From informing to remembering: ubiquitous systems in ...km2/projs/pervasive/... · The Exploratorium In most of the projects described earlier, the devices were deployed at museums

1536-1268/02/$17.00 © 2002 IEEE PERVASIVEcomputing 13

I N T E G R A T E D E N V I R O N M E N T S

From Informing to Remembering:

Ubiquitous Systems inInteractive Museums

Thanks to their enclosed nature and well-defined role, museums are a fertile groundfor studying visitor behavior and envi-sioning systems to enhance the visitor’sexperience.1 Early electronic guidebook

tools in the form of acoustic guides have been successfullydeployed in museums for some time. However, their appli-cation is limited because they can inform only by describ-ing; they require the user to make his or her associationswith an exhibit with no visual cues to assist. More recentmuseum guidebook systems have tried to improve on thisexperience with multimedia content.

The Exploratorium, an interactive science museum inSan Francisco (www.explorato-rium.edu), is a challenge because itsphysical environment alreadyrequires much of the user’s personalresources (eyes, hands, and mentalattention). A successful interactivetool must, therefore, provide a valu-

able service while making few demands on these resources. This article presents our experiences designing, imple-

menting, and deploying applications to enhance visits toan interactive museum. We conducted this research verymuch in collaboration with the Exploratorium and itsstaff. Our work is part of the Cooltown project,2,3 whichresearches infrastructure and applications for nomadiccomputing systems—ubiquitous systems in which peoplemove about while using portable devices to access ser-vices and applications integrated with the physical world.Our museum study has focused our attention on nomadiccomputing tools: applications that enhance users’ inter-

actions with the physical world but whose functionalityis sufficiently simple to make them largely transparent.

Museums as test bedsWith the introduction of inexpensive portable devices

that can render multimedia content, several projects havedeveloped prototype or commercial systems for museumaugmentation or navigation with handheld and wirelesstechnologies (see www.exploratorium.edu/guidebook).Technologies such as PDAs and pagers have provided ameans by which museums can communicate with theirvisitors in a more personal manner. The San FranciscoMuseum of Modern Art, for example, uses static and localcontent on a PDA, including a short video clip of the artistexplaining his or her work, to provide visitors with a dif-ferent perspective about the exhibits. The ExperienceMusic Project (see www.emplive.com) in Seattle usesinfrared beacons at which the user points the device toselect relevant content.

The Sotto Voce electronic guidebook, tested at the Filolimansion,4,5 also provides content about exhibits on ahandheld device. Users see images of the rooms in the his-toric house on their PDAs. They thus orient themselvesin both the physical and virtual world as they locate rel-evant content. Each image shows artifacts against a room’sphysical wall. The visitor can then select artifacts fromthe virtual image and learn more about them, principallyin the form of audio content. Visitors can also share con-tent in the relatively quiet environment, which plays a pos-itive role in social interactions.

Portable devices can also effectively deliver informa-tion to visitors in outdoor tourist locations. Systems such

Museums are excellent locations for testing ubiquitous systems; theExploratorium in San Francisco offers a unique and challengingenvironment for just such a system. An important design consideration ishow users switch between virtual and physical interactions.


Margaret Fleck, Marcos Frid, TimKindberg, Eamonn O’Brien-Strain,Rakhi Rajani, and Mirjana SpasojevicHewlett-Packard Laboratories

as Guide, for example, can replace the tradi-tional tourist guidebooks and maps.6 Onceagain, though, these electronic guidebooks aredesigned to support users who observe andlearn about artifacts rather than interact withthem. They are a closely comparable substitutefor the traditional paper guidebook, whichtourists often carry and consult during visits,switching their attention between the guide-book and the objects of interest. A major advan-tage of the electronic version is that it provideslocation-dependent information.

The ExploratoriumIn most of the projects described earlier, the

devices were deployed at museums where thereis not much hands-on activity allowed. In fact,in many situations, the visitor cannot touch theexhibits. The Exploratorium is a very differentenvironment.

The Exploratorium is a large, open spacepopulated with hands-on science exhibits (seeFigure 1). The building’s free-form, workshop-like structure contains several hundred exhibits,which frequently rotate on and off the floor.Users of all ages and levels of scientific knowl-edge roam from exhibit to exhibit, manipulat-ing the apparatuses, reading informationmounted on labels, and occasionally consult-ing “explainers”—museum guides who explainthe exhibits’ operation and the scientific phe-nomena behind them.

