Philip Agre Surveillance And Capture

download Philip Agre Surveillance And Capture

of 21

Transcript of Philip Agre Surveillance And Capture

  • 8/13/2019 Philip Agre Surveillance And Capture

    1/21

    51. Surveillanceand CaptureOriginal Publicationnjormation Society 10 2):101-127. Aprit-June 1994.SurveiUance andCaptureTwo Models of PrivacyPhilip E Agre1 IntroductionIdeas about privacy are. among other things, culturalphenomena. They are shaped through historical e>-'Petience,they conilltion perceptions of newly arising phenomena, andthey are reproduced or transformed in all of the samecomplicated ways as other elements of culture. Cultural ideasabout privacy are particularly significant right now, given therapid emergence of new technologies and new policy issuesaround privacy. ln this paper I propose to contrast twocultural models of privacy:

    Th'e sur\'elllance model." currently dominant in thepublic discour:;e of at least the English-speakings

    theNEWMEDIAREADER

    Section 3 takes up this observation more formally byintroducing and defining the surveillance model andthe capture model of privacy issues.Section 4 discusses the capture model in more depth,relating it to deeply ingrained aspects of appliedcomputing as a professional practice. It introduces theconcept of a grammar of action and provides severalexamples. It then describes an idealized five-stagecycle for the development of capture systems andreflects on certain computer-supported cooperativework systems in this light.Section 5 desclibes some trade-offs inherent in theconcept of capture. and consequently in the verydesign of computer systems as they are currentlyunderstood.Section 6 introduces the general question of captureas a social phenomenon, insisting that capture bestudied against the background of the largerinstitutional dynamics in which it is embedded.Section 7 offers a provisional analysis of the politicaleconomy of capture. starting with a discussion of therole of inform

  • 8/13/2019 Philip Agre Surveillance And Capture

    2/21

    buildings in which climate controls are integrated \vithnetworked digital systems.

    Active badges may be the best-known tracking tedmology,but they are hardly unique. Other tracking schemes involveradio-frequency beacons installed on materials inmanufacturing and distribution (Fales 1992. Sabetti 1993).And the trade press has reported on numerousimplementations of tracking systems:

    ~ ~ ? . u ~ e s bar-codes and a customized electronicclipboard to track the movements of packages: when apackage is delivered, the clipboard digitally records therecipient's signature and sends information about thepadcage's sta tus to a central compu ter through anationwide ce luJar telephone netwo rk (Duffy 1993.Eckerson 1991).The Canaclian Ministry of Transportation uses a\vireless pacltet raclio nehvork and a national database

    ~ ~ k e e p track of o m m e r d ~ 'vehicles in Toronto. Policeand inspectors use information provided by the systemto checlc drivers' speed and watch for unlicensedvehicles, and they can call up a complete history of anyvehicle in a few seconds (Loudermilk 1993).A trucking firm caJled Americana Inc. uses \vireless.communications and the US military's G f ~ b : JPositioning System (GPS) to allow dispatchers toautomatically track its lrucks. Each tnJCk canies anApple Macintosh that periodically talws a readingfrom a GPS device and sends it lo headqu.mers byelectronic mail (Lawton 1992). Computer networks are increasingly ma.ldng possibleautomatic realtime d.1ta collection and analysis forla.rgescale accounting and control systems. and thisdevelopment is revolutionizing if belatedly) the fieldof managem ent accounting (Johnson and Kaplan1987,5-6 ff.).A system called Voice Frame is used to monitorpeople who have been convicted of crimes. Eachoffender wears a bracelet that notifies the authoritiesif it passes o u t s i d ; ~ ~ c r t a i n boundary (Leib01.vitz1992) In a wide range of "virtual r e a l i t y ~ and "telepresence"systems, some mechanism continually informs a /(:computer about the locations of certain parts of a

    Lperson's body. The locations might be computed and transm itted by devices that are physically attached tothe relevant body parts, or they might be computed

    51. Surveillanceand Captureby a stationary device that observes the body'smotion, per haps through a video camera (Meyer.Applewhite, and Biocca 1992). One division of NCR has integrated its just-in-timemanufacturing systems with a plantwide system ofbar-code readers. The statu s of a c h job is availablefrom computer terminals throughout theorganization (Anonymous1990). Fast food chains are rapidly integrating theiroperations through pointof-sale (POS) terminals andbookkeeping systems for tracking individual stores'activities by interconnecting their local computers\vi th mainframes at headquarters, which performsintensive analysis of the resulting data (l'vkPartlln1992. Simpson 1989). items captured and stored inthe databas e include "product mi.'\, sales statistics,labor information. food costs." "bank deposits, cashregister information, sales totals. average orderamounts at different points in the day. and customertraffic (Baum 1992). Most such systems have replacedbranch managers' functions with centralized control(Walton 1989: 42 ff.). a pattern found throughoutmass retailing (Smith 1988).Numerous projects are currently building S)'Stemsfor "design rationale capture" (DRC) (Carroll andMoran 1991). The idea is that design changes in largeengineering projects are o ften made difficult by inadequate institutional memory about the reasoningbehind previous design decisions. A DRC system fillsthis gap by allowing designers to maint ain a runningaccount of their reasoning duling the design process.using a taxonomy of typ s of reasoning and acomplex system of datastructures for representingthem all This material is then stored for laterreference. For Carroll (1992). design rationale captureis the culmi nation of an underlying logic of computerdesign activities. Design practice, he argues, can beviewed as reifying a partietUar work practice, anddesign rationale capture similarly involves thereification of the design process itself, with all of itselements of hermeneu tic inquiry.Several vendors have built sofn.o,rare systems fortracking job applicants through the whole applicationand inte rview process. The systems can keep track ofeach individual's papen ork generate routine letters,and maintain a database of applicants that can besearched in a \vide variety of ways, includinggenerating documents for affirmative action

  • 8/13/2019 Philip Agre Surveillance And Capture

    3/21

    51 Surveillanceand Capturereporting and the like. Employees participating in thehiring process update the database upon each step ofthe process (Romei 1991). In Thailand. the Ministry of the Interior isdeveloping a centralized da tabase to maintaininformation on each of the country's citizens. Eachindividual will have a unique identification numberrecorded on a card with a magnetic strip (Hoffman1990).

    The cases vary among themselves in several ways. Althougheach system keeps track of significant changes in a trackedentity:S state, the nature of these changes valies. In somecases the changes are simply changes in physical location,reckoned against some kind of stationaty coordinate grid;the system may well place some kind of interpretation onthese locations, perhaps relative to a street map. In othercases the changes are defined in institutional terms, forexample whether a package has been formally received orwhether someone has been formally offered or turned dovmfor a job. In the former case. the t e n T I ( t r ~ ~ f d ~ g ; t a l ~ e ? on_ amore literal sense of tracking through space. In the latter

    caSe, the. term "tracking" is a metaphor the entity in questiontraces a trajectory through a more abstract space whichmight have numerous "dimensions."

    One might further distinguish between systems that trackhuman beings and systems that t r : ~ c k physical objects. Such adistinction \vould be misleading, though. Systems are indeedfound at each e::treme-fO ' example radio transmittersattached to shipping crates or fastened to prisonets' limbs.But many of the systems track both people and objects, andothers track objects as stand-ins for people. A system thattracks people by means of identification cards, for example. isreally tracking the cards; any connection between the cardand person wilJ have to be made in some other way, such asan official or supervisor checking each individual'sappearance against a photograph upon each significantevent. Similarly. a system that tracks trucks can generallydepend on a stable correspondence, at least over shortperiods. between truch and their drivers.

    Systems that track physical objects. for their part. varyconsiderably in the means by which they detect significantstate changes. Some depend on complex schemes forreckoning absolute or relative location; these systems mayonly require an approximate location, and thus may onlyreceive a periodic update from a location measuring device.

    theNEWMEDI RE DER

    Other systems depend on a clistlibuted system of passivesensors. Yet others might involve sensors that actively seekout the entities being tracked. But a large number of systemsinvolve human intervention: a human being executes somephysical action that closes a causal chain between the trackedentity and the centralized system, thereby signifying that

    s u c h ~ a n d s u c h a state-change has taken place.In general, the various tracking systems v ry widely in the

    way they divide their computational labor between themoving entity, some stationary computer system, andvarious human or mechanical intermediaries. A GPS device,for example, performs all of the necessary computation atthe location of the object being tracked. At another extreme,a tracking system might employ an algorithm to locate thetracked entity within each successive video image it receivesfrom a stationary camera. And in the middle ground betweenthese extremes lie numerous schemes for splitting theburden of tracking. for example by placing bar codes or LEDson the entity being tracked. or by testricting the entity'smovements so that it necessarily comes into contact with reiatively simple sensors (Ucloka 1991) . (For a generaltreatment of this trade-off in the design of robots. seeDonald (fmthcoming).)

