Our extensions also have implications for theories ontrust.

Therefore, training should focus on learning how toquickly recognize volunteers’ volition in participating inan emergent group, the tasks they might engage in, andthe support they might need to carry out those tasks.Such training could also help people to recognize thebenefits and dangers of generalized trust. It could alsohelp people to quickly evolve a coordination mecha-nism that does not rely on what people know, but oncompiling and communicating a narrative of the actionsthat volunteers take, so that others are able to assess forthemselves what actions they could take to help.

ur examination suggests that by expandingthe context in which TMS theory is applied to includeemergent response groups, insights can be gained intotheir internal dynamics. The three indicators of the levelof development of a TMS provide a useful frameworkfor organizing these insights in the exhibit.

The urgency of time may make it too onerous forthe extra effort of articulating actions as they are beingperformed, yet most emergency response requires somecommunication.

Explicitly articulated narratives mayalso make clearer that multiple sequences of actions maybe occurring simultaneously, thus resolving role conflictsby allowing multiple ways to accomplish a task

Emergent response groups may also use a mechanismof creating a community narrative (Boland and Tenkasi1995), which is a running narrative of the actions takenand not taken, the decisions made, and the theories inuse. Narratives do not represent a single shared under-standing of a domain; rather they represent the mul-tiplicity of events and actions a community is taking,as members are taking them. Narratives may be articu-lated explicitly or understood implicitly.

Whenemergent response groups first come together, membersare likely not to ask one another about who knows what;instead, they are likely to ask about what is knownabout the situation and about the actions taken thus far(Dyer and Shafer 2003, Hale et al. 2005). The cogni-tive structure that they develop for the group centersnot around people, but on action-based scenarios thateither have been or might be carried out. These scenariosinclude decisions, actions, knowledge, events, and feed-back (Vera and Crossan 2005).

On the surface, the lack of sta-ble membership suggests that a shared mental modelmay not be viable or even desired in emergent responsegroups. Time may be too precious to seek consensus onevents and actions, and agreements may make the groupless flexible to accommodate to changing inputs.

hus, we believe challenges occur in all three indica-tors of the level of development of a TMS—expertisespecialization, credibility, and expertise coordination—requiring a need to consider extending theorizing abouteach indicator for emergent response groups.

Moreland and Argote(2003) suggest that the dynamic conditions under whichthese groups form and work together are likely to havenegative effects on the development of transactive mem-ory.

Research on TMS has identified three indicators of thelevel of development of a TMS (Lewis 2003, Morelandand Argote 2003):1.Memory (or expertise) specialization:the tendencyfor groups to delegate responsibility and to specialize indifferent aspects of the task;2.Credibility:beliefs about the reliability of mem-bers’ expertise; and3.Task (or expertise) coordination:the ability of teammembers to coordinate their work efficiently based ontheir knowledge of who knows what in the group.The greater the presence of each indicator, the more de-veloped the TMS and the more valuable the TMS is forefficiently coordinating the actions of group members.

A TMS can be thoughtof as a network of interconnected individual memorysystems and the transfer of knowledge among them(Wegner 1995). Individuals who are part of a TMSassume responsibility for different knowledge domains,and rely on one another to access each other’s expertiseacross domains. Expertise is defined in the TMS litera-ture to broadly include the know-what, know-how, andknow-why of a knowledge domain (Quinn et al. 1996),what Blackler (1995) refers to as embodied competen-cies. Expertise specialization, then, reduces the cognitiveload of each individual and the amount of redundantknowledge in the group, while collectively providingthe dyad or group access to a larger pool of knowl-edge. What makes transactive memory transactive arethe communications (called transactions) among individ-uals that make possible the codifying, storing, retrieving,and updating of information from individual memorysystems. For transactive memory to function effectively,individuals must have a shared conceptualization of whoknows what in the group.

