Augmenting Museums and Art Galleries

Chris Baber, Huw Bristow, Sean-Lee Cheng, Anna Hedley,

Yuri Kuriyama, Marc Lien, James Pollard and Phil Sorrell

School of Elec. & Elec. Eng., The University of Birmingham. UK

c.baber@bham.ac.uk
Abstract: This paper is concerned with the application of context -aware computing to museums and art galleries.
The paper reports three studies addressing this issue. The first study involved a survey of visitors to an art
gallery and focused on visitor activity and information requirements. This led to the conclusion that one could
define visitor types and relate these to the amount, and level of detail, of information for the visitor types. The
second study involves a comparative evaluation of different platforms for augmenting museums and art galleries
through physical hyperspace. The evaluation demonstrates that the approach is viable. The third study
considers different media for presenting information to different user groups. British or Japanese students were
presented with either textual or text + picture presentations of information. Differences in performance across
groups and media were found. Implications for context -sensitive computing in museums and art galleries are
discussed.

Keywords: Augmented Reality, Context -aware computing, Museums and Art Galleries

Typically, labels are positioned adjacent to an

1 Introduction exhibit. A label might contain 50 to 500 words,
printed on a card the size of a postcard.
The nature of museums and art galleries has been Alternatively, an exhibit might contain a large artifact,
changing considerably in recent years. Rather than such as a map, which could have a variety of labels
merely repositories of artifacts (to which members of attached to it. The labels could be printed on paper
the public have limited access), these institutions or could take the form of small lights that are
have been developing into interactive and illuminated when buttons are pushed on a labeled
information-rich environments that allow visitors to panel.
experience the artifacts on display. Museums and art
galleries contain collections of artifacts and
information, presented in different locations and Visual Auditory
using different media. Visitors move through the Superimposed Labels of Audio
museums and art galleries receiving, and searching parts of an commentary on
for, information via these disparate media. The artifact exhibit
relationship between different media and artifacts Adjacent Labels next to Commentary from
could be seen as a (pseudo)-example of ‘augmented artifact ‘audio-guide’
reality’, i.e., information is presented to the visitor to
augment the artifact. Discussion with curators and Table 1: Conventional Augmentation
exhibition designers suggested the classification
scheme for contemporary augmentation of museums
and art galleries in table 1.
Recently, exhibits are being ‘labeled’ through than one hour. Thus, we can distinguish between
auditory information. There might be a button next types of visitor in terms of age group and first-time or
to the exhibit that will cause the replaying on a short repeat visit.
commentary. Alternatively, a visitor might carry a In terms of time spent viewing art works, the data
handset or wear a ‘Walkman’, to hear commentary for are less clear-cut. 16-25 year olds and 26-40 year olds
specific exhibits as they move through the museum. will typically spent 0.5 to 2 minutes viewing an
The handsets, for instance, require the visitor to artwork (around 85% of respondents), while the over-
enter a code number to replay a short commentary for 40 groups exhibit two distinct patterns, i.e., 0.5-1
a specific exhibit. The use of technology in museums minute (around 45% for 41-65 and 55% for >65) and
and art galleries highlights potential applications of more than 2 minutes (around 45% for the 41-65, and
context -aware computing. In this paper, we report 28% for >65). The duration of the visits vary across
the design and prototyping of platforms for context - age groups, as does the time spent viewing an
aware guides. The main aim of the paper is to artwork. This will have implications for the amount of
compare different platforms as the basis for future information that people will want to receive during
development. their visit. Furthermore, 65% of all respondents did
not look at every single artwork. Reasons for
2 User Requirements choosing to look at particular artworks included:
Visual appeal: colour, quality, subject; Familiarity:
The initial stage of this project has involved previous experience of artwork; Genre: favourite /
investigations of the behaviour and requirements of familiar artist, style, era; Curiosity: new exhibit,
visitors to museums and art galleries. This work has unusual features. This suggests that it will be
taken the form of observations of different visitor possible to identify characteristics of artworks and to
groups, e.g., parties of schoolchildren, young adults guide visitors to artworks, which share these
visiting alone or in small groups, elderly visitors, a characteristics. Given that selection of artwork to
short survey over the internet, and an interview with view can be related to particular factors, we were
visitors to a gallery. In this section, some results interested in discovering what sort of information we
from the latter study will be considered. This survey could give to visitors.
represents a preliminary investigation, but the
resulting information has proved very useful. Fifty
four respondents participated in the survey. The
majority of respondents were aged between 16 and Information Content 16-25 26-40 41-65
