8
10 PERVASIVE computing Published by the IEEE Computer Society 1536-1268/07/$25.00 © 2007 IEEE Physical, Social, and Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and re- searchers often create services and applications to help people record their experiences. At the same time, cheap, small, and easy-to-deploy recording technologies are quickly emerging throughout public spaces. In many ways, these technologies are pervasive computing realized. Understanding how people deal with audio and video recording is therefore a good way to explore how people might adopt, adapt, and react to pervasive com- puting technologies in general. Selective archiving is a method for recording data in an environment in which the recording devices (for example, video cam- eras and microphones) are always on and available. The de- vices record automatically into a short cache of data. If no ex- plicit action is taken, the record- ing system throws the data away when it becomes older than a set time. However, a user wanting to archive some data from this cache can retrieve it retroactively before that set time. BufferWare is a selective-archiving application we deployed in an informal space and studied for almost two years. This extended study helped us understand three types of knowledge people use to form impressions of new technologies: Physical knowledge relates to a particular piece of technology’s affordances—that is, the design elements that inform people about the tech- nology and how to use it. 1 Social knowledge is a community’s embedded knowledge of, for example, the conventions, laws, and uses of its space. It involves the set of social interactions occurring in or around a space housing pervasive computing technolo- gies. Social knowledge also includes elements of trust and understanding—or suspicion and confusion—gathered from other social actors in the environment. Experiential knowledge includes the range of information from past experiences, both in new spaces and with similar spaces and technolo- gies. These experiences let potential users and stakeholders learn about their worlds and build models into which the new experiences fit. Here, we aim to add significantly to the re- search surrounding security and privacy concerns by focusing on them rather than just noting them as a side effect of testing an application’s utility and usability. The BufferWare project Open and casual spaces present particular chal- lenges to people wishing to record experiences: People might not be able to predict when events of interest will occur and often aren’t prepared for manual recording. A long-term deployment of a system for recording experiences in informal spaces demonstrates that people use physical, social, and experiential knowledge to determine new technologies’ relative utility and safety. SECURITY & PRIVACY Gillian R. Hayes, Erika Shehan Poole, Giovanni Iachello, Shwetak N. Patel, Andrea Grimes, Gregory D. Abowd, and Khai N. Truong Georgia Institute of Technology

Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

10 PERVASIVEcomputing Published by the IEEE Computer Society ■ 1536-1268/07/$25.00 © 2007 IEEE

Physical, Social, andExperiential Knowledgein Pervasive ComputingEnvironments

Pervasive computing designers and re-searchers often create services andapplications to help people record theirexperiences. At the same time, cheap,small, and easy-to-deploy recording

technologies are quickly emerging throughoutpublic spaces. In many ways, these technologiesare pervasive computing realized. Understandinghow people deal with audio and video recordingis therefore a good way to explore how peoplemight adopt, adapt, and react to pervasive com-puting technologies in general.

Selective archiving is a method for recordingdata in an environment in which the recording

devices (for example, video cam-eras and microphones) arealways on and available. The de-vices record automatically intoa short cache of data. If no ex-plicit action is taken, the record-ing system throws the data awaywhen it becomes older than a set

time. However, a user wanting to archive somedata from this cache can retrieve it retroactivelybefore that set time.

BufferWare is a selective-archiving applicationwe deployed in an informal space and studied foralmost two years. This extended study helped usunderstand three types of knowledge people useto form impressions of new technologies:

• Physical knowledge relates to a particular piece

of technology’s affordances—that is, the designelements that inform people about the tech-nology and how to use it.1

• Social knowledge is a community’s embeddedknowledge of, for example, the conventions,laws, and uses of its space. It involves the setof social interactions occurring in or around aspace housing pervasive computing technolo-gies. Social knowledge also includes elements oftrust and understanding—or suspicion andconfusion—gathered from other social actorsin the environment.

• Experiential knowledge includes the range ofinformation from past experiences, both in newspaces and with similar spaces and technolo-gies. These experiences let potential users andstakeholders learn about their worlds and buildmodels into which the new experiences fit.