TMS theory, a theoryof group-level cognition, explains how people in collec-tives learn, store, use, and coordinate their knowledge toaccomplish individual, group, and organizational goals.It is a theory about how people in relationships, groups,and organizations learn who knows what, and use thatknowledge to decide who will do what, resulting in moreefficient and effective individual and collective perfor-mance.

The urgency of the situation meansthat the objective of coordination is to achieve minimallyacceptable and timely action, even when more effec-tive responses may be feasible—but would take longerand use more resources.

hese characteristics require thatemergent response groups adopt specific approaches forknowledge coordination. One such approach commonlydocumented in studies of such groups is their use ofa learn-by-doing (versus decision making) action-basedmodel of coordinated problem solving, in which sensemaking and improvisation are the norm rather than theexception

isaster researchers havedefinedemergent response groupsas collectives of indi-viduals who use nonroutine resources and activities toapply to nonroutine domains and tasks, using nonroutineorganizational arrangements (Bigley and Roberts 2001,Drabek et al. 1981, Drabek 1986, Drabek and McEntire2003, Kreps 1984, Tierney et al. 2001).

Disasters have wideimplications for expertise coordination because the pre-conditions known to facilitate expertise coordination arelimited or nonexistent in disaster response. Such precon-ditions include but are not limited to, a shared goal; aclear reward structure; known group membership, exper-tise, and skills to accomplish the task; and time to sharewho knows what.

Although the con-ventional indicators of efficient coordination—expertisespecialization, credibility in expertise, and coordinationof expertise—are relevant in disaster response, disasterspresent a unique operational environment. Disasters are“events, observable in time and space, in which societiesor their subunits (e.g., communities, regions) incur phys-ical damages and losses and/or disruption of their routine

Experimentation, the third affordance, refers to theuse of technology to encourage participants to try outnovel ideas.

Recombinability refers to forms of technology-enabled action where individual contributors build oneach others’ contributions.

Reviewability refers to the enactment of technology-enabled new forms of working in which participantsare better able to view and manage the content offront and back narratives over time (West and Lakhani2008). By allowing participants to easily and collab-oratively review a range of ideas, technology-affordedreviewability helps the community respond to tensionsin disembodied ideas, because the reviews can provideimportant contextual information for building on others’ideas.

Technology platforms used by OCs can providea number of affordances for knowledge collabora-tion, three of which we mention here: reviewability,recombinability, and experimentation. These affordancesevolve as new participants provide new ways to use thetechnologies, new social norms are developed around thetechnology affordances, and new needs for fresh affor-dances are identified.

Given the fluid nature of OCsand their rapidly evolving technology platforms, and inline with calls to avoid dualistic thinking about tech-nology (Leonardi and Barley 2008, Markus and Silver2008, Orlikowski and Scott 2008), we suggest technol-ogy affordance as a generative response, one that viewstechnology, action, and roles as emergent, inseparable,and coevolving. Technology affordances offer a relationalperspective on human action, where neither the technol-ogy nor the actor is dominant in the sense that the tech-nology does not define what is possible for the actor todo, nor is the actor free from the limitations of the tech-nological environment. Instead, possibilities for actionemerge from the reciprocal interaction between actor andartifact (Gibson 1979, Zammuto et al. 2007). Thus, anaffordance perspective focuses on the organizing actionsthat are afforded by technology artifacts.

third response to manage tensions is to promoteknowledge collaboration by enacting dynamic bound-aries. In social sciences, although boundaries divide anddisintegrate collectives, they also coordinate and inte-grate social action (Bowker and Star 1999, Lamont andMolnár 2002). Fluidity brings the need for flexible andpermeable boundaries, but it is not only the propertiesof the boundaries but also their dynamicity that helpmanage tensions.

We have observed in OCs that no single narrative isable to keep participants informed about the current stateof the OC with respect to each tension. These commu-nities seem to develop two different types of narratives.Borrowing from Goffman (1959), we label the two nar-ratives the “front” and the “back” narratives.