65, i.e., <16 = four; 16–25 = twelve; 26–40 = fifteen; Basic details 3.9 3.6 4.3
41–65 = sixteen; >65 = seven. The majority of Artist biography 3.0 3.5 3.3
respondents (around 75% of all respondents) Symbolism 3.4 3.2 2.9
indicated an interest in art and art history. While Historical context 3.6 3.5 3.3
most respondents were British (80%), other visitors Motive of artist 3.8 3.4 3.1
were Bengali, Nigerian, Danish, Belgian, Indonesian, Theme of painting 3.7 3.3 3.4
Spanish, Iranian, Australian, and Irish. The majority Technique 3.5 2.8 3.0
of visitors (66%) were in parties of two or more. Of Materials 3.8 2.7 3.0
particular interest, the survey was concerned with Significance of painting 3.8 2.9 3.3
the average duration of a visit and the type of Donor 1.9 2.2 2.0
information that visitors would like to receive Exhibit history 2.5 2.2 2.4
concerning art works.
Respondents spent between 0.5 to 2 hours in Table 2: Responses by Age Group
the art gallery, and this seems to vary with age. The
16-25 year old group were equally divided between
0.5-1 hour and 1-2 hour visits, while the 26-40 and 41- Respondents were asked to rate information on a
65 year olds tended to spend 1-2 hours in their visit, scale of 1 (not interested) to 5 (very interested). The
and the >65 year old group tended to spend 0.5-1 results for three groups are presented in table 2 (the
hour. Furthermore, first time visitors tended to spend <16 and >65 groups have excluded due to small
longer than returning visitors, i.e., around 75% of sample size). Assuming a cut-off of 3.5, i.e., of the
first time visitors spend more than one hour in their maximum, one can see a variation in the type of
visit, while 55% of returning visitors will spend less
information that the different groups wish to see. returned. A personalized web-page is created,
Bearing in mind that the majority of people in the 16- detailing the artefacts that the visitor saw and
25 age group were studying art, it is not surprising to providing further information. The visitor is given a
note that they rated more information as being of card with the URL of this web-page.
interest than the other groups. Furthermore, noting The design concept captures the complete visit
that around half of the 41-65 age group spent less and demonstrates the assumption of minimal
than one minute looking at each artwork, one can dialogue with the technology. This means that the
appreciate that they might not wish to be burdened visitor’s attention can be maintained on the musem
with the task of reading information beyond the basic and its artifacts rather than on the device. At the end
naming of the work and artist. of the visit, a relationship is established between
institution and visitor (through the web-page)
3 Design Concepts through which the visitor can maintain contact, e.g.,
to inquire about artifacts or future exhibitions. In
The project has involved discussion with curators, order to explore the potential designs of an electronic
Public Relations officers, education liaison officers tour-guide, the classification scheme used in table 1
and exhibit designers from four museums (Science was reapplied (see table 3).
Museum, London; Barber Institute, Birmingham;
Birmingham Museum & Art Gallery; Millennium
Point, Birmingham). The work is ongoing and the Visual Auditory
input of these staff is helping to shape the Superimposed Project display Commentary
technology in terms of museum requirements. For onto artifact played when
example, it is important for curators to know how artifact
attractive an exhibit is and the usefulness of the ‘selected’
labels (Boisvert and Slez, 1995). An electronic device Adjacent Handheld Commentary
that can record visitor activity might provide such display linked played at
information. As far as visitors are concerned, the to hypertext location
initial task model assumes that visitors require
instant access to the device (i.e., they will not want Table 3: Computer Augmentation
to engage in a lengthy dialogue to define their ‘visit
type’). Consequently, a set of predefined ‘badges’
will be used for specific types of visit. The ‘badges’ In these concepts, it is assumed that the visitor will
will interact with artifacts via IR or radio-frequency access a hypertext document containing information
tagging. Information will be made available to the on the museum’s artifacts. A head-mounted display
visitor when a ‘badge’ interacts with a transmitter. could be used to project information onto the artifact.
There will be some dialogue between visitor and This is a version what might be termed
device, e.g., to call up floor plans, but the dialogue ‘conventional’ augmented reality and has been
will be minimal. The following scenario has been demonstrated in (Baber et al., 1999). Alternatively,
used to present the overall design concept to visitors the visitor could carry a handheld device that
and museum staff, to a generally positive response: presents text when they approach the artifact. The
When a visitor arrives, they are given a badge. visitor could have an audio commentary that plays
Each badge defines a particular ‘type’ of visit. The when they approach the artifact. This is the basis of
visitor collects an electronic tour-guide and checks the HyperAudio project (Petrelli et al., 1999; Sarini
their location. The visitor is guided around the and Strapparava, 1998; Opperman et al., 1999; Not
museum, when pausing in front of an exhibit the and Zancanaro, 1998). Finally, in a gesture-based
electronic tour-guide presents information about that system, the visitor could point to an artifact and