Although audio guides are available toExploratorium visitors, a very small percentage

of visitors use them. The Exploratorium is thuspromising ground for nomadic computing toolsbecause it is an environment rich in physicalartifacts and is explored by mobile users. Elec-tronic guidebooks are an example of tools thatwork well in art museums, historic buildings,and historic towns (see www.exploratorium.edu/guidebook/forum/report/index.html).4,6,7

However, those environments often do notenable interaction with exhibits and are some-times less boisterous and more clearly organizedthan the Exploratorium.

The Exploratorium presents a substantialchallenge to creating a nomadic computinginfrastructure. Its size and tendency to changewould present a huge problem for museum-wide, long-term deployment. Our resources lim-ited us to temporary deployment for up toroughly 15 exhibits and 10 simultaneous users.However, certain characteristics complicatedeployment, whatever the scale. Specifically, theExploratorium is

• Hands-on. The exhibits invite physical inter-action, which might involve, for example,two-handed manipulation of a mechanismor putting your head inside an echoing tube.Some exhibits contain materials hostile toelectronic devices and wireless networking,such as sand, water, and large metal objects.

• Boisterous. The Exploratorium is popularwith children, who can be uninhibited intheir urge to race between exhibits andexclaim discoveries to one another. (Adults

are also known to get excited.) The combi-nation of voices, mechanical noise, andechoes creates an extremely high level ofbackground noise.

• Hard to navigate and identify. The exhibitsare sometimes clustered by topic but have noobvious order. There are few distinctiveareas, landmarks, or backgrounds in theopen-plan layout to aid in locating them.Identifying the exhibits without close inspec-tion is hard because they tend to be unfa-miliar and could be obscured by other visi-tors and exhibits.

Our main goal on this project was to under-stand the strengths and limitations of nomadiccomputing technologies while enhancing the vis-itor’s experience in this demanding environment.We could deploy a wide range of nomadic com-puting tools, but there seems to be only a smalllist of basic functionalities to consider:

• The Informer provides the user with infor-mation related to the exhibits, particularlythe exhibit that the user is currently visiting.This information is more detailed than thatwhich is on the physical exhibit labels or istailored to the user.

• Suggester offers ideas for what to try at anexhibit—for example, “Try moving the leverto see the effect on the waveform.” This sup-plements the short list of suggestions on theexhibit’s physical label.

• The Exploratorium’s open organization doesnot indicate any order for visiting theexhibits. For those who would like morestructure, Guider suggests exhibits to visitnext, along with navigational advice.6

Exhibit order might be fixed or could adaptto a user’s interests and past behavior.

• There is often a strong social dimension tovisiting museums.1,5 Communicator helpsusers communicate in the context of theexhibits. Communication might include elec-tronic bulletin boards for individual exhibits,instant messaging, or information beamedbetween handheld devices.

• The vast amount of information presentedoften overwhelms Exploratorium visitors.

14 PERVASIVEcomputing http://computer.org/pervasive


Figure 1. A corner of the Exploratorium.

Rememberer helps them build a record oftheir experiences, which they can consultduring or after their visit. The user deliber-ately selects the phenomena to record, in con-trast to passive recording systems such asForget-Me-Not.8

After discussions with staff at the Explorato-rium, we implemented a hybrid implementationof Informer, Suggester, and Rememberer skewedtoward real-time information provision (ourelectronic guidebook). Studies of museum visi-tors and volunteers using the guidebook, as wellas observations of regular visitors without anytechnology, however, led us also to consider apure Rememberer tool, which postpones muchof the visitors’ virtual interactions until aftertheir physical interactions with the exhibits.

A study of visitors withouttechnology

We carried out an informal study of howmuseum visitors ordinarily act—that is, with-out any extra technology—at exhibits. Weobserved 30 individuals and groups, randomlyselected from the visitors using one of fourexhibits, and briefly interviewed them immedi-ately after they finished.

Our goal was to broadly characterize theirinteractions. For various practical reasons, thestudy took place in parallel with the deploy-ment of our initial prototype instead of beforeit. By itself, the study did not help us reach anyfirm design implications. However, it helped usunderstand our design opportunities and con-straints in terms of typical user interactions withexhibits, labels, and other users.

We took away from our study a constraint, anobservation, and a question. The constraint wasthat explorative play and discussion among vis-itors are key characteristics of the Explorato-rium and that our tools should not interfere withthem. The observation was that information hasa role, but, for many visitors, the informationon the labels was secondary to the experience.The question was, could we improve access toinformation without breaking our constraint?