Based on our collective research on to date, we haveidentified that as tensions ebb and flow, OCs use (or,more precisely, participants engage in) any of the fourtypes of responses that seem to help the OC be gen-erative. The first generative response is labeledEngen-dering Roles in the Moment. In this response, membersenact specific roles that help turn the potentially negativeconsequences of a tension into positive consequences.The second generative response is labeledChannelingParticipation. In this response, members create a nar-rative that helps keep fluid participants informed ofthe state of the knowledge, with this narrative havinga necessary duality between a front narrative for gen-eral public consumption and a back narrative to airthe differences and emotions created by the tensions.The third generative response is labeledDynamicallyChanging Boundaries. In this response, OCs changetheir boundaries in ways that discourage or encouragecertain resources into and out of the communities at cer-tain times, depending on the nature of the tension. Thefourth generative response is labeledEvolving Technol-ogy Affordances. In this response, OCs iteratively evolvetheir technologies in use in ways that are embedded by,and become embedded into, iteratively enhanced socialnorms. These iterations help the OC to socially and tech-nically automate responses to tensions so that the com-munity does not unravel.

Tension 5: Positive and Negative Consequences ofTemporary ConvergenceThe classic models of knowledge collaboration in groupsgive particular weight to the need for convergence. Con-vergence around a single goal, direction, criterion, pro-cess, or solution helps counterbalance the forces ofdivergence, allowing diverse ideas to be framed, ana-lyzed, and coalesced into a single solution (Couger 1996,Isaksen and Treffinger 1985, Osborn 1953, Woodmanet al. 1993). In fluid OCs, convergence is still likelyto exist during knowledge collaboration, but the conver-gence is likely to be temporary and incomplete, oftenimplicit, and is situated among subsets of actors in thecommunity rather than the entire community.

ension 2: Positive and Negative Consequencesof TimeA second tension is between the positive and negativeconsequences of the time that people spend contribut-ing to the OC. Knowledge collaboration requires thatindividuals spend time contributing to the OC’s virtualworkspace (Fleming and Waguespack 2007, Lakhani andvon Hippel 2003, Rafaeli and Ariel 2008). Time has apositive consequence for knowledge collaboration. Themore time people spend evolving others’ contributedideas and responding to others’ comments on thoseideas, the more the ideas can evolve

We argue that it is the fluidity, the tensions that flu-idity creates, and the dynamics in how the OC respondsto these tensions that make knowledge collaboration inOCs fundamentally different from knowledge collabora-tion in teams or other traditional organization structures.

As fluctuations in resource endowments arise overtime because of the fluidity in the OC, these fluctua-tions in resources create fluctuations in tensions, makingsimple structural tactics for managing tensions such ascross-functional teams or divergent opinions (Sheremata2000) inadequate for fostering knowledge collaboration.As complex as these tension fluctuations are for the com-munity, it is precisely these tensions that provide thecatalyst for knowledge collaboration. Communities thatthen respond to these tensions generatively (rather thanin restrictive ways) will be able to realize this potential.Thus, it is not the simple presence of resources that fos-ter knowledge collaboration, but rather the presence ofongoing dynamic tensions within the OC that spur thecollaboration. We describe these tensions in the follow-ing section

Fluidity requires us to look at the dynamics—i.e., thecontinuous and rapid changes in resources—rather thanthe presence or the structural form of the resources.Resources may flow from outside the OC (e.g., pas-sion) or be internally generated (e.g., convergence), sub-sequently influencing and influenced by action (Feldman2004). Resources come with the baggage of having bothpositive and negative consequences for knowledge col-laboration, creating a tension within the community inhow to manage the positive and negative consequencesin a manner similar to the one faced by ambidextrousorganizations (O’Reilly and Tushman 2004).