exhibit. The visitor can choose to receive the whole receive audio commentary. It is assumed that
commentary or move on (in which case the electronic storage of information will mean that one
commentary will stop). The artefacts that the visitor can tailor the information across different media and
chooses will be used to develop a ‘model’ of the visit platforms.
and this will help guide the visitor to similar artefacts.
The tour-guide will also advise the visitor on any 3.1 Prototypes
major or famous artefacts in the collection. After the All of the prototypes followed the same basic
visit, the badge and electronic tour-guide are architecture. The badge contains an IR transmitter
that responds to the IR receiver adjacent to a the evaluation was partly a consideration of the
painting (in a similar arrangement to Not and usability of the prototype devices, and partly a
Zancanaro, 1998). However, rather than signaling consideration of people’s behaviour when using
only the paintings unique ID, we tune the IR receiver these devices to augment their viewing of a painting.
to represent different users (in terms of different Twenty-eight undergraduate students of
styles of visit). Thus, the badge collects the ID of engineering (25 male, 3 female) participated in the
the painting and the ID of the visitor. These data are study for course credit. The mean age of participants
passed, via RS232, to the host computer. This was 23 years and none of the participants felt that
information is used to access a database in order to they were ‘knowledgeable about art’. All
select the appropriate media for the presentation participants indicated that they had previously
about a given painting. In the demonstration system, visited art galleries. Thus, while the sample is
the presentation is written in PowerPoint for the obviously biased, it was felt that participants
laptop and Stylistic. The individual systems are represented a particular ‘type’ of visitor, i.e., non-
briefly described in table three. The ‘pros’ and expert, highly educated. Finally, none of the
‘cons’ were produced following initial user trials. participants had previously seen the painting. There
The prototypes do not correspond directly to the were five conditions (see figure one for the
cells in table three for the simple reason that we technology used):
sought to optimize designs, rather than produce • Control: participants were asked simply to
‘straw-man’ versions for comparison. This meant that look at the painting as if they were visiting
the designs represent combinations and an art gallery;
compromises between the different dimensions. • Guide-book: a paper version of the
However, this is not seen as a weakness in the information was prepared as a four page
development process but a recognition of the booklet;
capabilities of the technology that is used for the • HMD: participants wore the 1st prototype;
platforms. We have also assumed that, where • PDA: participants used the 2nd prototype;
possible, the devices will be multimodal, i.e., will • Tablet: participants used the 3rd prototype.
combine visual with auditory display. The aim in Participants were allocated to condition on
producing concept demonstrators has been to prove presentation (6 each in the PDA, tablet and HMD
the devices rather than develop fully functioning conditions, 5 each in the control and guidebook
systems.. conditions). Participants received a brief introduction
on the aims of the study and a brief explanation of
4 Evaluation of Prototypes the equipment used in each condition (including the
guidebook). Control actions for the devices were
In order to ensure a fair comparison of devices, all kept to a minimum in order not to confound
conditions employed the same painting (Salvador performance. Participants were asked to look at the
Dali, The Temptation of St. Anthony), and, as far as painting until they felt that they were confident that
practicable, the same information. Following the they could answer questions on it. Following the
recommendations of ISO 9241, the evaluation sought question-answering session, participants were
to measure effectiveness, efficiency and attitude (cf. shown a brief video describing all of the prototypes
Frøkjær et al., 2000). Effectiveness was defined in and shown the prototypes they had not used, and
terms of time spent viewing the art-work – the aim of asked to rank the prototypes in order of preference in
the design is to encourage people to spend more time terms of ‘which device would I most like to use on a
interacting with a work of art than they might do visit to an art gallery?’
unaided; Efficiency was defined in terms of people’s
ability to answer questions about the artwork, and in
terms of their ability to recognise elements from the
painting in a subsequent test – the aim was to
determine whether people learnt anything from using
the devices and whether or not use of the device
interfered with their ability to remember parts of the
painting; Attitude was defined in terms of people’s
preference for one device over the others, and was
assessed through a simple ranking exercis e. Thus,
 System Description Pros Cons
A MicroOptical head-mounted Information in same Registration;
display is used to project an visual plane as Can’t share
image onto the painting. An painting; information;
auditory commentary then Highlighting draws Resolution;
describes what is being gaze to location; Binocular
highlighted. Selection of Personal; rivalry
Superimposed information is via badge. User Multimodal;
Visual+Adjacent can interact through buttons. Minimal dialogue;
Auditory