ExplorationThe most common approach to an exhibit is

to walk up to it and figure out what to do withit. A quantitative study (described later) showedthat the typical visit lasts one to two minutes,

until the visitors decide that they have “done”the exhibit and move on. However, that shorttime is a period of opportunity in which userssometimes become more engaged and extendtheir visits for up to 10 to 15 minutes.

The exhibits support a combination of playand science; many visitors, especially children,are creative in the way they use exhibits. Forexample, two girls invented a sophisticatedgame involving catching sand (using an exhibitintended to illustrate exponential decay) thathad no connection to the “script” on the exhibitlabel. Such play can lead to science: one adultmentioned that interacting with an exhibithelped him finally understand a scientific con-cept encountered 20 years earlier at school.

The role of the labelsVisitors often read the exhibit label only

when they have finished or if their attempt tofigure out the exhibit fails. Most visitors saidthat the exhibit itself was explicit enough andthat the information on the label was a bonus.Some adult visitors spent a fair amount of timereading the labels and discussing the sciencebehind the exhibits. However, a striking obser-vation was that visitors spent more time at theexhibit when they did not read the label.Bypassing the label seemed to correlate withplay and exploration.

Social behaviorWhen a pair or larger group of people visit

an exhibit, often one person will read the instruc-tions while the other “does” the exhibit. Peopleseem to enjoy helping each other and discussingthe exhibits, and this seemed to encourage addi-tional interaction with the exhibits.

A physical experienceWhen children see an exhibit they like, they

often just throw to the ground whatever theyare carrying and start playing with the exhibit.This shows why any electronic tools must berobust and also suggests a tendency to react

nonverbally to the exhibits. Although adults areoften more circumspect (they sometimes walkon the periphery of an exhibit before decidingwhether to interact with it), they also seem toreact spatially rather than verbally. They referto exhibits by location (“that one over there”),rather than by name or description.

The electronic guidebookIn parallel with the no-technology study, we

deployed and tested a prototype electronicguidebook. Our original goal was to discoverhow to add value for visitors by integrating elec-tronic resources with the physical exhibits and,in particular, to test technologies developed bythe Cooltown project for associating exhibitswith Web resources.

The prototypeOur guidebook prototype combined the

functions of informing, suggesting, and remem-bering (the latter because users sometimes jotmarginal notes in conventional guidebooks).The three functions were to be delivered prin-cipally in the form of Web pages, which userscan access on PDAs while at the exhibits andbefore and after the visit on any computer.

The guidebook prototype provides contentabout the exhibits to users as they visit them; italso provides them with a record in the form of

a personal scrapbook on the Web, for perusalafter the visit. We used only six exhibits due toequipment limitations and the effort requiredto develop content for each exhibit.

Supplementary exhibit content. The guide-book delivers static Web pages to users’ PDAswhen they visit exhibits. Each exhibit has ahomepage that contains links to four to 16pages, which contain information about theexhibit and the underlying phenomena and sug-gestions about what to do with the exhibit. Fig-ure 2 shows the homepages for two exhibitsand a map of exhibit locations.

APRIL–JUNE 2002 PERVASIVEcomputing 15

People seem to enjoy helping each other and

discussing the exhibits, and this seemed to

encourage additional interaction with the exhibits.

We developed audio content for severalexhibits, but users could not hear it withoutheadphones because the Exploratorium is verynoisy. We also believe that headphones wouldinhibit social interaction.

A personal scrapbook. The guidebook letsusers bookmark pages. This creates a personalscrapbook page that has links to the selectedpages, which users can access from anywhereon the Web. (Users are assigned pseudonyms toprotect privacy.)

The prototype system (see Figure 3) con-sists of PDAs, pi-stations (“point of infor-

mation” stations), and a content server, allconnected by an IEEE 802.11 wireless net-work.

PDAs. Most of our tests used HP Jornada 690s,a clamshell with a keyboard and a color screenapproximately 15 cm × 5.5 cm and 640 × 240pixels. We also used a Hitachi ePlate, anotherWindows CE device, without a keyboard butwith about twice as large a screen. Visitors inter-acted with a Web browser simplified to haveonly four control buttons: forward, back, home,and bookmark. A stylus selects buttons and fol-lows hyperlinks.

Pi-stations. A pi-station is a conspicuous,portable, freestanding unit that can be config-ured for and placed next to any exhibit, housingidentification technologies and sometimes cam-eras and placing them at a convenient height.