However, failure to examine the critical roleof even the inactive participants in the functioning of thecommunity is to ignore that passive (and invisible) par-ticipation may be a step toward greater participation, aswhen individuals use passivity as a way to learn aboutthe collective in a form of peripheral legitimate partici-pation (Lave and Wenger 1991, Yeow et al. 2006).

Fluidity recognizes the highly flexible or permeableboundaries of OCs, where it is hard to figure out whois in the community and who is outside (Preece et al.2004) at any point in time, let alone over time. Theyare adaptive in that they change as the attention, actions,and interests of the collective of participants change overtime. Many individuals in an OC are at various stagesof exit and entry that change fluidly over time.

We argue that fluid-ity is a fundamental characteristic of OCs that makesknowledge collaboration in such settings possible. Assimply depicted in Figure 1, we envision OCs as fluidorganizational objects that are simultaneously morphingand yet retaining a recognizable shape (de Laet and Mol2000, Law 2002, Mol and Law 1994).

a growing consensus on factors that moti-vate people to make contributions to these communities,including motivational factors based on self-interest (e.g.,Lakhani and von Hippel 2003, Lerner and Tirole 2002,von Hippel and von Krogh 2003), identity (Bagozzi andDholakia 2006, Blanchard and Markus 2004, Ma andAgarwal 2007, Ren et al. 2007, Stewart and Gosain2006), social capital (Nambisan and Baron 2010; Waskoand Faraj 2000, 2005; Wasko et al. 2009), and socialexchange (Faraj and Johnson 2011).

For instance, knowledge collaboration in OCscan occur without the structural mechanisms tradition-ally associated with knowledge collaboration in orga-nizational teams: stable membership, convergence afterdivergence, repeated people-to-people interactions, goal-sharing, and feelings of interdependence among groupmembers (Boland et al. 1994, Carlile 2002, Dougherty1992, Schrage 1995, Tsoukas 2009).

Knowledge collaboration is defined broadly as thesharing, transfer, accumulation, transformation, andcocreation of knowledge. In an OC, knowledge collab-oration involves individual acts of offering knowledgeto others as well as adding to, recombining, modify-ing, and integrating knowledge that others have con-tributed. Knowledge collaboration is a critical elementof the sustainability of OCs as individuals share andcombine their knowledge in ways that benefit them per-sonally, while contributing to the community’s greaterworth (Blanchard and Markus 2004, Jeppesen andFredericksen 2006, Murray and O’Mahoney 2007, vonHippel and von Krogh 2006, Wasko and Faraj 2000).

Online communities (OCs) are open collectives of dis-persed individuals with members who are not necessarilyknown or identifiable and who share common inter-ests, and these communities attend to both their indi-vidual and their collective welfare (Sproull and Arriaga2007).

The situated and emergent nature of coordinationdoes not imply that practices are completely uniqueand novel. On the one hand, they vary accordingto the logic of the situation and the actors present.On the other hand, as seen in our categorizationof dialogic coordination, they follow a recognizablelogic and are only partially improvised. This tensionbetween familiarity and uniqueness of response is atthe core of a practice view of work (Orlikowski 2002).

Recently, Brown and Duguid (2001, p. 208) sug-gested that coordination of organizational knowledgeis likely to be more challenging than coordination ofroutine work, principally because the “elements to becoordinated are not just individuals but communitiesand the practices they foster.” As we found in ourinvestigation of coordination at the boundary, signif-icant epistemic differences exist and must be recog-nized. As the dialogic practices enacted in responseto problematic trajectories show, the epistemic dif-ferences reflect different perspectives or prioritiesand cannot be bridged through better knowledge

nature point to the limitations of a structuralist viewof coordination. In the same way that an organi-zational routine may unfold differently each timebecause it cannot be fully specified (Feldman andPentland 2003), coordination will vary each time.Independent of embraced rules and programs, therewill always be an element of bricolage reflecting thenecessity of patching together working solutions withthe knowledge and resources at hand (Weick 1993).Actors and the generative schemes that propel theiractions under pressure make up an important com-ponent of coordination’s modus operandi (Bourdieu1990, Emirbayer and Mische 1998).

