intelligent fashion using abstract visual
cues and metaphors and have the possibility of acting as a complement
to information ltering systems. We describe the implementation of an
ambient display that contains elements representing time, weather, pub-
lic transport departure times, and the proximity of friends. An initial
impact study was undertaken and found a high sense of usefulness and
curiosity in the nished application and in the eld as a whole.
1 Introduction
Fig. 1. The Ambient Calendar
With the growth of the Internet,
users are increasingly bombarded
with larger amounts of visual and
textual information, including news
and weather, blogs, message boards,
emails and RSS feeds. It is becom-
ing increasingly dicult to convey
this information in an easily accessible
and understandable manner. One ap-
proach to conveying such information
in a more condensed and meaningful
way is the use of Ambient Displays,
which utilise visualization techniques
to publish information in an unobtru-
sive manner.
Ambient Displays uses abstract
metaphors and visual cues to inform
users of peripheral information sources, rather than an explicit text-based means
(Section 2). For example, Figure 1 shows a cartoon image that may be situated
in a living room or oce. This image could be decomposed into dierent el-
ements, with characters in the foreground, clouds, trees and mountains in the

background, and a sky with clouds and a sun, each associated with a data source.
By changing the elements in the image, e.g., by making them larger or changing
their colour or shape, we can convey something about the underlying informa-
tion.
We developed The Ambient Calendar, an ambient display that represents
calendar information, and information from other Internet and local sensors
(Section 3). Instead of explicitly telling a user that there are events in the calen-
dar, the system conveys this implicitly by putting clouds in the sky. By glancing
at the display and seeing a cloudy sky the user can tell that they have a busy
day ahead. Major in
uences for this project include the prototype ambient dis-
play outlined by Neely et al. [1], and the InfoCanvas developed by Stasko and
Miller [2]. A novel aspect of The Ambient Calendar is that it locates proximate
Bluetooth devices and personalises the display's content for nearby users.
Like traditional art, the appeal and perceived usefulness of an ambient device
is subjective and varies from person to person. A short preliminary user study
was undertaken (Section 4) to investigate the initial impact and acceptance of the
display, and to inform renements before a longer study is carried out. This was
done using a combination of observations as users interacted with the display,
and through the application of heuristics. The results of these early evaluations
show that users are highly intrigued by the concept, and nd the nished product
generally informative and useful.
2 Background
Wisneski et al. [3] highlighted the issue that interactions between people and
digital information are entirely conned to graphical user interfaces that in-
volve traditional input devices (e.g., keyboard and mouse). They proposed the
ambientROOM 1 | an environment populated with ambient devices, each rep-
resenting an individual data source (as shown in Figure 2). Wisneski et al. used
the concepts outlined by Weiser [4], who described a Ubiquitous Computing en-
vironment as being technologically saturated and interconnected. He described
good ubiquitous computing as consisting of calm technologies, which disappear
into the background of human attention and become second nature. Calm Tech-
nologies aim to be unobtrusive and should go unnoticed until an important or
unexpected event brings them into the foreground [5].
Vande Moere speculates that ambient displays could be considered appliances
of persuasive visualization [6]. The author proposes that the concept of ambi-
ent display is becoming more persuasive as increasing numbers of information
visualizations are used to highlight important issues. Examples include social
and environmental concerns, health, welfare and economic data. The true use-
fulness of ambient displays, the author argues, may lie in the conveyance of these
important issues in a more eective manner by informing users about personal
interests on a personal level.
1 More information on the ambientROOM available on their homepage http://
tangible.media.mit.edu/projects/ambientroom/