Content server. Exhibit content is static,served by a standard Apache Web server. TheCooltown Web Presence Manager handles eachuser’s personal information (for example, scrap-book entries).9 Pseudonyms are assigned to pro-tect the user’s privacy. All requests from PDAsare sent via an HTTP proxy that logs all activ-ity, indexed by the user’s pseudonym.

Physical hyperlinksA key design aspect of our guidebook is the

mechanism by which users obtain a givenexhibit’s homepage. We used the standardapproach in our Cooltown research—the userpicks up an identifier by pointing a handhelddevice at the physical object of interest. Attach-ing an identifier on or near the object andinstalling a sensor in the PDA makes this work.Alternatively, the identifier token can be givento the user and the sensor attached to the object.

Once the identifier is picked up, it is con-verted to a URL (if it isn’t one already) using aresolution service.10 The URL is dereferenced,and the resulting Web resource (the physicalobject’s Web presence) is rendered on the PDA’sWeb browser. We call our identification tech-nology physical hyperlinks because, from theuser’s viewpoint, the process is similar to click-ing on a hyperlink in a Web document; readingan identifier with a handheld sensor replacesclicking with a mouse.

Alternatively, users could be asked to find



Figure 2. Examples of content a visitorwould see in a modified Web browser.One exhibit page (a) shows detailed,largely graphical content developed byExploratorium staff; we developed theother exhibit’s page (b) with more basic,predominantly textual content, to compare user reactions to the differentmedia types. Visitors could also accessthe guidebook homepage, which gives(c) a map of exhibit locations.

(c)

(b)

(a)

the exhibit’s Web page using conventional Webnavigation or manually enter an exhibit num-ber (as in audio guides). However, we believethat these methods would be more awkwardfor the user than the single, coarse-grainedactions of reading a beacon or barcode. Theimage-based method that the Sotto Voce guide-book offers for selecting information in a his-toric house also seems unsuitable for theExploratorium because its exhibits are hard toidentify visually and its open-plan layout doesnot offer canonical 2D views.

Within the physical hyperlink paradigm, wewanted to experiment with different options foridentification hardware. Therefore, we equippedthe pi-stations with three types of identificationtechnology: an infrared beacon, a barcode, anda radio frequency identification (RFID) reader.The infrared ports on our Jornada PDAs canpick up the URLs transmitted by the beacon,usually within a cone of a few tens of degreesand up to about two meters away. A laser scan-ner on the Hitachi ePlate can pick up the bar-codes; a resolution service then converts themto URLs. Several other types of identificationtechnology exist.11

In choosing an identification technology, amajor issue is scaling up to large numbers ofusers and exhibits. Barcodes are much cheaperper exhibit than beacons. However, barcodereaders for PDAs are still relatively expensive

whereas most PDAs have integrated infraredreceivers. RFID tags are closer to barcodes thanbeacons in price, but these readers are relativelyexpensive and can be cumbersome (dependingon their range).

Testing the electronic guidebookWe carried out several rounds of informal

studies of users with the electronic guidebook toget a general sense of how users would react toit. We wanted to understand whether the toolbroke the constraint of not interfering withexploration and whether visitors perceived anyvalue in this type of tool above the informationthat labels offered them. At a more basic level,another goal was to uncover major usabilityand system issues with the core Cooltown tech-nologies of physical hyperlinks to Web pagesaccessed through PDAs.

The experimental setupTo reflect the diversity of Exploratorium vis-

itors, we recruited several users, including thosewith and without prior computer experienceand those who had not visited the Explorato-rium before. The 35 users consisted of 16 adultfemales, nine adult males, and eight children(aged 10 to 13; one female).

The adults ranged in age from about 25 toabout 50. Twelve were teachers; most of theothers came from backgrounds that involved

museums or computers. Some users were insmall groups (two to three) sharing a PDA,including two family groups each with two chil-dren. All users were fluent English speakers withno major disabilities.

We did most of the tests under the Explorato-rium’s normal conditions: noisy, with manypeople milling around. One subject used aHitachi ePlate; the others used Jornada 690s.

The tests began with a brief demonstration ofa pi-station and how the PDA works, includ-ing how to pick up a beacon (or barcode for theePlate). The users were then directed to a gen-eral area of six instrumented exhibits. Theinstrumented exhibits were in a reasonablywell-defined section of the Exploratorium butinterleaved with other exhibits. Orange flagsmade the pi-stations more conspicuous. Theusers interacted with the instrumented exhibitsand sometimes with other nearby exhibits.