    Despite all of these variations, the various trackingsystems have a gteat deal in common. In each case. someentity changes state, a computer internall y represents thosestates, and certain technical and social means are pro\'idedfor (intendedly at least) maint aining the correspondencebetween the representation and the reality. The computerm y maintain a centralized database (this is the usual case)or it may be more widely distributed. Each entity has adefinite identity that remains stable over time. and if se\eralentities are being tracked then the tracking system has somemeans of consistently "attaching" a given entity to itscorresponcLng representation. This represent ation \\ill beexpressed within some mathematically definablerepresentation scheme, \ '-'hich is capable of expressing acertain formal space of states of affairs. The computermaintains a representation only of certain aspects of theentity. In particular. the represe ntation scheme recognizescertain specific kinds of changes of state, namely those whichcorrespond to changes in the stored representation. A systemfor tracking an object's location, for example, should beunaffected by changes in its color; the recognized statechanges Arill all talo::e the form of transitions from. say. one

  • 8/13/2019 Philip Agre Surveillance And Capture

    4/21

    sequence of coordinates to another. As the entityScorresponding representalion changes, records may well bekept of its state transitions, yielding a ''history" of itstrajectory through time. And this trajectory, of course, can beeither literal or metaphorical, depending on what aspects ofthe entity are represented.In addition to the continual updating of a representation,each tracking system is capable of closing a causal loopbetween the entity and the computer. That is, informationdoes not simply flow from the entity to the computer; inaddition, certain human or mechanical agents, faced with agiven entity in a prescribed type of situation, are capable ofdetermining its identity and "calling up" the information inits "file." (These agents' activities may, of course. be tracked swell.) Again, the causal means that provide for this loopclosing vary vridely. from bar codes to identification cards tolicense plates to keys to paperwork of all sorts. and thecomputa tional division oflabor among the entity. agent.central computer. and so forth varies widely as well.

    Tracking systems like these can obviously be used forgood or ill. Other things being equal. it is probably a goodidea to track hazardous materiaJs. government money, andso forth. At the same time, research on computers andprivacy has emphasized the fe< lr, often perfectly justified.that the accumulated information about a tracked personmight be used for abusive purposes. for example stalking bya would-be assailant, irresponsible publication ofembarr assing facts. or oppressively detailed con trol of workactivities. fn particular, this research has focused on theelement of data-collection; its question is what becomes ofthe data once it is collected. Yet tracking schemes haveanother side: the practical arrangements through which thedata is collected in the first place, including thearrangements that maJ.::e human activities and physicalprocesses trackable. As human activities become intertwinedwith the mechanisms of computerized tracking. the notionof human interactions with a computer -understood as adiscrete. physically localized entity-begins to lose its force;in its place we encount er activity-systems that arethoroughly integrated with distributed computationalprocesses. It is this deeper implication of tracking that formsthe central mo tivation for this paper.

    3 Surveillance and capture

    51. Surveillanceand Capture

    Let us, then. foxmally introduce the surveillance model andthe capture model of privacy issues. A "model." for presentpurposes, is a way of looking at things; specifically, it is a setof metaphors. Distinct models do not divide the world'ssociotechnica.l phenomena into nonoverl apping classes;instead. they simply point out some potentially significantfeatures of the phenomena-features that may call for moreconcrete ana.l,\sis. ' \The surnilla.nce model has five components:

    1) 'isual metaphors. as in Orwell's "Big Brother iswatching you" or Bentham's Panopticon;2) the assumption that this "watching" is

    nonc Jsruptive and suJTeptitious (except perhapswhen going astray or issuing a threat);(3) tmitoriaJ metaphors. as in the "invasion" of a"pdtate- personal space. prototypically the familyhome. marked out by "rights" and the oppositionbetHeen -coercion" and "consent";\4 centralized orchestration by means of abureaucrac:y with a unified set of "files"; and(5) ider:c ikation with t h ~ S t < ~ t e and in particulartdth c:onsdous]y plannedout malevolent aims of aspedfic.?. ly political n a t u n ~

    When stated in this t.:ay. it becomes evident that thesurveillance model is a cultural phenomenon. Although itsearliest geneaJogy deserves further research. its modernhistory is clear ;- rooted in the historical eAperience of secretpolice organi2ations and their networks of listening devicesand informers. most prominentlr in the totalitarian states ofNazi German1 and the Soviet Union, and to a lesser but stillsignificant extent in the United States. George Orwell's 984gave these symbols their most vivid literary form, but thecultural legacy of this history is also evident in, for example,the unpleasant connotat ions associated vvith certain uses ofa word like rues." t'vloreover, philosophers and cultural criticshave generally held vision a nd visual metaphors in lowesteem through much of this cenhu-y, as Jay (1993) hasdocumented in the case of France. ln any case, it is importantto keep in mind that tJ1e surveillance model is a system ofmetaphors: in p p l j ~ n g the surveilJance model to a priva tecompany, for example, one is simply likening it to amalevolent state organization. and it ~ v i l l be important toe},:plon: the limits of this comparison.

  • 8/13/2019 Philip Agre Surveillance And Capture

    5/21

    51. Surveillanceand Capture

    The surveillance model is by far the most prevalent in theliterature on privacy. It is found. for example, in definitionsof privacy in terms of the right to be left alone, or inconcerns over informa tion being used for unintendedpurposes. Indeed, the vast majority of the existing literatureon computers and privacy employs the surveillance modelwithout critically analyzing it or considering alternatives,indexing it through the term s u r v e i l l a n c e ~ or references toBig Brother and other themes from Orwell (Burnham 1983.

    Clarke 1989. The conomist 1993. Flaherty 1989, Flynn 1993.Gandy 1993, Larson 1992. Piller 1993, Rabel1993. Robinsand Webster 1988, Rule 1974, Smith 1979, Ware 1993). Mypoint is not that this work is wrong, but rather thatalternative models might draw different, but equallyimportant elements into the foreground.

    One such alternat ive metaphor-system is the capturemodel. In naming this model. l have employed a commonterm of art among comput ing people, the verb to capture. 'Computationalist s' discourse rarely brings to t h ~ . s u r f a ~ e theconnotations of violence in the metaphor Of capture'':captured information is not spoken of as fle'eirig. e-scaping. orresenting its impri sonment. The term has two uses. The firstand most frequent refers to a computer system's (figurative)act 'of acquiring ce1tain data as input. whether fi.om a humanoperator or from an electronic or electromechanical device.Thus one might refer to a CJsh register in a fast-foodrest aurant as capturing a patron's order. the implicationbeing that the information is not simpl>' used on the spot.but is also passed along to a database. The second use ofcapture. which is more common in artificial intelligenceresearch, refers to a representation scheme's ability to fully.accurately. or cleanly express particular semantic notions ordistinctions. without reference to the actual taldng-in ofdata. Thus one might refer to the object classes of an objectoriented computer program as capturing the distinctionbetween standing orders and particular occasions on whichgoods are delivered. J)1is ambiguity between an

    e r s t ~ m o l o g i c a l idea (acquiring the data) and an ontologicalidea (modeling the reality it reflects) is remarkably commonin the vocabulary of computing. (Al researchers, for example.aPply the wor d epistemological in the second sense of''capture, not the first.)

    The capture model can be contrasted point-by-point vviththe surveillance modeL lt comprises:

    theNEWMEDI RE DER

    1) linguistic metaphors for human activities,assimilating them to the constructs of a computersystem's representation languages;(2) tl1e assumption that the linguistic parsingH ofhuman activities involves active intervention in andreorganization of those activities;(3) structural metaphors; the captured activity isfiguratively assembled from a c a t a l o { of partsprovided as part of its institutional setting;(4) decentralized and heterogeneous organization; theprocess is normally conducted \vi thin particular. localpractices which involve people in tlle workings oflarger social formations; and(5) the driving aims are not political butphilosophical. as acti\ity is reconstructed throughassimilation to a transcendent ( virtual ) order ofmathematical formalism.

    Since the capture model is less familiar than thesurveillance model. the next four sections vvill be devoted toexplaining it. The capture model, like the surveillance model.is a metaphor system and not a literal description. It can, forexample. be applied equally well to publk or plivateorganizations (or to the many activity-systems that cross theincreasingly permeable boundaries between d1ese twodomains). although my analysis will focus on workplacesettings. It is important to mal\e dear. with regard to point(5), that the capture model is a philosophical metaphor in thesame sense as the surveillance model is a political metaphor.The actual institutionaJ sites to which the capture modelmight be applied presumably have their politicaJ aspects: thepoint is simply that the capture model suggests using certainphiJosophical projects as models for understanding theactivities in these sites.