Here, we aim to add significantly to the re-search surrounding security and privacy concernsby focusing on them rather than just noting themas a side effect of testing an application’s utilityand usability.

The BufferWare projectOpen and casual spaces present particular chal-

lenges to people wishing to record experiences:

• People might not be able to predict when eventsof interest will occur and often aren’t preparedfor manual recording.

A long-term deployment of a system for recording experiences in informalspaces demonstrates that people use physical, social, and experientialknowledge to determine new technologies’ relative utility and safety.

S E C U R I T Y & P R I V A C Y

Gillian R. Hayes, Erika ShehanPoole, Giovanni Iachello, ShwetakN. Patel, Andrea Grimes, GregoryD. Abowd, and Khai N. TruongGeorgia Institute of Technology

Page 2: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

• To provide security and privacy, therecording technologies must take intoaccount technical-infrastructure re-quirements and social and cultural re-quirements such as notification andconsent.

• The need to record can be infrequent,and barriers to recording might limitthe system’s use.

We considered these challenges whenselecting an installation place for theBufferWare system. We chose a socialarea next to an open stairwell on thethird floor of an academic building hous-ing media-and-design, computer science,and engineering researchers (see figure1). The area contained three tables nextto three large clearboards (large trans-parent writing surfaces that work simi-larly to whiteboards). One table wasnear a window. We chose this spacebecause it was open, informal, and pub-lic�not closed off like meeting rooms,which other researchers have alreadystudied. We instrumented the table near-est the window, so that people wouldn’thave to walk through a recorded spaceto get to a nonrecorded area.

For six months before deployment, werecorded activity in the space using asemicontrolled manual sampling algo-rithm (n > 300 samples). The space’s pri-mary users were graduate students, staff,faculty members, and professional re-searchers with offices or meetings in thebuilding. A negligible number of other

people, such as undergraduate studentsand family and friends of workers in thebuilding, used the space occasionally.Most of the time, the space was unoc-cupied, but during the afternoon, allthree tables, all the chairs, and somestanding space were often filled. Typicalactivities in this space include

• small meetings;• lunch, snack, or coffee break gatherings;• individual reading and work sessions;• individuals and groups looking out the

windows;• telephone conversations;• eavesdropping on conversations on

other floors; and• group discussions unrelated to work.

BufferWare’s capture services supportrecording of these activity types. The sys-tem is inherently flexible, requiring min-imal infrastructure (see figure 2). Userscan access saved content online. A touchscreen embedded in a café table providedthe interface to BufferWare. A singlecamera, attached to a PC via video andsecurity cables, provided video data tothe application. Finally, a microphoneglued to the tabletop recorded soundwithin a space matching the cameraviewing angle.

We first deployed BufferWare fromSeptember 2005 to June 2006. Weplaced blue tape along the carpet todenote the recorded area, and postedsigns and sent emails to alert building

occupants to its presence. We also beganlogging requests to save or deletebuffered data or review clips. In April2006, we began using motion detectorsto log motion in the space (see figure 3).

Survey and interview input from thefirst deployment revealed several small,correctable problems. We fixed them tohelp disambiguate technological con-cerns from ease-of-use problems. Thesefixes included

• enlarging buttons to accommodatepeople whose fingers were too wide toinput easily, or who didn’t have theright manual dexterity,

• expanding the buffer length to onehour for longer meetings,

• doubling the video capture resolution,and

• providing a URL with a “magic key”that automatically authenticated usersto view newly saved clips.

Additional changes to the primaryinterface addressed concerns aboutvisual feedback and video data access,including an expanded viewing windowand notices about policies for protectingsaved data. We deployed a second ver-sion of BufferWare during Septemberand October 2006.

After the second deployment, weremoved the BufferWare hardware,which included the camera, microphone,and touchscreen interface. We continuedto gather motion detector data for the

OCTOBER–DECEMBER 2007 PERVASIVEcomputing 11

Figure 1. The public area used for BufferWare deployment: (a) two of the tables in the room (we instrumented the table nearestthe window), (b) a view of a table and clearboard, and (c) the open stairwell.