A project member, who observed the users’actions and reactions and helped them recoverfrom major problems, shadowed each user orsmall group of users who had a PDA and heldsemistructured interviews afterward. All Webaccesses from the PDAs were logged automatically.


802.11

Pi-station: beacon (infrared), barcodeRFID plus processing, I/O, and networking

Internet

RFID

Exhibit

Viewed at home or school

Contentserver

URLover

infrared

π

Figure 3. The guidebook systemoverview.

The findingsDue to the relatively small number and self-

selectivity of subjects, our findings represent“existence proofs” for certain reactions andphenomena rather than statistical evidence, sowe’ll summarize the proportions of usersinvolved instead of giving numbers or demo-graphic breakdowns.

There is an ongoing study we’re conductingthat focuses on the role and form of contentdesigned for special groups involved in self-guided learning activities. However, our initial

findings gave us an indication of the usabilityissues for the general visitor population andmotivated us to reconsider the tool that we hadimplemented.

Overall positive response. Most usersreacted positively to the electronic guidebookoverall. Some appreciated the ability to obtainmore information about the exhibits. Severaltried the suggested exhibit activities and likedhaving more options than those on the exhibitlabels. Some thought the ability to look backon “remembered” exhibits on the personalscrapbook would be useful—especially in theclassroom. Overall, users found the prototypestimulating and made suggestions, for exam-ple, about the ability to input comments intothe pages.

Not enough hands. The PDAs tended tointerfere with hands-on activities at the exhibits.They are too big for a pocket and tend to swingawkwardly on a neck strap. There is obviouslynowhere safe to put them down, so some userstried to manipulate the exhibit with one handwhile holding the PDA with the other.

Demands on the user’s attention. Someteachers were keen to explore the content—par-ticularly the suggestions—at the exhibits. Butfor many of the users overall, the informationand suggestions were extraneous, and the PDA

represented an undesirable demand on theirattention.

Lost in hyperreality. Each exhibit’s contentcontained hyperlinks to other, related exhibits.One user followed such a hyperlink withoutrealizing that she had done so and ended up try-ing to follow a suggestion intended for anexhibit that she was not observing. Some of thechildren became more focused on the content(and the PDA itself) than the exhibit, reversingthe behavior we saw without the PDA.

The wow factor. Part of the overall positivereaction was clearly due to the hardware’s nov-elty, rather than the system’s functionality. Con-trolling for this wow factor is a problem in test-ing most (all?) nomadic computing systems.

Beacons are OK, but.... Most users quicklygrasped the notion of receiving a beacon viatheir PDA’s infrared port to view the exhibit’spages and had little difficulty physically man-aging beacon reception. The one ePlate userwith a barcode scanner also seemed to get alongwell, although her PDA proved uncomfortablyheavy. Users occasionally picked up beaconsaccidentally, triggering a disruptive change inthe page the browser displayed. The mostcommon case involved accidentally reading abeacon soon after reading it deliberately. There-fore, for repeated pickup of the same beacon,we modified our software to ignore the secondpickup or query the user before changing pages,depending on the interval between pickups.Requiring users to press a button when pickingup a beacon might help, but our PDAs didn’thave an ergonomically suitable button thatusers could find and press accurately.

Browser interface. Some users unfamiliarwith PDAs had trouble manipulating the stylusto click on a hyperlink—some tried scratchingon the link rather than tapping it. Such usersmight need brief hands-on training.

Content design. Content style varied acrossthe exhibits. Most users correctly recognizedunderlined text as hyperlinks. However, in heav-ily graphical content, they seemed to have trou-ble guessing which items they could click. Explicitmarking (outlining) of graphical hyperlinksmight help,4 as would consistent content design.

Forgetting to remember. Despite statementsof enthusiasm about the potential uses of thepersonal scrapbook, few users pressed the but-ton on their PDA to add pages to it.

AnalysisThe studies of Exploratorium users without

the electronic guidebook and with it led us tothe following considerations:

• Interference with exploration. PDAs are toolarge and fragile to be convenient for userswho want to experiment with the exhibits,given the environment’s relative boisterousness.

• Value. For some, the guidebook was mostlya distraction, at least at the time of the phys-ical visit to the exhibit; others liked the ideaof being able to explore an already-familiarexhibit more deeply through Web pagesdelivered in real time.