    The t\vo sets of metaphors have significantly differentorigins. Whereas the surveillance model originates in theclassically political sphere of state action. the capture modelhas deep roots in the practical application of computersystems. As such. technical developments such as thetracking schemes described in Section 2 do not bring thecapture model into existence; rather, they express in a clearway something tha t has long been implicit in appliedcomputer work. whether or not its relevance to privacyissues has been recognized.

  • 8/13/2019 Philip Agre Surveillance And Capture

    6/21

    4 Grammars of actionComputers are frequently sajd to store and transmit"information." The term, though. conceals a significantambiguity. On one hand. infmmation can be defined (as perShannon and Weaver) as a purely mathematical measure ofinformation and informa tion-can ying capacity. withoutregard for the content. On the other hand information isinformation also bout something. A similar point applies tocustomary uses of the term "data.") Although it makes senseto speak of false information (for example. in a faulty cred itdatabase), the tacit assumption is most commonl? thatinformation is true-that it corresponds in sometransparent way to certain people. places, and things in theworld. This assumption does not. strictly speaking. detivefrom any inherent property of computers. It is, rather, atheory of representation that is embedded in the \ - ay thatcomputer s have customarily been used.To see this. consider a te;.:tbook of informationmanagement such as iVlartin (1989). Martin's goal is toinstruct MIS managers on the principled construction ofinformation systems. and spedficaUy on the principledselection of v1.rhat ought to be represented. In doing this. hedescribes an onto logy of entities and relations ;md functionsand activities. along with a set of procedures forsystematically representing the e:.jsting organization in theseterms. Having prepared this self representation. rhe next stepis to implement it on a computer. The purpose of thiscomputer will be to model the organization-that is. tomaintain a set of datastructures that mirror the dayto-dayactivities of the org.:mizrttion's members. In phiJosophicalterms. the resulting computer will embody 1 correspondencetheory of representation: the machine's internal states will be"true'' (so far as this theory is concerned) because theymaintain a certain fixed set of relationpresening mappingsto the extern,1l states of affairs in the world.

    The practice of construct ing systematic representations oforganizational activities is not at al ne\V, of course. nor is itinherently tied to computer systems development. Indeed.Martin emphasizes that it can be valuable in itself. evenwithout any computers being installed. simply for theredundancies and other inefficient pattern s of activity it canbring to management's notice. As such, it clearly stands in aline of descent that includes the elaborate representationalschemes of industrial time and motion studies (Gilbreth1912, Holmes 1938 and other fmms of systematic

    51. Surveillanceand Capturerationalization of work activities (Lichtner 1924). Whenapplled to the tracking of organizational processes, of course,these schemes relied heavily on paperwork (Yates 1989 oron the intrinsic controls buUt into the movements ofmachinery (Edwards 1979).

    Besides the creation of tracking systems, systematicactivity-mapping schemes have also been applied to theautomation of activities. Coug er (1973), for example, surveysa variety of such schemes from the early days of computing,each based on tracing the flows of information vvithin abusiness. A map of these information-flows, and of theinformation-processing operations that tal;:e place along theway, could be treated as a blueprint for a computer programthat automated those same operations.

    Yet another analogous representation practice is found inresearch on "knowledge representation" in the field ofartificial intelligence. Several of the entity-relationshipdiagrams in Martin (1989: 168 ff.) resemble nothing so muchas the "semantic networks" employed in AI knmvledgerepresentation research (Brachman and Levesque 1985). AIresearchers. more than their counterparts in other kinds ofapplied computer science, set about explicitly searching forontological sy stems that would a1low a computer torepresent clean }' and accurately a wide ran ge of humanknowledge-including knowledge about human activitiesand social organizrttions.

    Despite their varied surface forms. these lines of researchtogether constitute a coherent genealogy-a tradition ofapplied representational work thrtt has informedorganizational practice the world over. Its u nderlyi ngapproach is organized and reproduced largely through itspractical conduct: its methods. its language. its paradlgms ofgood practice, its training regimens. and so forth. Although ithas become deeply identified with organizational applicationsof informat ion technology. it is (at least in principle) neither anecessary nor a sufficient condition for the use of computers.At the same time. it has grovvn such deep roots incomputationa l practice that it is hard to imagine what nyalternative computational practice would be like.

    Among the many attributes shared by theserepresentation schemes, perhaps the most significant forpresent purposes is their use of linguistic metaphors: theyeach employ formal"languages" for representing humanactivities. Human activity is thus effectively treated as a kindoflanguage itself. for which a good representation scheme

  • 8/13/2019 Philip Agre Surveillance And Capture

    7/21

    51. Surveillanceand Captureprovides an accurate grammar. This grammar spedfies a setof unitary actions the words or }e;...ical i t 5 ~ of action,which AI people call plimitives and which Quinn (1993:103-109 ff.) calls minimum replicable units. It also specifiescertain means by which actions might be compounded-most commonly by arranging them into a sequence, althoughvarious languages provide more sophisticated means ofcombination (for example, conditional and iteratedsequences).

    These gr mm rs o ction are central to the capture modeLGrammars of action have many and varied manifestations.

    Accounting systems, for example, are b;15edongrammars of action; in order to keep a set of books. itis necessary to organize one's financial acthities witha view to categorizing every move as one of the,Ktion-types that one's particular accounting schemerecognizes Telemarketers and many types of telephone-basedcustomer service personnel employ scripts that arebased on a set of standard moves, manr of whosenames are drawn from the structured patter of salespeople (e.g., 'assumptive dose"). Some grammars ofsales interaction are extraordinarily complo:., (Millerand Heiman 1987). limited,1Ccess highway (such as the roads in theAmerican interstate highway system) enforces,through both physical and legal means. a simplegrammar of action whose elements are entrances,discrete continuous segments of traveled ro.1dway.,md exits. TolJ.collection systems for such roads oftenemploy a keypunch card which contains a table formapping "grammatical" trips to collectible tolls. The user interfaces of many if not a\1 computersare readily understood as supplying Lheir users withgrammars of action. The permissible unitary actionsare ASCJJ keystrokes, menu selections, sheUcommands. and so forth. Some projects haveattempted to formalize the inte actionpatternsdiscovered in empirical study of human conver-sations, and then to build computer programs tha tcan engage in these patterns (Luff. Gilbert, andFrohlich 1990).Waiters in large restaurants frequently employ anautomated system for passing orders to the kitchenand keeping track of tabs (Rule and Attewell1989,Quinn 1992: 142-145). The waiter might interact

    theNEWMEDIAREADER

    with the system by swiping a card through a readeron the cash register and enter ing commands on atouch-sensitive display. "Enterpri;;e i n t e g r a t i o n ~ EI) systems draw an organizationS computer systems together on a globalnehvork with a standardlzed set of communicationsprotocols and data models. One proposal for an EIarchitecture (Pan and Tenenbaum 1991) breaks anorganization's work activities down into tasks, eachrepresented within a common language, and automatically evaluates which tasJ.:s should be assigned tocomputation,1l"agents" and which s b o t ~ d be : ~ s s i g n e dto human workers.

    What matters in each case is not the sequences of "inputs"' toor o u t p u t s ~ hom a given machine, but rather the ways inwhich human activities have been structured. The capturemodel describes the situation that results V1.1hen grammars ofaction are imposed upon human activities, and when thenewly reorganized activities are represe nted by computers inreal time. It is con\'enient to subdivide this process into afive-stage cycle. This division is. of course. a gl'eatoversimplification: the phases frequently operateconcurrently. ,1d1.ances in one phase n1ay force revision of thework done in an e,1rlier phase. and work in each stage d r t ~ w son a wide ranoe of sociotechnical advances not necessarilv0

    r e i ~ ted to the other stJges.Analysis Somebody studies an existing form of activityand identifies its func Jmentrtlunits in terms of someontology (e.g . entities, relations. functions, primitive actionsand so forth). This ontology might draw on the participrtnts"terms for things. or it might not Programming languagesand systems analysis methodologies frequently su pply basicontologies (objects. variables. relations) upon which domainspecific ontologies can be buUt. The resulting ontologies aresometimes standardized across whole institutions,industries. or markets.Articulation Somebody articulates a grammar of the waysin which those units can be strung together to form actualsensible stretches of activity. This process can be complicated.and it often requires revision of the preceding ontologicalanalysis. It is typically guided by an almost aesthetic criterionof obtaining a complete, closed formally specified pictu re ofthe activity.Imposition The resulting grammar is then given anormative force. The people \vho engage in the articulat ed