Page 3: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

next 30 days to record activity in thespace after the equipment had beenremoved (the postdeployment condi-tion). The online service continues toallow access to saved content but not tolog and analyze the interactions.

Anyone in the building could partici-pate in the BufferWare study in variousways:

• using the service itself (36 people didthis),

• completing anonymous surveys (32people took surveys: 13 for the firstdeployment and 19 for the second),

• participating in interviews (we held 27interviews with 22 people: five peopleafter the first deployment only, 12 afterthe second only, and five after both).

Nearly all the interviewees also com-pleted surveys.

We used sensors and the BufferWaresoftware to log user interactions with ornear BufferWare. We attached black tapeover portions of the motion detectors tolimit their view to the zone depicted infigure 3. The log associated with thesemotion detectors contains the date andtime stamp for each sensor firing.

We examined the space’s use during thefirst 30 days of each deployment and the

first 30 days of the postdeploymentphase. Across all three time periods, thespace was occupied approximately 20percent of the time. (Although the systemwas on only 18 hours a day, from 6 a.m.until midnight, we used a full 24-hourday in these calculations. That’s becausewithout careful inspection, people didn’tnecessarily know the hours of service.)People used the space approximately 18seconds more per day during the seconddeployment than in the first (two-tailedmatched t-test, p < 0.005) and approxi-mately 10 minutes more per day after weremoved recording from the space (two-tailed matched t-test, p < 0.001). It’s hardto be certain why these differences exist.Although they’re statistically significant,they’re small. Furthermore, in self-reported survey data of space use, no sig-nificant differences existed between con-ditions. So, we would need furtherexploration or a replication of the studyto better understand these results.

We also recorded interactions with thesoftware through automatic logging.Across both deployments, people created174 archives. Users accessed these videoclips 257 times during the entire study(both deployments and postdeploy-ment). Most clips were accessed. Userscleared the buffer manually 598 times

by pressing the Delete button rather thanwaiting for the system to delete the data.Some of these presses were in a row—likely, a person trying to ensure that thedeletion had registered.

Knowledge in privacy and security determinations

Here’s how participants used the threetypes of knowledge in relationship tosecurity and privacy while interactingwith BufferWare.

PhysicalIdentifying appropriate ways to notify

people that they’re being recorded andto let them provide feedback about thatrecording are significant issues forresearchers in security and privacy andfor designers of new recording tech-nologies. Minimizing the recording’s no-ticeable impact in a space makes it hardfor people to know what’s happeningand how to control it, creating usabilityand sociocultural issues. At the sametime, designing for maximum notifica-tion about recording can be over-whelming, given the sheer volume ofrecording in public places, and can makeusers self-conscious. So, designers muststrike a balance to optimize users’ abil-ity to create appropriate mental models

12 PERVASIVEcomputing www.computer.org/pervasive

S E C U R I T Y & P R I V A C Y

Figure 2. The BufferWare system: (a) A touch screen in the table provides two buttons, one for saving and one for purging the datacache. (b) Users can choose how much to save using a double slider and can send the data to registered users. (c) A Web interfacelets users view and comment on saved videos.

Page 4: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

of recording from the available cueswithout creating other challenges.

We intentionally made visible notifica-tion and feedback in the BufferWare proj-ect’s recorded space plentiful. Specifically,BufferWare’s design included a visiblecamera and a small window, embedded inthe touch screen on the table, showing thelive video capture feed. As we mentionedearlier, we placed numerous signs in andoutside the space and changed them sub-stantially between the first and seconddeployments. As with the other issues weexplored, interview and survey partici-pants provided abundant informationabout physicality and visibility, includingnotification and feedback concerns.