• Physical hyperlinks. Cooltown physical hyper-links are an effective mechanism for invokingservices in places such as the Exploratoriumwhere navigation and identification are diffi-cult, although some details (for example, acci-dental beacon pickup) need improvement.The phenomenon of being “lost in hyperreal-ity” raises interesting research issues.

• Application complexity. The combination ofseveral functions (Informer, Suggester, andRememberer) was too complex. Users weretoo busy with the first two functionalities touse Rememberer.

We thus decided to design a simpler system,concentrating on one basic functionality.

One option would be to implement a pureversion of Informer or Suggester, but in a lessobtrusive form factor. For example, contentcould in principle be displayed on screensmounted on or near exhibits, with users carry-ing at most a simple electronic token to activatethem rather than a PDA. However, this was notan option at the Exploratorium, where policy isto avoid placing distracting screens next to



Rememberer is intended to aid personal recall

and stimulate discussions and other forms of

social interaction.

exhibits. Kiosks, away from exhibits, remainan option to explore. Another alternativewould be to deliver content on a small-screenwearable device that users could safely ignoreduring exhibit manipulation. Aside from thelogistics of producing such a device, we wereconcerned that this would restrict the contentto small amounts of text.

However, we considered that the researchquestion, “Which functionality could betterprovide value, taking the environmental prop-erties fully into account?” was likely to raisebroader issues than how to implementInformer or Suggester despite the hostile envi-ronmental characteristics. Our initial studiessuggest that a critical resource is the user’s atten-tion. Consulting the PDA requires the user torepeatedly shift attention between the virtualworld of content and the physical world of theexhibit, the user’s companions, and the sur-rounding environment. The transitions requirephysical manipulation (removing the PDA froma belt clip or pouch), shifting visual attentionand reestablishing visual context, and cognitiveeffort.

This analysis would suggest that physical–virtual transitions would be less distracting ifthey occurred less often or between exhibits,when users would already be transitioningbetween different physical objects, a likely usagepattern for Guider and Communicator.Rememberer should dramatically reduce thenumber of transitions because users can main-tain their attention on the physical world whilevisiting exhibits—except when taking actionsnecessary to record phenomena—and cross tothe virtual world when consulting theirrecorded information at home or at a museumkiosk. We chose to concentrate on Remembererfor this reason and because our guidebook usersexpressed a strong interest in it.

On to RemembererRememberer is intended to aid personal

recall, stimulate discussions and other forms of

social interaction, and support the user’sresearch or classroom work. It consists of

• A “remember-this” technology with whichthe user selects objects during the visit

• The visit record, consisting of a set of Webpages

• A physical artifact that reminds the user ofthe visit and contains a pointer (URL) to thevisit record (see Figure 4)

Because the remember-this technology per-forms a simple task, its handheld unit can be cor-respondingly simple and small. For our initialtests, we used RFID tags (some credit cardshaped and some mounted in watches). Bring-ing the tag within 10 cm of the pi-station’s reader(“swiping”) registers the exhibit under the user’spseudonym and briefly lights up an LED.

We also used a Jornada 567 as a remember-this tool. We enclosed it in a case for protectionand to prevent use of the screen. Pointing thedevice at a beacon registered the exhibit. ThePDA is much larger and heavier than the RFIDtags. However, this implementation choice letus scale our system to a higher number (15) ofexhibits because we could equip some exhibitswith only a beacon instead of an RFID readerand computer.

The visit record contains a list of exhibitnames in the order visited. In addition, we

included pointers to detailed content for eachexhibit and a field for users to record comments.Moreover, to make the record more specific tothe user’s personal experiences of the exhibits,we also equipped some exhibits with cameras.

Figure 5 shows a page created at the “Spin-ning Blackboard” exhibit. When a user regis-tered the exhibit with his or her remember-thisdevice, four photographs were taken at 1-sec-ond intervals. Cameras were positioned to takepictures of users at exhibits or phenomena thatusers create at the exhibit. The pictures were notdisplayed at the exhibit: users saw them onlylater when inspecting their visit records.

Before starting, users visit a special basesta-tion exhibit. This assigns them a pseudonymand creates and displays the beginning of theirvisit record, including a picture of the user vis-iting the basestation. This gives users briefhands-on practice with the system (includingthe remember-this technology) and introducesthem to what artifact the system is generatingduring the visit.