In order for ambient displays to reach their potential, articial intelligence
techniques must provide the basis of a dynamic back-end to provide and lter
information and determine when an unobtrusive element becomes more explicit.
In terms of traditional AI, Ramos argues that certain aspects of the eld can be
augmented to the ambient information paradigm to assist in supporting the users
activities and better decision making with access to essential knowledge [7]. This
notion is termed Ambient Intelligence (AmI). Ramos proposed that features of
traditional AI concepts such as Machine Learning, Planning, Knowledge Rep-
resentation, Speech Recognition, and Computer Vision can all assist in a more
useful human-interactive system [7].
2.1 Examples in the eld
Several ambient displays have been implemented in the past few years ranging
from the physical icons of the MIT ambientROOM, to wearable ambient displays
that relate to biometric data of the wearer [6] to the Apple Mac OS X bouncing
dock icons.
The ambientROOM uses a mixture of light, physical motion and air
ow, as
well as physical icons to represent activities in the local environment [3]. Above
a desk in the room sits a partially transparent glass pane with liquid encased.
As the lab's pet hamster runs in its wheel, the liquid encased in the glass pane
ripples informing its owner that the hamster is active.
The InfoCanvas allows users to create and personalise scenes based on data
sourced online. The display depicts a colourful beach panorama that could tell
the user that it is a late hour (as the boat is on the right-most edge), and that
the user's stocks are down (as the bird is closer to the ground). The artifacts
in the InfoCanvas are personal to the owner and the metaphors are deliberately
abstract and subjective. The InfoCanvas can be seen as much a piece of art as
an information display.
Ishii and Ullmer, Tangible Bits 5
on the map bound to the physical location of the Domephicon. (Fig. 11).
Simultaneously, the arm-mounted activeLENS displays a
spatially-contiguous 3D view of MIT campus (Fig. 11).By grasping and moving the activeLENS (a physically
embodied window), the user can navigate the 3D
representation of campus building-space.
The Great Dome phicon acts both as a container of bits
which represent the MIT campus, and as a handle for
manipulating the map. By rotating or translating the Dome
object across the desk surface, both the 2D desk-view and3D activeLENS-view are correspondingly transformed. The
user is thus interacting visually and haptically with three
spaces at once—the physical-space of the Dome object; the2D graphical space of the desk surface; and the 3D graphical
space of the activeLENS.
The user may then take a second phicon, this time of theMedia Lab building, and place it onto the surface of thedesk. The map then rotates and scales such that the secondphicon is bound to the location of the Media Lab building
on the map. Now there are two physical constraints andhandles on the MIT campus space, allowing the user to
simultaneously scale, rotate, and translate the map by
moving one or both objects with respect to each other.
Because each phicon serves as an independent locus of
control, the user may grasp and manipulate both objects
simultaneously with his/her two hands. Alternatively, two
users may independently grasp separate building objects,
cooperatively manipulating the transformation of theGeospace. In this fashion, there is no one locus of control
as is true in point-and-click mouse interaction; rather, theinteraction is constrained by the physics of the physical
environment, supporting multiple pathways of single- and
multi-user interaction.
By bringing a passiveLENS device onto the desk, the user
may interact with satellite-imagery or future/past-time
overlay views of the map space, or explore alternateinteractions consistent with physical instantiation of theMagic Lens metaphor [17].
With two phicon objects on the desk, there is an issue of
ambiguity that must be resolved. For instance, when one
or both phicons are rotated independently, how should the
application respond? We currently ignore this conflictinginformation, but could also imagine other interpretations
such as warping the map view. To resolve this ambiguity,
we designed a rotation-constraint instrument made of two
cylinders mechanically coupled by a sliding bar as shown inFig. 12. This instrument has mechanical constraints whichprohibit independent rotation and realize distinct axes of
scaling and rotation. By uilding in these physical
constraints, we resolve the question of ambiguity in thisparticular c se.
ambientROOMThe ambientROOM complements the graphically-intensive,
cognitively-foreground interactions of the metaDESK by
using ambient media – ambient light, shadow, sound,
airflow, water flow – as a means for communicatinginformation at the periphery of human perception. The
ambientROOM is based on Steelcase's Personal Harbor™
unit, a 6’ x 8’ freestanding room, which we have augmented
with MIDI-controllable facilities. The ambientROOM isdesigned to employ both the foreground and background of
users' attention.