People’s responses to the visible camerawere divided. General distaste and dis-comfort with cameras were some of thereasons reported for not liking Buffer-Ware. Interviewees often noted that secu-rity cameras are more acceptable becausethey’re often inconspicuous. Some com-mented that they became nervous or self-conscious when having their imagesrecorded. For example, one intervieweenoted that security cameras are less of aconcern, because “I’m not looking at thecamera directly and … sometimes I’m not

even aware that there’s a camera there. So,when I notice the camera I get nervous.”On the other hand, many intervieweescommented that they preferred visibilityand transparency to any recording.

One interesting issue is that even a pub-lic space isn’t one-dimensionally so. InGavin Jancke and his colleagues’ experi-ments, a vocal minority of users reporteddistaste for an always-on system linkingtwo public spaces.2 These users describedmany private activities that took place inthe public space (personal phone calls,eating lunch, meetings, and so on), andthat the private nature of this public spacewas disrupted.

Reflection of the capture stream isanother way to provide feedback aboutthe recording experience, a notion cen-tral to David Nguyen and ElizabethMynatt’s “privacy mirror.”3 You couldalert people who might be recorded tothe presence of these technologies byusing a video display of the data beingcaptured. In BufferWare, the small screenembedded in the touch screen served thispurpose. Our choice of a horizontal dis-play rather than a vertical wall displaywasn’t a trivial one. We wanted to givefeedback to individuals in the space

without providing too much of thatinformation to people outside the space,who might see a vertical display.

Some individuals reported that theinformation was harder to view on thehorizontal screen: “Maybe you need tohave an LCD on the wall that’s showingyou ‘Hey! This is what’s being recordedright now!’” This same individual com-mented that a vertically mounted displaymight resemble a mirror and could con-tribute to a “surveillance feel” that mightmake a person uncomfortable or self-conscious—further demonstrating users’conflicting sentiments.

People also reported that simple phys-ical differences between the BufferWare-enabled table and the other tables in thespace added to their feelings of being partof something new, different, and possiblyuncomfortable. For security reasons—both information security and physicaltheft—the server and display units werelocked to the table. We also used hard-wired cables to transport the data insteadof less secure wireless signals. These dif-ferences often had little to do with therecording itself, but still marked the tableas different and somehow “not for[them],” as one participant commented.

OCTOBER–DECEMBER 2007 PERVASIVEcomputing 13

Clearboards

Stairwell

Motion detector

Windows Motion detector and cameraTable Table

Figure 3. Two motion detectors let us track movement in the camera angle and near the table separately.

Page 5: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

Furthermore, usability considerationsand interface choices impacted how peo-ple understood the system. The systemintentionally hides complex informationabout data storage and deletion to makeit more comprehensible. Similarly, to make

the interface usable given a touch screen’sconstraints, we made certain compro-mises. For example, rather than ask usersto type an email address to save data (aswe originally designed), which more max-imally protects privacy, the BufferWareinterface displayed registered usernamesas quick buttons. To save data, users sim-ply pressed the buttons to select theiraccounts. Although some users createdusernames that weren’t easily identifiable,this display still presents potential risk.

Institutional considerations stronglyinfluenced how we notified people aboutour recording. A legal representative forGeorgia Tech’s Office of Sponsored Pro-grams prescribed the wording on the ini-tial notification signs as well as the agree-ment at the time of archiving indicatingthat anyone in the space is comfortablewith recording. During the first deploy-ment, however, people perceived thesigns as intimidating and scary. For thesecond deployment, we used larger, morecolorful signs that stressed the technol-ogy’s potential uses. The required word-ing remained on the signs, however. So,the signs might have continued to con-vey a sense of danger.

The blue tape line in the first deploy-ment notified people about recording inthe space. Many people commented thatthis explicit notification made them un-comfortable because it seemed to indicate

potential risk. To combat these concerns,we removed the tape during the seconddeployment. People interviewed after thesecond phase who had been in the build-ing during the first phase almost unani-mously agreed that its removal made the

space seem less threatening. So, these clearindicators might increase people’s aware-ness at the expense of a reasonable abil-ity to judge the level of concern theyshould have about the recording tech-nologies in the space.