  • 8/13/2019 Philip Agre Surveillance And Capture

    8/21

    activity are somehow induced to organize thei r actions sothat they are readily ~ p a r s e a b l e in terms of the grammar.The s o m e h o w is typically both social (explicit proceduresbacked up by certain relations of authority) and technical(whether through m a c h i n e ~ y or simply through physicalbaniers); participants in the activity m y or may notparticipate in the process and may or may not resist it.Institutions frequently impose grammars on activities forreasons other than real-time capture-for example, forsecurity, efficiency, protection from liability, and simplecontroLInstrumentation Social and technical means areprovided. 1.vhether through paperwork or machinery, andpotentia11}' with a complex division oflabor for maintaininga runnin g parse of the ongoing activity. This phase maycoincide with the imposition phase, or it may follow by yearsor decades. Mterward. the participants begin, of necessity, toorient their acti'ities toward the capture machinery and itsinstitutional consequences.Elaboration The captured activ ity records. which are ineconomic terms among the products of the reorganizedactidty. can nm be stored. inspected. audited. merged withother records. subjected to statistical analysis. employed asthe basis of Pareto optimization. and so forth. Likewise.concurrent computationaJ processes can use captured recordsto "watch t h ~ ongoing activities for purposes of errordetection. advice giving. per formance measurement . qualitycontrol. and so forth. These additional processes might arisesimultaneously with the instrumentation phase, or they mayaccumulate long afterward.This cycle is normally attended by a kind of mythology,according to tvhich the newly constructed grammar of actionhas not been invented"" but '"discovered." The activity inquestion. in other '''ords. is said to have already beenorganized according to the grammar. Of course this is notwholly false: imposing a grammar that radkaJJy andarbitrariJy misrepresents the activity will probably lead tocalamity. But even when a grammar of action is relatively"good in this sense. its imposition \Nil generally require hardwork both for the people who are imposing the grammarand the people upon whom the grammar is imposed. Thework of these latter participants consists in part of findingways to organize one's activities, even in the tricky andexceptional cases, so that they can be parsed within suchand-such a vocabulary of discrete units.

    51. Surveillanceand CaptureIndeed, it is crucial to appreciate the senses in which the

    imposition and instrumentation phases constitute areorganization of the existing activity, as opposed to simply arepresentation of it. Let us distinguish eight such senses, inincreasing order of significance for the current argument:1) The introduction of new technologies. whether

    they involve the capture of activities or not. isfrequently the occasion for a wide variety of otherkinds of changes to the activity, for example due toextrinsic economic changes (e.g., Iacono and Kling1987). Indeed technological change is generallyinseparable from broader social changes.(2) The representations constructed in thearticulation phase (based to some extent on empiricalstudy of the activity. but mostly on informalspeculation and scenario-making) and then in theelaboration phase (based on the newly accumulateddatabase of parsed activity) frequently suggestrearrangements of the activity (Quinn 1992. Taylor1923). Some of these rearrangements may bedesigned in part to facilitate the capture process, as inHammer's (1990: 112) dictum, "Capture informationonce and at the source:{3) Grammars of ;" Ction frequently oversimplify theactivities they .1re intended to represent, if onlybecause the people who articulate the grammars areonly superficially acquainted with its actualcomplexities and the actual social forces thatdetermine its form (Suchman and Jordan 1989). Theontology may fail to make enough distinctions. or elsewhole subcategories of ""invisible" activity might gounrepresented. The grammar might impose overlyrestrictive ordering constraints on the unitaryactions. it might neglect the interleaving of distinctforms of activity. or it might mistake prescribedprocedures for an accurate descriptive account (or atleast a practicable form) of the activity (Suchman1983). As a result, the participants in the newlyinstrumented activity will find it necessary to evolve asystem of ""work-arounds"" to keep things going(Gasser 1986).(4) Grammars of action can be ""mistaken"" in otherways. Most especially, ther can encode asystematicaJJy distorted conception of the activity. Forexample, Kling (1991) argues that extant computersupported cooperative work (CSCW) S}'Stems arebased on ontologies that recognize cooperation but

  • 8/13/2019 Philip Agre Surveillance And Capture

    9/21

  • 8/13/2019 Philip Agre Surveillance And Capture

    10/21

    radicalize it through the application of philosophicalconcepts (Winograd and Flores 1986). The idea is cl1at adeeper-than-normal ontology can provide a firmer and moreaccurate basis for system-butlding. Winograd and Flores taketheir inspiration for this project from the existentialhermeneutics of Martin Heidegger's eing and Time (1961[1927]). Heidegger's project was to employ successive cyclesof phenomenological description to uncover successivelydeeper layers of ontological structure in human e>..-perience,eventually yielding some kind of authentic engagement tviththe ultim te ontological category of Being itself.

    For Winograd and Flores, this method promises a kind ofultimate authority to their project of pliori clatification(Suchman 1993). Human activities, they argue, go astraywhen they depart from the essential structures that rigorousphenomenological analysis reveals, and computer-mediatedtools can prevent such mistakes by imposing particular gram*mars of action upon their users. Although this idea isaltogether natural within the tradJtions of computer science,and while VVinograd and Flores' philosophy is alert to someof the oversimplifications implicit in conventionalcomputational practices, Heidegger would have been aghastat the idea of formalizing ontologic,1l categories incomputational (or othenvise mathematical) terms, much lessemploying machinery to enforce compliance with th('ffi.

    Winograd and Flores' ontology, moreover, has l ittle to dowith Heidegger's, being drawn principalJy from the quite unHeideggerian theory of speech acls (Searle 1969). In theirdesign for The Coordinator, they provide a grammar oflinguistic action in conventional state-graph notation. Usersexchange electronic messages in conducting their work andthey are supposed to label each message with a particularspeech act. The system, meamvhile, can capture the speechact structure of each sequence of interactions. Althoughsome research groups have presented equivocal evaluationsof The Coordinator's success in practice e.g., BulJen andBennett 1990), it is not my purpose to argue that suchsystems cannot work. Quite the contrary, I wish to portrayThe Coordinator and its more sophisticated successors(Marshak 1993, Medina-Mora et al1992) as deeply rootedin a social and technical tradition. Although computersupported cooperative work systems such as TheCoordinator require thei r designers to perform particuJarlyrigorous ontological work in the analysis and articulationphases cf. Clawson and Bostrom 1993), this work is no

    51. Surveillanceand Capturedifferent in kind from the generations of systems analysisthat have gone before it.

    To summarize, the phenomenon of capture is deeplyingrained in the practice of computer system design througha metaphor of human activity as a kind of language. VVithinthis practice, a computer system is made to capture anongoing activity through the imposition of a grammar ofaction that has been articulated through a project ofempirical and ontological inquiry. Advances in computerscience have thus gone hand-in-hand with ontologicaladvances. Furthermore, the phenomenon of capture alsounderlies the tracking systems discussed in Section 2.Tracking is impossible without a grammar of states andstate-changes and the technical means to detect the states(or the state-changes) when they occur. Except in thesimplest cases, this will require that the grammar beimposed through one means or another: location trackingdevices, paperwork ident ity cards, and so forth. Theresulting technology of tracking is not a simple matter ofmachinery: it also includes the empirical project of analysis,the ontological project of articulation, and the social projectof imposition.5 Capture and functionalityA variety of projects, particularly in the participatory designmovement (Schuler and Namioka 1993}, have sough talternatives to the engineering strategy of thoroughgoingcapture. through schemes that allow people to recordinformation in the form of computerized text (and in other

    o m p t ~ t e r i z e d media as well) without imposing any detailedgrammar on it. The stored materials can later be retrievedand interpreted by others. Simple electronic matl andhypertext systems work this way. as do certain moresophist icated systems (MacLean, Young, and Moran 1989).

    But these systems all participate in a trade-off that goes tothe core of computing: a computer at least as computersare currently understood can only compute with what itcaptures; so the less a system captures. the less funct ionali tyit can provide to its users. To understand this trade-off.consider the contrast beh...,een voice mail and electronic mail.Both media are routinely emplo}'ed to convey stretches oflanguage from one person to another, as well as a variety ofother functions: storing the messages, reviewing them later,replying to them, attaching timestamps and labels to them,and so on. Nonetheless, they capture different aspects of the

  • 8/13/2019 Philip Agre Surveillance And Capture

    11/21

    51. Surveillanceand Capturelanguage: whereas voice mail captures spoken language at thelevel of sampled frequency spectra. electronic mail captureswlitten language at the level ofASCII keystrokes. Eachmedium thus has capacities that the other does not: voicemail. unlU\:e email, can transmit singing and languageswithout Latin orthographies. and email archives. unlikevoice mail can be searched by keyword. Finally, neithermedium captures the grammatical structure of the messagesit transmits, much less anythin g about the content of thosemessages. Thus neither medium can offer features based onautomatic recognition of agramrnaticality, urgenq orrelevance to a given topic. (It may be possible to heuristicallyinfer such matters from e-mail messages, but systems fordoing so are far from practical at this moment.) Someanalogous examples include the contrast benveen paintingand drawing programs, and between ASCII-based te.\. t editorsand \NYS[WYG word processors.