A Philips Nino PDA is used to Unobtrusive; Text only;

display text relating to the Portable; Small display
painting. Selection of Social; space;
information is via badge. User Games and other Separation of
can interact through PDA. User tasks; Limited display and
can play games or perform dialogue art-work
Adjacent learning tasks.
Visual

Adjacent A Fujitsu Stylistic 2300 was Interactive; Lengthy

Visual+Adjacent used to highlight parts of the Portable (heavy?); dialogue;
Auditory painting and provide audio Social; Separation of
commentary. User can play Games and other display and
games or perform learning tasks. tasks; art-work
Multimodal

Figure 1: Examples of Prototype Designs

4.1 Results Efficiency A one-way ANOVA was conducted on

The results will be reported in terms of the three the results of the question-answering test. The
sets of measures employed, followed by a summary results failed to reach significance but a breakdown
table ranking the prototypes on the different of the results is of interest (see figure two). The
measures. labels on the graph correspond to whether the
Effectiveness The average (mean) time spent questions could be answered by looking at the
viewing the painting ranged from 100s for control painting (Qo), by using the commentary (Qc) or by
condition to 230s for the PDA {100s: control – 185s: using both sources (Qb), and whether answers
guide book – 200s: tablet – 215s: HMD - 230s: required recall of elements of the painting (Qp).
PDA}. A one-way analysis of variance (ANOVA) Attitude In terms of the mean rank-ordering of
indicated a significant main effect of device [F94,29) prototypes, participants placed PDA and HMD
= 5.67, p<0.005]. Thus, there was significant equal first, followed by the tablet, followed by the
variation in the time that people viewed the painting guidebook, with control ranked last. Thus, having
under different conditions. The HMD and PDA no information was ranked worse for the question
resulted in the longest viewing times. ‘which device would I most like to use on a visit to
an art gallery?’ Post-hoc justification for ranking by
participants included statements that the PDA and
HMD were light, easy to carry, personal, easy to the painting. This raises the question of how the
move around with and gave a sense of interacting information could be presented.
with the painting.
Overall selection of device Taking all of the results,
the following table ranks the prototypes from best 5 Media and User Groups
(1) to worst (5) for each measure. Taking the overall
The first study compared different platforms for
results, it is possible to produce a final rank-order
presenting information about paintings. It was clear
(F) shown in the last column. This allows the final
that the HMD and PDA were preferred by
ranking of devices to be derived.
participants. It was not clear as to whether the
The results suggest that people in the control
differences were due to technology or to the
condition fared poorly on questions requiring
information presented on the devices. A second
access to the paintings description (D); this is
study was conducted in order to consider whether
obviously because these participants were not
users performed more efficiently with text only or
presented with this information and had no prior
with combined text and images for physical
experience of the painting. More interestingly,
hyperspace. In this study, participants used the
people in the HMD and tablet conditions fared
HMD and viewed either a text only description that
poorly on questions requiring observation of the
was superimposed over the top of the painting, or a
painting (O), and those using HMD also
combination of graphical highlighting with text. In
performance poorly on recall of elements (P). This
all slides, the font used was 54 point. Also, the
might be because the presentation of information
problems relating to observation in study one could
guided people around the painting (and minimized
be related to obscuring of the artwork. In this
the need for participants to ‘search’ the painting
study, the presentation was designed so that the
themselves). Finally, comparison of conditions that
presentation slides were interspersed with clear
used textual display of commentary (guidebook and
slides. This meant that participants were able to
PDA) with those using audio display of
view the information and then have an
commentary (HMD and tablet) does not reveal any
unobstructed view of the painting. The study also
obvious advantage for one medium over another,
provided an opportunity to consider differences
on questions relating to the commentary (C).
between English and Japanese script for the
presentations.