The use of multiple indicators warn-ing people that they might be recordedcould have had an effect opposite to thatintended. We thought this intentional vis-ibility would enhance the trust level be-tween people using the space and theresearchers. However, it might have pre-vented potential users and the researchersfrom showing each other that they couldbe trusted through other means. ThomasErickson and Wendy Kellogg argued thatthis natural trust developed over time intheir Babble chat system deployment.4

How best to notify people about re-cording has been a significant challengefor BufferWare and for other capturesystems. Because people differ so widelyin their notification preferences (for ex-ample, some want to be notified everytime a recording occurs, others wish tobe notified the first time only, and stillothers prefer no notification), a one-size-fits-all solution is unlikely.

These experiences and concerns exem-plify the conflicted responses we receivedwhen querying about physical knowl-edge. People want to make informed deci-sions about video capture, and part of

that process is knowing (and sometimesseeing) what’s being captured. How-ever, constant feedback in the physicalenvironment can create an unnecessaryaura of danger, making people moreself-conscious and concerned than theymight have been with less feedback.

SocialSocial cues can be extremely impor-

tant for building models of security, pri-vacy, and trust in a system. Knowingwhat other people think, talking withother people affected by the system (or incharge of it), and the general social pres-sures of belonging to a group can allaffect people’s perceptions of technology.The BufferWare project was no excep-tion to this pattern. During the two testdeployments, people reported changingtheir minds about the system after see-ing other people using or avoiding thespace or after talking with other peopleabout it. People often queried otherswith whom they were interacting whendeciding whether to use the space. Ifsomeone wanted to use it, the entiregroup often would. If one person wasuncomfortable, the entire group mightacquiesce and avoid it.

We used the email distribution lists forbuilding occupants to communicate sys-tem status changes and other information.During the second deployment, however,interview and survey responses indicatedthat many people, particularly those whowere newest to the building when they firstencountered BufferWare, didn’t knowwhat was happening in the space. Afterfurther probing, we learned that some peo-ple who had moved into the building inthe past year weren’t on our email lists.Instead, they reported relying on whatother people were saying, a situation sim-ilar to that encountered when a commer-cial product is released or a government-sponsored initiative is put in place todeploy new technologies, and productbuzz determines reactions.

14 PERVASIVEcomputing www.computer.org/pervasive

S E C U R I T Y & P R I V A C Y

Knowing what other people think, talking with

other people affected by the system, and the

general pressures of belonging to a group can

all affect people’s perceptions of technology.

Page 6: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

Other people’s use of BufferWare gaveimportant social cues to potential usersthat it was safe in terms of security andprivacy. These cues include witnessingother people in the space or receiving anemail stating that a user had saved someclips. If the email’s receivers didn’t havea BufferWare account, the messageasked them to create an account andbecome a registered user. We designedBufferWare to be flexible in terms of useby individuals or groups. Interestingly,nearly every reported use of the systemfor archiving involved multiple people.People used BufferWare to share infor-mation. For example, users sent snippetsof meetings when they wanted to saveor share exact wording. For the mostpart, however, interviewees commentedthat such moments are rare and currentpractices for documenting informalmeetings, such as taking notes or pho-tographing large shared writing surfaces,are often sufficient.

Almost all people who reported view-ing archived video segments viewedmore segments sent by other people thanthose they saved themselves. Even whenthey didn’t expect an archive, peoplewere intrigued when unknown videoclips appeared in their accounts. Mostunknown clips came from people whoappeared to be experimenting with thesystem but chose not to register for anaccount.

Particularly in pervasive computingsystems, where a technology’s workingsmight be hidden, people often rely onthose around them for indicators of asystem’s use and safety. So, it’s importantto exploit existing social practices whendesigning and deploying new pervasivecomputing technologies.