    This trade-off is also found in systems for tracking humanactivities through automatic capture. Simply put. a systemcan only track what it can capture, and it can only captureinformation that can be expressed within a grammar ofaction that has been imposed upon the activity. t\umeroussystems, including many of the examples in Section :Z r e s i ~toward the minimal end of this trade-off since they onlytrack simple position information. Systems like these are notparticularly convincing cases of the capture model since theydo not usually require much imposition beyond theinstallation of the tracking instruments themselws. Butposition tracking is frequently a precursor to morequ,1litatively complex kinds of capture, for example whenpositional information is stored along with other e\'ents ortransactions that might be captured: arrival at a certaindestination, crossing a certain boundary, changes in thestatus of materials or participants. encounters \\ith otherparticipants. and so forth.

    The inherent trade-off of computer systems for capturinghuman activities underlies the most significant technicaltrend in their ongoing historical development: the tendencytoward v r deeper'' articulation and capture of activities. AsQuinn (1992: 104 ff. has put t in the case of businesses inservke industries. the minimum replicable unit has gottensteadily smaller:

    Early in the life cycle of many service industries. thesmallest truly replicable unit seemed to be anindJviduill office. store. or franchise location. Later. as

    theNEWMEDI RE DER

    ;:olume increased, it often became possible for cl1eheadquarters to develop and replicate greaterefficiencies within locations by managing andmeasuring critical performance variables at individualdepartmental. sales counter, activity, or stock-keepingunit (SK'U) levels Then the successful reduction of l

  • 8/13/2019 Philip Agre Surveillance And Capture

    12/21

    through habits of selective perception and inquiry. As KJJngand Dunlop (1993) have pointed out, analysis of the place ofcomputer technology in society has often been i m p o \ ~ i s h e dthrough a bifurcation into t\\'o structurally opposed genres,which they cal1 utopian and anti-utopian. The utopian genreas its name suggests, emphasizes good things: efficiency. theamplification of various professions' powers, and otherbeneficial consequences of computing. The anti-utopiangenre, for its part, draws on a stock of cuJturaJ images of classconflict and totalitarian domination. Both genres areprevalen t in journalistic and academic "vriting alD;:e.

    In the parti cular area of workplace computing. one srrandof anti-utopian writing is found in union-oriented criticismsof managerial control imperatives (Garson 1989, Ho\\ard1985. Shailcen 1985: cf. Rule and Attewell1989). Theargument. first formalized by Braverman 1974) butpossessing deeper historical roots in the American unillnmovement. was originally motivated by real historicalconflicts over production knowledge, which consisted o theappropriation of craft knmvledge through scientificmanagement and the replacement of craft work-ways \::rhfragmented and riltionalized forms of work enforced tl1 'L1ughdirect surveillance and control. An extensive school ofthought has generalized this experience into a theory rhehistorical development of

  • 8/13/2019 Philip Agre Surveillance And Capture

    13/21

    51. Surveillanceand Capturelocation, the capture model is agnostic on this matter. Indeed.the capture model emphasizes that capture. as a speciflcallysocial process, is not a unitary phenomen on. To the contrary.every domain of activity has its own histmicallogic, its ownvocabulary of actions. and its mvn distinctive social relationsof representation. As a result. information gathered th roughcapture in one domain of activity may or may not becommensurable with information captured in anotherdomain. Even \ \ ~ t h o u t this element of unification andcentralization. though. this pictttre of a tot?lly capturedsociety offers plenty of opportunit y for utopian and antiutopian s p e c u l r ~ t i o n . In particular. it has a millenarian flavorof perfect transparency of correspondence between digitalrepresentations and the now fully ordered affairs ofembodied activity

    Nonetheless. this picture is wholly unsatisfactory. since itprovides no notion of the larger social dynamics that captureprocesses \\ill panidpate in or inte 'ilCt with. Indeed. withouta workedout conception o f ho\': re :> l activities might actuallybe reorg;mized during the various phases of the capture cyckthis sketch of a h ~ p o t h e t i c a l \\odd of total capture is hard tndistinguish in an:: convincing way from the utopins

  • 8/13/2019 Philip Agre Surveillance And Capture

    14/21

    sequences of action exist in the first place. The generalpicture of empowerment and measurement is consistentvvith a wide range of power relations, from the intensiveproduction pressures placed on fast food workers bycentralized monitoring of captured information to therelatively gentl e bureaucratic negotiations e>.-perienced bydoctors dealing wit the medical-activity capture schemes ofhospitals, health maintenance organizations, and insurancecompanies, In particular, the measurements that are derivedfrom representations of captured work can be put to avariety of uses, including piece-rate pay and periodicadjustments of work methods. The ult imate use of suchmeasurements is the establishment of bidding for services inreal-time markets, whereby the control function previouslyprovided by bureaucracy is transferred to the inherentdiscipHne of the market A useful theory of the capturemodel's place in society would provide a way ofunderstanding when the capture model is employed. how itis employed. and how its employment affects relationshipsamong people. A great deal of work is required before such atheory can be formulated, but at least some plausible startingpoints are available.7 The political economy of captureThese general considerations provide the background for ananalysis of capture as an institutional phenomenon. Whileinstitutions deserve analysis on numerous levels, I propose tofocus on Ciborra's (1983. 1987) economic mode of theinsti tutional effects of information technology. The analysisin this- section will be considerably more speculative than inprevious sections, and should be understood as the outline ofa research agenda.

    Ciborra's model is based on Cease's (1937) notion of~ t r n s c t i o n costs.H Coase begins by looking at ll productivehuman interactions as economic exchanges. and then he askswhy some such relations are organized through marketmechanisms and others are organized within the authorityrelationships of hierarchical firms. His ansvver. roughlyspeaking, is that the boundaries of firms are drawn aroundtransactions which are less costly to perform within ahierarchy than they are to perform in a market. Transactioncosts, which are the costs associated with the use of marketexchange. include the costs of locating and evaluating thevarious goods and services for sale in the market, definingthe precise nature of the goods and services to be exchanged,

    51. Surveillanceand Captureand negotiating and enforcing the contracts that govern theexchange. (For the subsequent h istory of the theory oftransac tion costs, see Williamson and Winter (1991).)

    Markets work over time to reduce transaction costs, forthe simple reason that competition tends to reduce ll costsof production. For example, improved computing andcommunications technologies make it easier to collect andanalyze information on offerings available in the market. Astechnological changes permit decreases in transaction costs,the theory predicts that the boundaries of firms and thecontractual basis of various economic arrangements willchange as well in p r t i c u l ~ r ways. For example. firms maybegin t purchase certain goods and services on the openmarket instead of organizing their production in-house, andpatterns of vertical integration may change as it becomesrridre efficient to coordinate certain institutional interfacesthrough market mechanisms rather than bureaucraticorganization.

    Ciborra (1987: 258-260 applies this themy to the role ofinformation technology in organizational change. FollowingOuchi (1979), he extends Coase's framework slightly bydistinguishing three organizational types-markets.bureaucracies, and clans-depeniling on the degree ofambiguity or uncertainty p resent in a given tvay of producingand selling a given produc t or service. and on the degree ofcongruence among the interests of the various parties to theinteraction. According to this analysis, an economicrelat.ionship \ \rill be organized as a market v.rith completelyseparate transactions on every occasion of exchange) whenambiguity and uncertainty are low and the interests of theparties are in conflict; it will be organized as a rule-boundbureaucratic organization when ambiguity and congruity ofinterests are moderate; and it will be a closely knit, informal"dan" when ambiguity and congruity of interests are high.(Numerous intermediat e and hybrid forms are found aswell.) Formulated in this way, the theory specifically predictsthat. as transaction costs are reduced, industries willdemonstrate a historical trajectory in the direction fromclans to hierarchies to markets.

    If true, this theory ought to be invaluable to managersfaced with planning the organizational concomitants oftechnological change. With the transition &om dan tohierarchy, or hierarchy to market. or with other significanttransitions in work-organization within each of thesecategories, the theory of transaction costs prescribes in some

  • 8/13/2019 Philip Agre Surveillance And Capture

    15/21

    51 Surveillanceand Capturedetail the economically most efficient and stable contractualform that can be given to the social relationships in the newform of work In particular, Ciborra (1987) argues, shategiesfor designing and managing infonnation technology oughtto depend on ,.vhich category of economic relationships ispresent (that is the relationships that will e present oncethe system is working).