Figure 13 ambientROOM based on Personal Harbor™
In normal day to day interactions, we get information intwo main ways. First, we get information from what we
are focusing on, where our center of attention is directed.When we are speaking with a colleague in the office, we are
consciously focusing on that person and receivinginformation directly from them. But at the same time, we
are also getting information from ambient sources. We
may have a sense of the weather outside from ambient cues
such as light, temperature, sound, and air flow from nearby
windows. We may also have an idea of the activities of
colleagues in the area from the ambient sound and the
visible presence of passers-by.
In contrast to the conscious foreground processing occurringin discussions with a colleague, much of this ambientinformation is processed through background
communication channels. Our goal for the ambientROOMis to explore how we can take advantage of this natural

Figure 11 Phicon and activeLENS
Figure 12 Sc ling and Rotation Device with embedded
mechanical constraints
Fig. 2. Stasko and Millers InfoC nvas, the MIT ambientROOM and Neely et al.'s
Visual Calendar [1]
Neely et al. proposed a Visual Calendar prototype (shown in Figure 2) that
they used to demonstrate how d erent types of information might be conveyed
in an ambient fashion [1]. Their prototype depicts a small town with artifacts

that move from left to right of the display depicting time, as well as other arti-
facts that represent trac and weather information. The image conveys location
information by using graphical representations of school, oce, and work, along
with avatars of friends or family members of the user. This approach is similar
to Microsoft Research's Whereabouts Clock, which displays family members' re-
mote location on a clock display situated in the home [8]. The work presented
here is based on Neely et al.'s original Visual Calendar proposal.
3 Implementation
Fig. 3. A scene from South Park, and some states of the Ambient Calendar
The main approach considered was how to convey information in a meaning-
ful manner using a single image, whose constituent parts can be decomposed and
treated as individual metaphors, while generating an intelligent data gathering
and ltering back-end that could manipulate each of the images' constituent
parts. The visualization is intended to mimic of a piece of art or scene that is
generally familiar to users. As such, we decided to use a screenshot from the
popular television show South Park2 (the rst image in Figure 3) as inspira-
tion. The image shows an everyday scene, with a variety of artifacts, including
avatars, which have been shown to be useful in ambient displays [9].
A mapping between the individual elements and potential data sources can
then be considered. The scenes in Figure 3 show the Ambient Calendar in vari-
ous states. The display owner has ve calendars subscribed (represented by the
trees), is associated with two blogs (represented by the mountains), has three
remaining events today in their calendar (represented by the clouds), their pub-
lic transport is approaching (shown in the modied bus stop symbol), and there
is a nearby friend associated with a Bluetooth Device name (represented by the
female avatar on the right).
The other images in Figure 3 represent other times of the day. The middle
image conveys to the user that it is mid evening, that they have a busy evening
2 Image taken from http://www.southparkstudios.com/ - South Park is a copyright
of Viacom Inc.