ExperientialPeople often use their past experiences

with technologies to determine appro-priate reactions to new technologies.Furthermore, a group’s past experiences

that become shared cultural under-standing impact responses to new tech-nologies; people might already havebeliefs about the “right” response withinthis culture. For example, previous expe-rience with research projects contributedto people’s respect for them and pre-vented even those individuals who feltnegatively about recording from activelyworking against the BufferWare project.Past experiences with similar researchprojects also led many participants to be-lieve that the researchers might examinethe video clips, because past researchprojects had focused on such questionsas the content of saved data.

The past experiences of this study’sparticipants aren’t representative of thegeneral population. Many of the peopleimpacted had computing-related back-grounds and interests. This group’sunderstanding of computing adds newchallenges, such as how to ensure thatproper security mechanisms are in placefor a group of people highly familiarwith the potential security risks inherentin networked computing. It also createdan environment in which people who arearguably knowledgeable about and con-

cerned with security and privacy issuescould articulate these concerns.

Governmental and commercial entitieshave encountered other considerationswhen deploying systems in environmentswith technologically savvy users. Theseexperiences provide a view into the poten-tial range of reactions to new technolo-gies, particularly as the world’s popula-tion becomes more accustomed tocomputing and attuned to the issues

related to pervasive computing. Manysolutions let concerned and technicallyinclined individuals manage and analyzerisks for themselves. For example, theLinux open source community lets peo-ple compile their own operating systems.Similarly, when the wireless FastTrak sys-tem was installed to speed toll collectionin Northern California’s Bay Area, secu-rity and privacy details were made publicso that those concerned could check thesystem details (see http://traffic.511.org/privacy.asp). Although this solution putsome stakeholders at ease, in this andother pervasive computing environmentsin which a third party maintains service(such as toll collection or BufferWare),users must still trust that the inspectedalgorithms are the ones running.

Given these constraints, pervasivecomputing service providers must buildthe trust required for adoption and ac-ceptance. Through repeated positive ex-periences with a particular service pro-vider, users develop stronger feelings oftrust, and the brand associated with thatprovider will likely become trusted. Forexample, the success of Google docu-ments, spreadsheets, and mail demon-

strates the Google brand’s impact. Peo-ple will store sometimes sensitive andpersonal information on the corporateservers largely because of the repeatedpositive experiences they’ve had with thesearch engine and other applicationsassociated with the Google brand. In theBufferWare project, our research group’sbrand carried trust for certain stake-holders, enabling the system’s adoption.However, rather than use a lengthy server

OCTOBER–DECEMBER 2007 PERVASIVEcomputing 15

A group’s past experiences that become shared

cultural understanding impact responses to new

technologies; people might already have beliefs

about the “right” response within this culture.

Page 7: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

16 PERVASIVEcomputing www.computer.org/pervasive

S E C U R I T Y & P R I V A C Y

address and difficult-to-remember IPaddress, as is automatically assigned foruniversity servers, we registered and useda short, easily remembered domain name.The divorce of the server address from atrusted set of server addresses associatedwith the Georgia Tech brand was a con-cern for some users.

Many stakeholders reported that re-cording in BufferWare was complicatedand they didn’t immediately understandit. They often ignored the notificationsand descriptions provided and tried toreconcile this new mode of recording totheir previous understandings. Ratherthan map BufferWare to the familiar andsimilar analog tape loops of security andother systems, they typically chose a dif-ferent familiar technology�the stream-ing webcam. It isn’t surprising, given thiscomparative technology, that they didn’timmediately understand the automaticor even manual deletion of buffered orcached data. As designers, we must con-sider how much information peoplebring to their interactions with a capturedspace (for example, assumptions aboutuse and retention policies), and find astraightforward way to educate andinform potential stakeholders, particu-larly when learning about a service isn’ta high priority.

Not only is past experience with simi-lar technologies relevant to understand-ing a new technology, but experienceswith the new technology itself are alsoimportant. Giovanni Iachello and JasonHong describe a “privacy hump” inwhich people start out largely pessimistictoward a new technology but becomeoptimistic over time.5 Repeated positiveencounters with technology let peoplebuild experiential knowledge that leadsto their accepting rather than fearing it.