8
O D
6
4 B C Device Time O D B C P F

2 PDA 1 2 1 1 1 1 1

0 HMD 1 4 2 2 5 1 2
Tablet 3 5 4 5 4 3 3
id o l
k
D
A
et
oo
M
PD
gu n t r

bl
eb

Guide 4 3 3 4 3 4 4
ta
H
co

Control 5 1 5 3 2 5 5

Figure 2: Performance on recall tests

Table 4: Final Ranking of Prototypes

The study indicates that performance is affected by

Eighteen students participated in this study. Nine
the device used. In broad terms, the PDA tended to
of the participants were British (6 male; 3 female)
out-perform the other prototypes. The use of the
and nine were Japanese (5 male; 4 female).
HMD (and also the tablet) impaired the participants’
Participants were assigned to either text only or to
ability to recall painting specific information, i.e., Qo
graphics + text groups. Assignment was conducted
and Qp. It is possible that the HMD and tablet led
on order of appearance to the experimental session,
to similar problems of highlighting parts of the
with people being assigned to alternate groups.
painting leading to the participants’ attention being
Participants were presented with a projected image
drawn to these areas and away from other parts of
of a painting (Ford Madox Brown, Work). All
participants were seated 4m from the painting and
fitted with a Seattle Sight monocular, monochrome Information recall Picture recall
head-mounted display. The HMD was adjusted Picture + text 5.3 / 10 2.1 / 5
until the participant was comfortable and could see Text only 5.3 / 10 2.6 / 5
the whole of the image from the headset projected
onto the painting. The presentation consisted of Table 7: Performance by Medium
around a dozen slides. Participants were able to
advance the slides by pressing the spacebar. Three
dependent variables were used: i. time on task; ii. Rather than comparing across two variables, table
ability to answer questions; iii. preference for seven shows the recall performance for all
presentation media. participants using either presentation medium. It is
The initial comparison of time on task (see table interesting to note that, despite having more textual
five) shows differences on both independent information, the text only condition does not yield
variables. demonstrably superior performance on the
information recall task. It is also interesting to note
a slight deterioration in picture recall performance in
Picture + text Text only the picture + text condition. This can be related to
British 105.6s 126.8s the poor performance on the recall task in study
Japanese 75.8s 121.0s two, and suggests that the user of highlighting
attracts the user’s attention to specific parts of the
Table 5: Mean time on task painting (at the expense of other parts). However,
the results are superior to those found in study two
and it is proposed that the reason why the effect is
British and Japanese participants spent a similar less pronounced in this study was because
length of time viewing the text only presentations, participants were repeatedly shown the complete
and took less time when using picture + text (with picture during their session.
Japanese participants in this conditions taking the Participants rated the devices on a scale of 1
least amount of time). In terms of ability to answer (poor) to 5 (good). As previously, we assume a cut-
questions, tables 6 and seven reveal some off of maximum rating. All participants felt that
differences between groups. the font size was good (i.e., a mean rating of 4.1 / 5),
and that the order of presentation of information
was good (4 / 5). The information presented was
Information recall Picture recall felt to be useful (3.7 / 5) and the interaction was
British 6 / 10 2.9 / 5 entertaining (3.6 / 5). While the information was