ahead with numerous events. The user is associated with three blogs, as well
as numerous calendars. The nal image shows a display at night time. There is
one more event in the calendar for that day, the user in question has one blog
associated with them, and their public transport (represented by a bus) has just
arrived (the public transport information comes from the Dublin Luas3 online
timetable).
We use Construct4, a distributed context-aware framework [?,10], to store,
distribute, and query for information used by the display. Construct can be de-
ployed on multiple systems, and uses zero-conguration networking technologies
to allow for resource discovery on local networks. This is useful for distributing
ambient information around a network and will make it easier to implement
intelligent collaborative or public ambient displays in the future.
4 Evaluation
Manko et al. [11] outline methods and tools for dening characteristics of an
eective Ambient Display. We initially planned to assess the common ambient
properties of usability, aesthetics and peripherality using a heuristic evaluation.
This involves the use of users who would analyze and critique the device based
on a set of criteria, for example, visibility of structure. In the eld of ambient
information systems, evaluation aims to explain how often a display is used,
how much it improves the life of its users, and how the system becomes learned
suciently well by its users.
Manko et al. designed a list of heuristics specically engineered for analysis
of ambient displays. They argued that an earlier list of heuristics derived by
Nielsen and Molich [12] did not relate to ambient displays, but rather human-
computer interaction (HCI) as a whole. The evaluation of ambient displays is
inherently problematic; they require relatively little traditional physical interac-
tion beyond initial setup and visual contact. We generated a questionnaire from
these seven heuristics that allowed our test subjects to quantify to what extent
they agreed with each. These are outlined below:
{ Q1: Did you nd the kinds of information relevant and useful?
{ Q2: Was the display obtrusive or interrupting?
{ Q3: Did the display attract attention of others?
{ Q4: Was there a feeling of Information Overload whenever you interacted
with the display?
{ Q5: Did you understand the information conveyed after you were introduced
to the display?
{ Q6: Did you understand the overall content when glancing at the display?
{ Q7: Was the display aesthetically pleasing?
3 The Luas is Dublin's public transport tram system
4 More information is available from Construct's homepage: http://www.
construct-infrastructure.org/

Manko et al. discuss the proper conduct of these evaluations, and argue
that tests should be longitudinal, lasting several weeks/months and be closely
evaluated by trained usability experts of at least 5 years experience. It is our
opinion that although the method using trained experts would produce accurate
results, it is not an entirely denitive or complete manner of evaluating the
usability of such devices. This is because these devices are user-centric by design,
and if they were to reach any manner of prevalence it would be among users who
have a wide variety of experiences using Information Technology, from Novice
to Expert.
With this premise under consideration, a questionnaire and usability study
was drafted to evaluate the usefulness of the display. It was targeted at four
\user groups", namely Novice (those who rarely use computers), Medium (those
who use it for a single purpose, perhaps work), Experienced (those who use IT
at home and work frequently), and High (these users were fourth year Computer
Science students) The following sections discuss this experiment in greater detail,
along with the results and analysis.
4.1 Experiment
To evaluate the Ambient Calendar, a qualitative Heuristic evaluation outlined
in Manko et al. was used. The display was placed in numerous situations for
short periods of time. There were eight test subjects, (termed Subjects A, B,
C, D, E, F, G, H). Subjects A and B were novice users, C and D were oce
workers who had basic training with IT interactions, E and F were experienced
computer users, and subjects G and H were Computer Science students with an
introductory understanding of ambient technologies, and an advanced knowledge
of Information System design. This spread of user qualications and backgrounds
intends to highlight the dierent responses due to the qualitative study. The
display was positioned in various peripheral locations. Subjects A and B took the
experiment in a domestic setting, and the remaining subjects took the evaluation
in their workplace or lab. Each subject evaluated the display over a period of
one hour.
Experiment Results and Conclusions
The overall evaluation results (Figure 4) show high satisfaction among all the
subjects. The overall results show the display is a generally useful, clearly mapped
and aesthetically pleasing application. There was also little feeling of informa-
tion overload during interactivity. Although these are generally positive results,
some problems were highlighted. For instance, there was an average score of 3.1
out of 5 in terms of misunderstanding the information after the visual elements
were initially explained.
Those who identied themselves as being Medium or Experienced users said
the information was highly useful, unobtrusive and aesthetically pleasing. During
post-evaluation interview, all but one of the subjects (Subject E) stated the
display had \attracted signicant attention from colleagues", which spawned