For our initial tests, we observed and infor-mally interviewed 14 adults, all but one ofwhom were employees or volunteers at theExploratorium. Six visited the exhibits alone,and eight wandered in groups of two to three.One group shared an RFID tag but the rest ofthe users had their own tags. Users viewed theirvisit records at the end of the visit. A laptop was


Figure 4. A reminder refrigerator magnet (top left) and remember-thistechnologies: a radio frequency identification card and wristwatch and a PDA that receives and invokes beaconed URLs.

also available for viewing records during thevisit, but only some chose this option; thosewho did not still swiped their tags at theexhibits.

In the later tests with a PDA as a remember-this device, we observed 23 individuals andfamily groups (of up to four members) solicitedthrough the Exploratorium membership pro-gram. Of these, two individuals and four groupsrepresented a control group (used no technol-ogy). Each of the other groups shared a PDA.

Users reacted positively to the system over-all. They were especially stimulated by the pho-tographs, but many of them mentioned that theimages were often blurred and of low resolu-tion. They also wanted more control over theposition and timing of pictures—the only feed-back was the LED on the PDA or RFID reader,which lit at about the time of the first photo-graph in the sequence of four. Users almostalways registered the exhibits, sometimes sev-

eral times to capture particular phenomena. Inthe case of the RFID tags, they did so casuallybut accurately, with no indication that this dis-turbed their engagement with the exhibit ortheir companions.

The PDAs required more practice and con-centration from the users. As expected, theirlarger size made them harder to carry and moreof a hindrance to manipulating the exhibit.

We gave the users the URLs of their pagesand logged visits after they left. Most revisitedthe pages, some several weeks after the event(10 out of 17 individuals or groups from thesecond round of tests). Several of them savedcomments within their pages, referring to thephotographs. At least one user emailed hispage’s URL to relatives. All this suggests thatRememberer could have value for personal andsocial uses.

We have preliminary evidence that both pro-totyped remember-this devices distracted our

users much less than our guidebook tool. In par-ticular, the RFID cards were unobtrusive. Webelieve that we could modify our beacon designto produce a remember-this device as unobtru-sive as the RFID tags, while enabling us to equipmany exhibits just with beacons.

A ll of us will eventually need no-madic computing tools for a widerange of environments. Our initialelectronic guidebook seemed ap-

propriate for particular classes of users, such asteachers or explainers, who wanted to gobeyond their familiarity with the exhibits. How-ever, our studies also provide preliminary evi-dence that bookmarking of physical objects,together with a basic photographic capability,might be sufficient to provide a valuable serviceto other nomadic users. Simple tools such asRememberer seem worthy of further researchfor other environments that place high demandson a user (for example, shopping with childrenin tow).

Finally, this study illustrates the value ofincorporating prototype deployment and usertesting in the early stages of developing anomadic computing tool, despite the sub-stantial investment of time required to do so.Demos inside HP Labs gave us little prepa-ration for the conditions we found inside theExploratorium.



Figure 5. A Rememberer page showing auser’s visit to the “Spinning Blackboard”exhibit.

ACKNOWLEDGMENTSWe thank the rest of the team (both HP and Explorato-rium staff) who helped design and implement theseexperiments: Katherina Audley, Wesley Chan, DougConaway, Philippe Debaty, Caroline Gattein, JoshuaGutwill-Wise, Ron Hipschmann, Sherry Hsi, FranciMcFarland, Mike Petrich, Natalie Rusk, Rob Semper,Larry Shaw, Quan Tran, Adrian Van Allen, Noel Wan-ner, and Karen Wilkinson. The Exploratorium’s work onthe electronic guidebook was funded by NSF grantESI-9901985. We also thank Nigel Davies and theanonymous reviewers for their comments on earlierdrafts of this article.

REFERENCES1. D. vom Lehn, C. Heath, and J. Hindmarsh,

“Conduct and Collaboration in Museums andGalleries,” to be published in Symbolic Inter-action, vol. 24, no. 2, 2002, pp. 189–216.

2. T. Kindberg and J. Barton, “A Web-based NomadicComputing System,” Computer Networks, vol.35, no. 4, Mar. 2001, pp. 443–456; www.cooltown.hp.com/papers/nomadic/nomadic.htm.

3. T. Kindberg et al., “People, Places, Things: WebPresence for the Real World,” Proc. 3rd Work-shop Mobile Computing Systems and Applica-tions, IEEE CS Press, Los Alamitos, Calif., 2000,pp. 19–28.

4. P.M. Aoki and A. Woodruff, “Improving Elec-tronic Guidebook Interfaces Using a Task-Ori-ented Design Approach,” Proc. 3rd ACM Conf.Designing Interactive Systems, ACM Press, NewYork, 2000, pp. 319–325.