At times, even after repeated exposureto a potentially risky technology hasproved safe, people still have concerns,demonstrating an internal conflict be-tween what they inherently believe and

what they logically deduce. For exam-ple, one participant commented,

There’s the rational part of mybrain that says … there’s nothingnefarious going on. But then there’sthe paranoid part of my brain thatsays … wait, but they could be. Howdo I know they’re not?

These conflicting sentiments indicatethat there’s often more to acceptabilityissues than risk-and-reward analyses orother rule-based decision criteria. Fur-thermore, people make decisions on asubconscious level that they can’t artic-ulate, creating a large design challenge.

A comparison of BufferWare’s deploy-ment with that of other pervasive com-puting technologies reinforces this con-clusion. The deployment of a similarexperimental “office memory” systemat an Electricité de France research labshows the effects of context.6 EDF em-ployees developed and used an audio-video recording system that continuouslyarchived everything happening in the lab.Unlike BufferWare, the archiving was per-manent, and all lab members could accessthe recordings. An interesting privacycontrol was that EDF tracked every accessto the recordings, similarly to the opti-mistic security protocol,7 and informedeach individual of the identity of the per-son accessing the recordings of his or herworkstation. These interactions’ trans-parency let people build over time a set ofexperiences that led to an optimistic, safemodel of the system’s use.

Services in pervasive computingenvironments can perform ac-tions for users without requiringmuch attention and effort. How-

ever, we must take care not to make“invisible computing” a literal and ex-plicit goal. Instead, designers shouldleverage physical, social, and experien-tial knowledge to help users decide howto adopt and adapt to new pervasive

computing technologies. As educationabout and experiences with technologieschange over time, organizations shouldcarefully plan the introduction of newtechnologies.8 Understanding the threeknowledge types can greatly enhance thedesign and adoption of pervasive com-puting technologies with privacy andsecurity implications.

REFERENCES1. D.A. Norman, The Design of Everyday

Things, Doubleday, 1990.

2. G. Jancke et al., “Linking Public Spaces:Technical and Social Issues,” Proc. 2001SIGCHI Conf. Human Factors in Comput-ing Systems (CHI 01), ACM Press, 2001,pp. 530–537.

3. D.H. Nguyen and E.D. Mynatt, “PrivacyMirrors: Understanding and Shaping Socio-technical Ubiquitous Computing Systems,”Georgia Inst. of Technology tech. reportGIT-GVU-02-16, June 2002.

4. T. Erickson and W. Kellogg, “Social Trans-lucence: An Approach to Designing Systemsthat Mesh with Social Processes,” Trans.Computer-Human Interaction, vol. 7, no.1, 2000, pp 59–83.

5. G. Iachello and J. Hong, “End-User Privacyin Human Computer Interaction,” in review.

6. S. Lahlou, “Living in a Goldfish Bowl:Lessons Learned about Privacy Issues in aPrivacy-Challenged Environment,” Proc.Privacy Workshop at Ubicomp Conf.,2005, http://people.ischool.berkeley.edu/~jensg/Ubicomp2005/papers/2-Lahlou.pdf.

7. D. Povey, “Optimistic Security: A New Ac-cess Control Paradigm,” Proc. New Secu-rity Paradigms Workshop, ACM Press,1999, pp. 40�45.

8. G. Iachello et al., “Prototyping and Sam-pling Experience to Evaluate UbiquitousComputing Privacy in the Real World,”Proc. 2006 SIGCHI Conf. Human Factors inComputing Systems (CHI 06), ACM Press,2006, pp. 1009–1018.

For more information on this or any other comput-ing topic, please visit our Digital Library at www.computer.org/publications/dlib.