detailed data with a highly abstract image. Our future work will focus on the
trade-os that exist between presenting abstract personal information, and in-
formation that is easily interpretable to others. We will also investigate the most
appropriate visual metaphors for dierent types of information.
The evaluation was valuable in verifying the usefulness of the display, how-
ever one limitation of the study is that similar evaluations involved longitudinal
studies (over many weeks or months), carried out by trained usability evalu-
ators [11]. Although this would produce high quality results, our contention
is that involving users of varying knowledge and experience will provide valu-
able information. User-centric technologies should be tested among the cohorts
whose later uptake is the key to them becoming prevalent. We are planning a
more longitudinal study which will investigate Weiser's view that if people learn
a technology suciently well it will simply disappear into the background.
The Ambient Calendar can be extended in several ways. As speculated by
Vande Moere [6], we believe that ambient displays can be useful in persuading
users to take actions. With AI techniques such as those discussed in Ramos, and
particularly recommender algorithms, more intuitive and useful displays can be
devised. We will exploit other online sources, such as social networking sites, for
rich information to aid in the personalization of the display's content. We are
also investigating how other nodes of the Ambient Calendar running on the local
network can be used to convey collaborative or shared information.
References
1. Neely, S., Stevenson, G., Nixon, P.: Assessing the suitability of context information
for ambient display. In Hazlewood, W.R., Coyle, L., Consolvo, S., eds.: Proc. of
1st Workshop on Ambient Information Systems. Colocated with Pervasive 2007,
Toronto, Canada. Volume 254 of CEUR Workshop Proceedings ISSN 1613-0073.
(2007)
2. Miller, T., Stasko, J.: Artistically conveying peripheral information with the in-
focanvas. In: Advanced Visual Interfaces (AVI) '02 Conference, Trento, Italy, Ad-
vanced Visual Interfaces (AVI) '02 Conference (2002)
3. Wisneski, C., Ishii, H., Dahley, A., Gorbet, M., Brave, S., Ullmer, B., Yarin, P.:
Ambient displays: Turning architectural space into an interface between people
and digital infomation. In: Proceedings of the First International Workshop on
Cooperative Buildings (CoBuild '98), MIT Media Laboratory, Springer (1998)
4. Weiser, M.: The computer for the 21st century. Scientic American 265(3) (1991)
66{75
5. Miller, T., Stasko, J.: The infocanvas: Information conveyance through personal-
ized, expressive art. Technical report, College of Computing/GVU Center (2002)
6. Moere, A.V.: Towards designing persuasive ambient visualization. In Hazlewood,
W.R., Coyle, L., Consolvo, S., eds.: Proc. of 1st Workshop on Ambient Information
Systems. Colocated with Pervasive 2007, Toronto, Canada. Volume 254 of CEUR
Workshop Proceedings ISSN 1613-0073. (2007)
7. Ramos, C.: Ambient intelligence - a state of the art from articial intelligence
perspective. In: EPIA Workshops. (2007) 285{295
8. Taylor, A.S., Harper, R., Swan, L., Izadi, S., Sellen, A., Perry, M.: Homes that
make us smart. Personal Ubiquitous Comput. 11(5) (2007) 383{393

9. Forlizzi, J., Li, I., Dey, A.: Ambient interfaces that motivate changes in human
behavior. In Hazlewood, W.R., Coyle, L., Consolvo, S., eds.: Proc. of 1st Work-
shop on Ambient Information Systems. Colocated with Pervasive 2007, Toronto,
Canada. Volume 254 of CEUR Workshop Proceedings ISSN 1613-0073. (2007)
10. Dobson, S., Nixon, P., Coyle, L., Neely, S., Stevenson, G., Williamson, G.: Con-
struct: An open source pervasive systems platform. In: Consumer Communications
and Networking Conference (CCNC), 2007, 4th IEEE, Las Vegas, NV, USA. (2007)
1203{1204
11. Manko, J., Dey, A.K., Hsieh, G., Kientz, J., Lederer, S., Ames, M.: Heuristic
evaluation of ambient displays. In: CHI 2003. (2003)
12. Nielsen, J., Molich, R.: Heuristic evaluation of user interfaces. In: CHI '90: Pro-
ceedings of the SIGCHI conference on Human factors in computing systems, New
York, NY, USA, ACM (1990) 249{256