    VVhen applied in accordance vvith the capture model.information technology reduces transaction costs byimposing more clearly defined-less ambiguous and lessuncertain-relationships upon the parties to economicinteractions, thereby decreasing the overhead associatedwith coordination of various individuals' activities. Morespecifically, once a grammat of action has been imposedupon an activity. the discrete units and individual episodesof the activity are more readily identified. verified. counted.measured, compared. represented. tearranged. contractedfor, and evaluated in terms of economic efficiency This is aparticularly simple matter when the interactions in questionare already organized by market relationships, and Ciborraconjectures (1987: 263) that market transactions ratherthan bureaucratic firms are at present the main field ofapplication of DP technology. since the structured andstandardized nature of those transactions make them moresuitable to automation. Indeed som of the mostspectacular applications of information technology arefound in the operation of global markets in stocks.commodities. currencies . nd derivatives built upon thesethings Kurtzman 1993). and this increasingly includesgeneralized markets in debt streams of all sorts Lenznerand Heuslein 1993)

    But information technology can also reduce informationcosts when applied to work activities in bureaucracies andclans, perhaps leading these activities to change theireconomic organization. When designed and introduced inaccord with the capture model. through the use of anontology and grammar of social interaction. computersupported cooperative work (CSCW) tools are particularlywell suited to this purpose. The grammars which such toolsimpose upon an organization's activities necessarilystructure, to some substantial extent, the relationshipsamong the organization's members (CibotTa and Olson1988). In particular. Ciborra (1987: 263) recommends furthis purpose the framework of speech acts andcommitments (Flores and u d l o ~ v 1981) later employed by

    theNEWMEDI RE DER

    the Coordinator (see above, Section 4). (As a general matter,of course, such systems are more readily implemented whensupporting activities that are already organized through agrammar of action. The point is simply that they may prove

    s u i t ~ b l e for less structured activi ties as well.) Once thesequalitative structures of work interactions have beenformalized and successfully submitted to automatic trackingand enforcement, it costs less to coordi nate them all.

    The analysis of transaction costs has politicalconsequences for the design of computer systems thatsupport cooperative work For example. Ciborra and Olson1988) assert that. in reducing transaction costs, new

    technologies can be designed to reinforce clanlike structuresas opposed to creating economic pressures for a transitionto hierarchical or market structures. But this is contrary toCease's (1937) original argument. which is that the use ofmarkets or hierarchies (or, br extension. clans) isdetermined by their relative costs. with nonmarkctorganizations savings in transaction costs being balancedagainst their comparati ve economi c inefficiencies.(Incidentally. eLms should not be confused v ~ r i t h thecontemporary phenomenon of frequently reorganizedmultifunctional teams. which are. economical ly speaking.really a kind of internal labor market.) By this logic,technologies that reduce transaction costs wilL other thingsbeing equal. necessarily shift the balance in the direction ofmarket relations. But, as Co;Jse points out. new technologiescan also reduce the costs of organizing (he cites thetelegraph and telephone. though the point clearly applies tocomputer networks as \vell), thus potentially preserving oreven extending the economic scope of firms even in the faceof reduced transaction costs. The broader point hasconsiderable significance for designers vvho wish toencourage clan-like forms of sociJ\ relations as opposed tomarket or hierarchicJl relations: a focus on the reduction oftransaction costs \vill not serve this goal.

    These propositions on trJnsaction cost economics,together with the foregoing analysis of captttre, suggest arudimentary theory of the political economy of capturedinformation. To place this theory in perspective, it will helpto consider Schiller's 1988) analysis of information as acommodity. Schiller argues against deriving the economicproperties of information simply from its inherent qualities(its lack of inherent physical form. ease of duplication. andso forth). Instead. he asserts that the specific historical form

  • 8/13/2019 Philip Agre Surveillance And Capture

    16/21

    of information depends on its embedding in a particular setof social relationships. Information, in particuJar, hasincreasingly become a commodity. that is, somethingproduced, exchanged, and used within the framework of amarket economy. Indeed, as a rapidly e:.:panding sector ofthe market, i nformati on has become a fundamental sourceof growth for the market economy as a whole (1988: 27).Information, though. has not always been a commodity inthis strict sense. and Schiller points to the historical processby which the production of informatio n has become amarket-based industry largely comparable to any other, aprocess that has involved the progressive reorganization ofthe human activities through which information isproduced and used.

    The economic theory of capture presented here makes itpossible to extend Schiller's analysis in the case of oneconsiderable category of information commodities. namelycaptured information. Regardless of its particular content.captured information is cllstinguished by its dualrelationship both product and representation to thehuman activities upon which particular grammars of actionhave been imposed In particular. the capture process makes"visible" a great deal of infom1ation-creating activity whichhnd formerly been left impHcit in the production of other,historically prio r commodities. Moreover, the phen omenon ofcapture extends market relations not simply through thecommodification of the captured information itself (if in factthat information is marketed). but also through themovement toward market relations. through a reduction intransaction costs, of the human activities that theinformation represents. In othe r words, by imposing amathematically precise form upon previously unformalizedactivities. cap ture standarcllzes those activities an d d1eircomponent elements and thereby prepares them (again, otherthings be ing equal) for an eventual transition to market-basedrelationships. This transition is not a mechanical process, tobe sure; attempts to impose grammars of action upon existingforms of activity are themselves forms of activity pursued byfully blown human agents, and they regularly fall prey to tech-nical or economic miscalculations, or to the resistance of theparticipants. The tendency of information technology tocontribute to the spread of market relations into previouslyhierarchical or informal territories of activity shouJd thus beunderstood as cl1e historically contingent confluence of adisciplinary practice and an economic ~ l a w , on the same basis

    51. Surveillanceand Capture

    as the mutua accommodation of supply and demand inperfect markets.

    That said. the role of information technology in thegeneralization and e.'i:tension of market relations isformidable and not to be underestimated The process isextraordinarily systematic. At the level of the professionalpractice of computer people it takes the form of a kind ofrepresentational crusade the conscious formulation of athoroughgoing srstem of ontological categories for the ullrange of productive activities, at every level from the globaleconomy as a whole to the most refined unitary action. Nomatter how forbidding their discursive forms may be and nomatte r how esoteric much of their specific content routinelyis, these ontological schemes must nonetheless beunderstood fundamentally not as "technical" but as "social"In other words, the practice of formulating these ontologiesis, all disciplinary insularity aside, and regardless of anyone'sconscious understandings of the process. a form of socialtheorization. And this cl1eorizing is not simply a scholasticexercise but is part of a much larger matelial process throughwhich these new social ontologies. in a certain specH1c sense,become real

    The rele,ant sense of .reality" must be defined wi th care,since, as Section 4 has argued at length, the articuJation andimposition of gramma1s of action routinely involves a kindof mythology: the idea that the activity in question hasalready been organized in accord , ,rith the grammar, and thatthe subsequent capture scheme simply reads off. in real time,a representation of this pree>..isting forma structure. Thiskjnd of mythology is frequently associated with theconstitution of novel commodities. and may even help definethe commodity-form as a social phenomenon. Indeed. thetheory of transaction costs exhibits the same form ofmythology. inasmuch as it presupposes that the entire worldof productive ac tidties can be conceptualized, priori interms of extremely numerous episodes of exchanges amongeconomic actors.

    The truth. of course, is more complicatecl Theintroduction of capture systems into e;...isting activitiesrequires a great deal of effort, including not simply thetechnical \mrk of buikLng and instaliJng the system but thesocial work of imposing the system and t hen IJving ... .rith it. Inparticular, the \Vork of imposing a capture system frequentlyinvolves conflict, as the affected parties organize resistanceto it and its beneficiaries organize to overcome, dissolve, or

  • 8/13/2019 Philip Agre Surveillance And Capture

    17/21

    51. Surveillanceand Capturecircumvent this resistance (Agre in press). Normally theseconflicts take place within the context of e..xistingorganizational structures, but if the transaction cost analysisis any guide then many of these conflictswill become largelymoot as the contest ed social relat ionships move increasinglytoward the market. The growth in temporary employment(Negrey 1993, Sacco 1993) and the trend toward outsourcingof noncore functions Quinn 1992) may be, at least in part.one reflection of this movement. Be this as it may, a rapidlygrowing literature is exploring the pot entiall y considerablestructural changes to firms and markets in whichinf01mation technology may participate (Davis 1987, Quinn1992, Scott Morton 1991).