Page 8: Physical,Social,and Experiential Knowledge in Pervasive ...shwetak/papers/b4hay.pdf · Experiential Knowledge in Pervasive Computing Environments P ervasive computing designers and

OCTOBER–DECEMBER 2007 PERVASIVEcomputing 17

the AUTHORSGillian R. Hayes is an assistant professor in the Department of Informatics at theUniversity of California, Irvine’s Donald Bren School of Information and ComputerSciences. Her research interests are human-computer interaction and ubiquitouscomputing, specifically in the areas of record-keeping and surveillance technologies.She received her PhD in computer science from the Georgia Institute of Technology.She’s a member of the IEEE, ACM, and SIGCHI. Contact her at the Donald Bren Schoolof Information and Computer Sciences, Univ. of California, Irvine, 6210 Donald BrenHall, Irvine, CA 92697-3425; [email protected]; www.gillianhayes.com.

Erika Shehan Poole is a doctoral student in the Georgia Institute of Technology’sCollege of Computing and Graphics, Visualization, and Usability Center. Her re-search focuses on usability issues associated with networking and security. She re-ceived her BS in computer science from Purdue University. She’s a member of theACM and IEEE. Contact her at the GVU Center, Georgia Inst. of Technology, Tech-nology Square Research Bldg., 85 5th Street NW, Atlanta, GA 30332-0760; [email protected]; www.cc.gatech.edu/~erika.

Giovanni Iachello is an associate consultant at McKinsey & Co., focusing on high-tech and media strategy. His research interests relate to the human aspects of pri-vacy and off-the-desktop technologies. He received his PhD in computer sciencefrom the Georgia Institute of Technology. He’s a member of the IEEE, ACM, Interna-tional Federation for Information Processing working group 9.6/11.7 “IT-Misuse andthe Law,” and SIGCHI. Contact him at [email protected].

Shwetak N. Patel is a PhD candidate in the Georgia Institute of Technology’s Col-lege of Computing and Graphics, Visualization, and Usability Center. He’s also an as-sistant director of the Aware Home Research Initiative. His research focuses on low-cost sensing systems, context-aware mobile phone applications, and technology tosupport ubiquitous computing applications. He received his BS in computer sciencefrom the Georgia Institute of Technology. He’s a member of the IEEE Computer Soci-ety and the ACM. Contact him at the GVU Center, Georgia Inst. of Technology, Tech-nology Square Research Bldg., 85 5th St. N.W., Atlanta, GA 30332-0760; [email protected]; www.cc.gatech.edu/~shwetak.

Andrea Grimes is a PhD student in human-centered computing at the Georgia In-stitute of Technology. Her research examines the sociocultural nature of eating inlow-income, urban communities to design technologies that promote healthy eat-ing in this context. She received her BS in computer science from NortheasternUniversity. Contact her at the GVU Center, Georgia Inst. of Technology, Technol-ogy Square Research Bldg., 85 5th St. NW, Atlanta, GA 30332-0760; [email protected]; www-static.cc.gatech.edu/~agrimes.

Gregory D. Abowd is a professor in the Georgia Institute of Technology’s School ofInteractive Computing and Graphics, Visualization, and Usability Center, the directorof the Aware Home Research Initiative, and the associate director of the Health Sys-tems Institute. His research focuses on an application-driven approach to ubiquitouscomputing concerning both human-centered and technology-driven themes. Hereceived his DPhil in computation from the University of Oxford. He’s a member ofthe IEEE Computer Society and the ACM. Contact him at the GVU Center, GeorgiaInst. of Technology, Technology Square Research Bldg., 85 5th St. NW, Atlanta, GA30332-0760; [email protected]; www.gregoryabowd.com.

Khai N. Truong is an assistant professor in the University of Toronto’s Departmentof Computer Science. His research lies at the intersection of human-computer inter-action and ubiquitous computing, and focuses on usability and acceptance issuessurrounding automated capture and access and context-awareness applications. Hereceived his PhD in computer science from the Georgia Institute of Technology. Con-tact him at the Sanford Fleming Bldg., 10 King’s College Rd. Rm. 3302, Toronto, ONM5S 3G4, Canada; khai@ cs.toronto.edu; www.cs.toronto.edu/~khai.