5. A. Woodruff et al., “Electronic Guidebooks andVisitor Attention,” Proc. 6th Int’l Cultural Her-itage Informatics Meeting (ICHIM ’01), A&MIPublishing, Pittsburgh, 2001, pp. 437–454.

6. K. Cheverst et al., “Experiences of Developing andDeploying a Context-Aware Tourist Guide: TheGUIDE Project,” Proc. MOBICOM 2000: 6thAnn. Int’l Conf. Mobile Computing and Net-working, ACM Press, New York, 2000, pp. 20–31.

7. S. Long et al., “Cyberguide: Prototyping Con-text-Aware Mobile Applications,” ACM SIGCHIConf. Companion, ACM Press, New York, 1996,pp. 293–294.

8. M. Lamming and M. Flynn, “Forget-Me-NotIntimate Computing in Support of Human Mem-ory,” Proc. FRIEND21, Int’l Symp. Next Gen-eration Human Interface; tech. report RXRC TR94-103, Cambridge, UK, 1994.

9. P. Debaty and D. Caswell, “Uniform Web Pres-ence Architecture for People, Places, and Things,”IEEE Personal Comm., vol. 8, no. 4, Aug. 2001,pp. 46–51.

10. T. Kindberg, “Implementing Physical HyperlinksUsing Ubiquitous Identifier Resolution,” Proc.11th Int’l World Wide Web Conf., ACM, NewYork, 2002, pp. 191–199.

11.R. Want and D.M. Russell, “Ubiquitous Elec-

tronic Tagging,” IEEE Distributed SystemsOnline, vol. 1, no. 2, 2000, http://dsonline.computer.org/archives/ds2000/ds2wan.htm.

For more information on this or any other computingtopic, please visit our digital library at http://computer.org/publications/dlib.


the AUTHORS

Margaret Fleck is a researcher at Hewlett-Packard Laboratories in Palo Alto, California.Her research interests include computer vision, computational linguistics, and pervasivecomputing. She received a BA in linguistics from Yale University and an MS and PhD incomputer science from the Massachusetts Institute of Technology. Contact her at HPLabs, 1501 Page Mill Rd., MS 1138, Palo Alto, CA 94304-1126; [email protected].

Marcos Frid is a senior R&D engineer at HP’s Embedded and Personal Systems. Hisresearch interests include computer architecture, hardware design, wireless systems, loca-tion-aware technologies, and ubiquitous computing. He received a BS in electrical engi-neering and computer science from Rice University, and an MS in electrical and computerengineering from Stanford University. He is a founding member of MEXPRO, the Associa-tion of Mexican Professionals of Silicon Valley. Contact him at HP Labs, 1501 Page MillRd., MS 1601, Palo Alto, CA 94304; [email protected].

Tim Kindberg is a senior researcher at HP Labs. His research interests include ubiquitouscomputing systems, distributed systems, and human factors. He has a BA in mathematicsfrom the University of Cambridge and a PhD in computer science from the University ofWestminster. Contact him at HP Labs, 1501 Page Mill Rd., MS 1138, Palo Alto, CA 94304-1126; [email protected].

Eamonn O’Brien-Strain works at HP Labs, investigating distributed infrastructures tosupport applications and services that are usable and useful to nomadic individuals in apervasive computing environment. He has a BE from University College Dublin, Ireland.Contact him at HP Labs, 1501 Page Mill Rd., MS 1138, Palo Alto, CA 94304; [email protected].

Rakhi Rajani is studying for a PhD in the area of collaborative work in medical practiceat Brunel University in the UK. Her research interests include cognitive psychology,ethnography, and the role of space, place, and the environment in design. Contact her at26 High Worple, Harrow, Middlesex, HA2 9SU, London, UK; [email protected].

Mirjana Spasojevic is a project manager of the Cooltown Program at HP Labs, focusingon pilot deployments of nomadic computing systems. Her research interests includewide-area file systems, storage systems, and distributed systems. She has a BS in mathe-matics from the University of Belgrade and an MS and PhD in computer science fromPennsylvania State University. Contact her at HP Labs, 1501 Page Mill Rd., MS 1138, PaloAlto, CA 94304; [email protected].

Documents

From informing to remembering: ubiquitous systems in ...km2/projs/pervasive/... · The Exploratorium In most of the projects described earlier, the devices were deployed at museums