    The analysis in this section, once again, should not beunderstood as a finished theory but as a conjectural outlineof a program of research. Lest the theory be overgeneralized,several qualifying points, already implicit in the argumentabove, should be emphasized. Informat ion technology is notsynonymous with the capture model at least not inprinciple), the application of information technology canhave other consequences besides the reduction of transactioncosts, and reductions in transaction costs do not necessarilyinduce transitions to market relations if other, countervailingfactors are present. Changes that reduce the costs of sometransactions may be accompanied by, even linked to, otherchanges that simultaneously increase the costs of othertransactions. (Indeed. Allen (1994) suggests that increasedintegration of production processes requires new, lessroutinized kinds ofrelationships among the people involved)Applications of information technology are invariably accompanied by other developments and other agendas that caninfluence the shape and consequences of narrowlytechnological changes. Finally, all of these phenomena aresubject to contestation on a wide variety of fronts.

    These qualifications having been stated. the hypothesisseeking validation can be formulated in t he largest possibleterms: the compute r practitioner's practice of capture isinstrumental to a process by which economic actors reducetheir transaction costs and thereby help transformproductive activities along a trajectory towards anincreasingly detailed reliance upon (or subjection to) marke trelations. The result is a generalized acceleration of economicactivity whose soci l benefits in terms of productiveefficiency are dear enough but whose social costs ought to bea matter of concern.

    theNEWMEDI RE DER

    8 ConclusionThe previous sections have outlined a political economy ofworkplace privacy, building on an analysis of the professionalpractice of computer people. This discussion provides someresources for a more careful consideration of the relationshipbetween the two models of pri\ 2cy, the surveillance modeland the capture model. that I introduced at the outset. Let usreview these models' respecti\'e definitions, recalling onceagain that they are intended as metaphor-systems and not asmutually exclusive categories:

    1) The surveillance model employs visual me taphor s.most famously OmreU s Plg Brother is watching o u ~the capture mode l employs linguistic metaphors bymeans of va1ious grammars of action.2) The surveillance model emphasizes nondlsruptive,

    surreptitious data collection; the capture mode ldescribes the readily apparent instrumentation thatentails the reorganization of existing activities.3) The surveillance model is concerned to mark off a

    p r i v a t e ~ region by meall5 of territorial metaphors ofinvasion and the like: the capture model portrayscaptured acthities as bcingconstructed in real-timefrom a set of institutionally standardized partsspecified by the captured ontology.4) The surveillance model depicts t he monitoring of

    activity as centrally organized and presumes that theresulting information is centrally stored; the capturemodel emphasizes the locally organized nature of contests over the captur e process and their structuring\Vithin particular institutional contexts.(5) The surveillance modd takes as its prototype themalevolent political actiltities of state organizations:the capture model takes as its prototype the quasiphilosophical project of ontological rec onstruc tionundertaken by computer professionals in privateorganizations.

    The body of the paper has introduced a reasonablysubstantive theory of capture as part of the historicaldynamics of a market economy. This theory does notpretend to cover ll uses of commodified information, and itwould be worth e.xploring the possibilities of a parallel theoryof information formed into commodities through processesbetter understood through the surveillance modeL Such atheory is available in the work of Gandy (1993), whoemphasizes the now v st market machinery around the

  • 8/13/2019 Philip Agre Surveillance And Capture

    18/21

    personal information that people leave behind in a widerange of public records and economic transactions. tv1uch ofthis information, no doubt, alises in the first place throughthe capture of activities of valious kinds. One possibility isth t market pressures of various sorts tend to induce atransition in the manner in which information is collected.away from the surveillance model and toward the capturemodeL Such a trend, if it exists. will presumably be mostmarked in workplaces, where the relations of powernecessary to impose grammars of action are the most fullydeveloped But the accumulation o f personal informationthrough medical care, the contractual conditions of insurancecoverage, and driving on increasingly inshumented publicroads (Bender 1991, Jurgen 1991) also provides promisingsites of investigation along these lines.

    Some additional topics invite further research. It would bevaluable to catalog the kinds of organizationaltransformations that can accompany the imposition ofgrammars of action. Capture, and particularly the sharing andstandardization of ontologies, may provide a vocabulary forexploring some of the interlocking. overlapping, and r o s s ~fertilization among various forms of computer-mediatedwork that are evolving thin the global economy (Rosenberg1982). The processes of articulation and imposition should bestudied empirically in a va1iety of settings. particularly n t hregard to the forms oCparticipation that they exhibit. Thegenealogy of the capture model should be sought in thehistory of ideas an d in the hjstorical development of thecomputer profession and its practices. The transaction costmodel of capture economics should be evaluated andextended with reference to detailed case studies.

    The analysis of the capture model has significantimplications for designers. It provides some tools for placingtechnical design-styles in larger political and economiccontexts, and thereby for more consciously setting researchpriorities in accordance \vith democratic g o < ~ J s . This analysismight also provide some impetus for investigations of theunderlying struc tures of design practices, and it might providea prototype for research into the political and economicdimensions of various specific fmmatlons of design. Finally. itwould seem important to articulate various countertraditions of design and their associated counter-visions ofhuman activity, keeping in mind the trade-offs that arestubbornly inherent in computers and computational designas these things are currently constituted.

    . ..... VCII LCand CapAcknowledgementsThis paper originated in comments that I prepared as a discussanat the Symposium on Basic Research Topics at the Conference onComputerwHuman Interaction in />\ay 1992 in / ~ o n t e r e y California.Thanks to John Carroll and Jim Hollan for their roles in organizinthis meeting. I presented a subsequent version at the ThirdConference on Computers Freedom and Privacy in t ~ a r c h 1993 inSan Francisco. Thanks to numerous participants in this conferencfor their useful comments. The paper has also benefitted fromcomments by Jonathan Allen Rick Crawford Bruce Donald /

  • 8/13/2019 Philip Agre Surveillance And Capture

    19/21

    51. Surveillanceand CaptureJohn t ~ . Carroll and Thomas P t ~ o r a n , Introduction to this specialissue on design rationale, Human-Computer Interaction 6{3-4),1991, pages 197-200.John M Carroll, Creating a design science of human-computerinteraction, in Alain Bensoussan and Jean-Pierre Vedus, eds.,Future Tendencies in Computer Science, Control. and AppliedMathematics: 25th Anniversary of INRIA, New York: Springer Verlag,1992.Noam Chomsky Problems of Know/edge and Freedom: The RussellLectures, New York: Pantheon 1971.Claudio U Ciborra, Information systems and transactionsarchitecture, International Journal of Policy Analysis andInformation Systems 21(3), 1983, pages 145-160.Claudio U. Ciborra, Research agenda for a transaction cost approachto information systems, in Richard J. Boland, Jr. and Rudy A.Hirschheim, eds .. Critical Issues in Information Systems Research,Chichester, UK: Wiley, 1987.Claudio U. Ciborra and f ~ a r g r e t h e H. Olson, Encountering electronicwork groups: A transaction costs perspective, Proceedings of theConference on Computer-Supported Cooperative Worl:, September26-29, 1988, Portland, Oregon, pages 94-101.Roger A Clarke, Information technology and dataveH\ance,Communications of the ACM 31(5), 1989, pages 498-512.Victoria K Clawson and Robert P Bostrom, Facilitation: The humanside of groupl iore, Paper presented at the Groupware '93Conference, San Jose, California, 1993.Andrew Clement, Office automation and the technical control oinformation wor\:ers, in Vincent 1\osco and Janet Wasko, eds .. ThePolitical Economy oj Information, 1\adison: University of WisconsinPress, 1988.Ronald H Coase. The nature of the firm, Economica NS t,, 193i,pages 385-405.J. Daniel Couger. Evolution of business system developmenttechniques, Computing Surveys 5(3), 1973, pages 167-198.Richard tl. Cyert and James G t\arch, A Behavioral Theory of theFirm, Englewood Cliffs, NJ: Prentice-Hal\, 1963.Stanley t

  • 8/13/2019 Philip Agre Surveillance And Capture

    20/21

    Century French Thought, Berkeley: University of California Press,1993.H. Thomas Johnson and Robert 5. Kaplan, Relevance Lost: The Riseand Fall oj Management Accounting, Boston: Harvard BusinessSchool Press, 1987.Ronald K Jurgen, Smart cars and highways go global, IEEESpectrum 28 5), 1991, pages 26-36.Rob Kling, Social analyses of computing: Theoretical perspectivesin recent empirical research, Computing Surveys 12 1), 1980, pages61-110.Rob Kling and Suzanne Iacono, Computing as an occasion for socialcontroL Journal oj Social Issues 40(3), 1984, pages 77-96.Rob Kling, Theoretical perspectives in social analysis ofcomputerization, in Zenon W. Pytyshyn and Liam J. Bannon,Perspectives on the Computer Revolution, second edition, Norwood,NJ: Ablex, 1989.Rob Kling, Cooperation, coordination and control in c o m p u t e r ~supported work, Communications of the ACN 34(12), 1991, pages83-88.Rob Kling and Charles Dunlop, Controversies about computerizationand the character of white coHar worklife, The Information Sodety9 1). 1993, pages 1-30.Joel Kurtzman, The Death of Money, New York: Simon and Schuster,19g3.Erik Larson, The Naked Consumer: How Our Pn'vate U1es BecomePublic Commodities, New York: Henry Holt, 1992.George lawton, /lacs hit the road to keep trucks on track, help cutcosts, NocWEEf: 6(41), 16 November 1992, pages 32-33.Ed Leibowitz, Halt: Op;;n architecture detains prisoners underhouse arrest and ~ : e e p s tabs on parolees, Teleconnect 10 11), 1992,pages 36-38.Robert lenzner and William Heuslein, How delivatives aretransforming Wall Street. Forbes 151 7), 29 March 1993, pages62-72.William 0. Lichtner, Planned Control in Manujoctudng, New York:Ronald Press. 1924.Stephen loudermill:, Toronto tracks vehicles over wireless net, PCVee : 10(6), 15 February 1993, pages 47-48.