Japanese 4.7 / 10 1.9 / 5 deemed useful (3.5 / 5), there was some concern
over the technology used, i.e., the headset was
Table 6: Recall Performance by Group uncomfortable (1.7 / 5) and this made the ‘system’
difficult to use (2.6 / 5 for usability), and also made it
difficult to find parts of the painting (1.7 / 5).
In these tests, British participants scored higher
than their Japanese counterparts. Post-test 6 Discussion
debriefings suggested that there were two main
reasons for these differences: i. the British In this paper, we have presented three studies
participants could readily draw on previous exploring aspects of augmenting museums and art
knowledge of English Victorian society; and ii. the galleries. The first study supports the assertion that
British participants were able to identify elements in ‘one size will not fit all’, and emphasizes the need
the painting on the basis of this prior knowledge. for context -adaptive technology to suit this domain.
The second study suggests that some form of
context -aware computing might be a useful means
of augmenting art galleries. In the third study, two
issues raises in the first two studies were explored:
medium of information presentation, and differences
in user groups. Information was presented in either relationship between exhibit characteristics
text only or picture + text to British or Japanese and learning-associated behaviours in a
participants. While there was little effect of science museum discovery space, Science
presentation medium on recall of information, there Education, 79
was some impact on recall of pictures by the
overlaying of picture + text onto the artwork (see Frøkjær, E., Hertzum, M. and Hornbæk, K.,
also study two). Furthermore, the Japanese 2000, Measuring usability: are effectiveness,
participants tended to under-perform on the recall efficiency and satisfaction really correlated?,
tests (despite having all information presented in CHI’2000, New York: ACM, 345-352
Japanese). This was explained by differences in
cultural experience, i.e., knowledge of Victorian Not, E. and Zancanaro, M., 1998, Content
English society. Consequently, the different user adaptation for audio-based hypertexts in
groups required different ‘background’ information. physical environments, Proceedings of the
In conclusion, this paper has presented our Workshop on Adaptive Hypertext and
initial work into the augmentation of art galleries Hypermedia, 20-24 June, Pittsburgh, PA
through electronic media. Our primary interest is in
the provision of visual support that can be used to Opperman, R., Specht, M and Jaceniak, I., 1999,
interact with paintings. Hippie: a nomadic information system, In H-
W. Gellerson (ed) Handheld and Ubiquitous
Acknowledgements Computing, Berlin: Springer-Verlag, 330-
333
We are grateful to Jane Arthur and Simon Redgrave
for their assistance in conducting stud one, and to Petrelli, D., de Angeli, A. and Convertino, G.,
all of the museum staff who have helped with the 1999, A user-centred approach to user
project. modeling, In J. Kay (ed.) UM99 User
Modelling, New York: Springer Verlag
References
Baber, C., Harris, T. and Harrison, B., 1999, Sarini, M. and Strapparava, C., 1998, Building a
Demonstrating the concept of physical user model for a museum exploration and
hyperspace in an art gallery, In S. Brewster, information-providing adaptive system,
A. Cawsey and G. Cockton (eds.) Human Proceedings of the Workshop on Adaptive
Computer Interaction INTERACT’99 (volume Hypertext and Hypermedia, 20-24 June,
II), Swindon: British Computer Society Pittsburgh, PA

Boisvert, D.L. and Slez, B.J., 1995, The