    Paul luff. Nigel Gilbert. and David Frohlich, eds., Computers andConFersation, london: Academic Press, 1990.Ronni T. t ~ a r s h a k Action Technologies workflow products, Wor :f/owComputing Report 16(5), 1993, pages 1-20.James tlartin, Suategic information Planning Methodologies, secondedition, Eng e.iOOd Cliffs, 1/J: Prentice Hall, 1989.John P. / ~ c P a r t l i n Press 1 for cheeseburger: IS chiefs scramble foradvantage in the fast food fracas, Information Week 374, 18 /lay1992, page 78.Raul lledinaf.lora, Terry Winograd, Rodrigo Flores, and FernandoFlores, The action workflow approach to workflow managementtechnology, Proceedings of CSCW-92, Toronto, Ontario, 1992, pages281-288.

    51. Surveillanceand CaptureKenneth Heyer, Hugh L Applewhite, and Frank A. Biocca, A surveyof position trackers, Presence 1 2), 1992, pages 173-200.Peter A. D. Mill, Volker Hartkopf, Vivan Loftness, and PleasantineDrake, The challenge to smart buildings: U s e r ~ c o n t r o U e d ecologicalenvironments for productivity, in David V Gibson, Georg eKozmetsky, and Raymond W. Smilor, eds., The TechnopofisPhenomenon: Smart Cities, Fost Systems, Global Networks, Lanham,MD: Rowman and littlefield, 1992.Robert B. Miller and Stephen E. Heiman, Conceptual Selling, NewYork: Warner, 1987.Cynthia Negrey, Gender; Time, and Reduced War/:, Albany: StateUniversity of New York Press, 1993.Richard R. Nelson and Sidney G. Winter, An Evolutionary Theory ofEconomic Change, Cambridge: Harvard University Press, 1982.Wanda J. Orlikowski, Learning from Notes: Organizational issues ingroupware implementation, The Informab'on Society 9 3), 1993,pages 237-250.William G. Ouchi. A conceptual framework for the design oforganizational control mechanisms, Management Science 25 9),1979, pages 838-848.Jeff Y . ~ C . Pan and J. Marty Tenenbaum, An intelligent agentframework for enterprise integration, IEEE Transactions on Systems,Man and Cybernetics 21 6), 1991, pages 1391-1,08.Charles Piller, Privacy in peril: how computers are making privatelife a thing of the past, Macworld 10 7), 1993, pages 121,-130.James Brian Quinn, Intelligent Enterpdse: A Knowledge and ServiceBased Paradigm for Industry, New York: Free Press, 1992.Timothy R. Rabel, Software and privacy: revising OrweH, IEEESoftware 10 3), 1993, pages 92-93.Kevin Robins and Frank Webster, Cybernetic capitalism:Information, technology, everyday life, in Vincent Mosco and JanetWasko, eds., The Political Economy o Information, Madison:University of Wisconsin Press, 1988.Lura K Romei, Software helps find the applicants you need, NodernOffice Technology 36 8), 1991, pages 52-53.Nathan Rosenberg, Technological interdependence in the Amen'coneconomy, Inside the Block Box: Technology and Economics,Cambridge: Cambridge University Press, 1982.James B Rule, Private Lives and Public Surveillance: Social Controlin the Computer Age, New York: Schocken Books, 1971.James Rule and Paul Attewell, What do computers do?, SocialProblems 36 3), 1989, pages 225-241.Charles F. Sabel, Work and Politics: The Division o Labor in Industry,Cambridge: Cambridge University Press, 1982.Tony Sabetti. RF/ID creates dynamic links between people, objects,and processes, I&CS Instrumentation Control Systems) 66{10),1993, pages 41-44.Samuel R. Sacco, All employment relationships are changing,Modern Office Technology 38 5), 1993, pages 47-48.Dan Schiller, How to think about information, in Vincent /oloscoand Janet Wasko, eds., The Political Economy o Information,

  • 8/13/2019 Philip Agre Surveillance And Capture

    21/21

    51 Surveillanceand Capturef.\adison: University of Wisconsin Press, 1988.Douglas Schuler and Aki Namioka, eds., Participatory Design:Prindples and Practices, Hillsdale, NJ: Erlbaum, 1993.r.lichae\5. Scott Morton, ed., The Corporation f the 1990s:Information Technology and Organizational Transformation, NewYork: Oxford U n i v e ~ s i t y Press, 1991.John R Searle, Speech A c ~ . S : An Essay in the Philosophy ofLanguage, Cambridge: Cambridge University Press, 1969.Harley Shaiken, Work Transformed: Automation and Labor in theComputer Age New York: Holt, Rinehart, and Winston, 1985.Rcibert Ellis Smith, Privacy: How to Protect What's Left of It. NewYork: Anchor Press, 1979.Steve Smith, How muCh change at the store?: The impact of newtechnologies and labour process on managers and staffs in retaildistribution, in David Knights and Hugh Willmott, eds., NewTechnology and the Labour Process, london: HacmiUan, 1988.Arthur L Stinchcombe, Information and Organizations, Berkeley:University of California Press, 1990.lucy A Suchman, Office procedure as practical action: Hodels ofwork and system design, ACM Transactions on Office InformationSystems 1 4), 1g83, pages 320-328.lucy Suchman and Brigitte Jordan, Computerization and women sknowledge, in Kea Tijdens, 1\ary Jennings, Ina Wagner, and

    ~ a r g a r e t Weggelaar, eds., Women, Work and Computerization:Forming New Alliances, Amsterdam: North-Holland, 1989.lucy Such man, Technologies of accountability: Of lizards andaeroplanes, in Graham Button, ed., Technology in Working Order:Studies of Work, interaction, and Technology, london: R o u t l e d g e ~1992.Lucy Such man, Do categories have politics?: The language/actionperspective reconsidered, in Georgie de ~ l i c h e t i s , Carla Simone andKjeld Schmidt, eds., Proceedings of the Third European Conferenceon Computer-Supported Cooperative Work: ECSCW 93, Dordrecht. TheNetherlands: Kluwer, 1993.

    theNEWMEDIAREADER

    Frederick Winslow Taylor, The Prinr:ip es of Scientific Management,New York: Harper, 1923.Paul Thompson, The Nature of Work. An Introduction to Debates onthe Labour Process, second edition, london: Macmillan, 1989.5. J. Udoka Automated data capture techniques: prerequisite foreffective integrated manufacturing systems, Computers andIndustrial Engineering 21, 1991, pages 217-222.Charles Von Simpson, Food fight: Business strategies,Computerworld 23 40), 2 October 1989, pages 522-26.Richard E Walton, Up and Running: Integrating InformationTechnology and the Organization, Boston: Harvard Business SchoolPress, 1989.Roy Want, Andy Hopper, Veronica Falcao, and Jonathan Gibbons,The active badge location system, CM Transactions on Information?ystems 10 1), 1992, pages 91-102.Willis H. Ware, The new faces of privacy, The Information Sodety9 3), 1993, pages 195-212./4ark Weiser, Ubiquitous computing, Computer 26 10), 1993, pages71-72.Oliver E Williamson and Sidney G Winter, The Nature of the Firm:Origins, Evolution, and Development. Oxford: Oxford UniversityPress, 1991.Terry Winograd and Fernando Flores Understanding Computers andCognition: New Foundation for Design, Norwood, tlJ: Ab\ex, 1986.JoAnne Yates, Control through Communication: The Rise of System inAmerican Management, Baltimore: Johns Hopkins University Press,1989.