These practices are highly situated, emer-gent, and contextualized and thus cannot be prespec-ified the way traditional coordination mechanismscan be. Thus, recent efforts based on an information-processing view to develop typologies of coordina-tion mechanisms (e.g., Malone et al. 1999) may be tooformal to allow organizations to mount an effectiveresponse to events characterized by urgency, novelty,surprise, and different interpretations.

Our findings also point to a broader divide in coor-dination research. Much of the power of traditionalcoordination models resides in their information-processing basis and their focus on the design issuessurrounding work unit differentiation and integra-tion. This design-centric view with its emphasis onrules,structures,andmodalitiesofcoordinationislessuseful for studying knowledge work.

Boundarywork requires the ability to see perspectives devel-oped by people immersed in a different commu-nity of knowing (Boland and Tenkasi 1995, Star andGriesemer 1989). Often, particular disciplinary focilead to differences in opinion regarding what stepsto take next in treating the patient.

The termdialogic—as opposed to monologic—recognizes dif-ferences and emphasizes the existence of epistemicboundaries, different understandings of events, andthe existence of boundary objects (e.g., the diagnosisor the treatment plan). A dialogic approach to coordi-nation is the recognition that action, communication,and cognition are essentially relational and highlysituated. We use the concept of trajectory (Bourdieu1990, Strauss 1993) to recognize that treatment pro-gressions are not always linear or positive.

A dialogic coordination practice differs from moregeneral expertise coordination processes in that itis highly situated in the specifics of the unfoldingevent, is urgent and high-staked, and occurs at theboundary between communities of practice. Becausecognition is distributed, responsibility is shared, andepistemic differences are present, interactions can becontentious and conflict laden.

xpertisecoordination refers to processes that manage knowl-edge and skill interdependencies

we describe two categories ofcoordination practices that ensure effective work out-comes. The first category, which we callexpertise coor-dination practices, represents processes that make itpossible to manage knowledge and skill interdepen-dencies. These processes bring about fast response,superior reconfiguration, efficient knowledge shar-ing, and expertise vetting. Second, because of therapidlyunfoldingtempooftreatmentandthestochas-tic nature of the treatment trajectory,dialogic coordina-tion practicesare used as contextually and temporallysituated responses to occasional trajectory deviation,errors, and general threats to the patient. These dia-logic coordination practices are crucial for ensuringeffective coordination but often require contentiousinteractions across communities of practice. Figure 1presents a coordination-focused model of patienttreatment and describes the circumstances underwhich dialogic coordination practices are called for.

We found that coordination in a trauma settingentails two specific practices.

Based on a practice view, we suggest the followingdefinition ofcoordination: a temporally unfolding andcontextualized process of input regulation and inter-action articulation to realize a collective performance.

Theprimarygoalispatientstabilizationandini-tiating atreatment trajectory—a temporally unfolding

rauma centersare representative of organizational entities that arefaced with unpredictable environmental demands,complexsetsoftechnologies,highcoordinationloads,and the paradoxical need to achieve high reliabilitywhile maintaining efficient operations.

We sug-gest that for environments where knowledge work isinterdisciplinary and highly contextualized, the rele-vant lens is one of practice. Practices emerge from anongoing stream of activities and are enacted throughthe contextualized actions of individuals (Orlikowski2000). These practices are driven by a practical logic,thatis,arecognitionofnoveltaskdemands,emergentsituations,andtheunpredictabilityofevolvingaction.Bourdieu (1990, p. 12) definespracticesas generativeformulas reflecting the modus operandi (manner ofworking) in contrast to the opus operatum (finishedwork).

In knowledge work, several related factors sug-gest the need to reconceptualize coordination.

Usinga practice lens (Brown and Duguid 2001, Orlikowski2000), we suggest that in settings where work iscontextualized and nonroutine, traditional models ofcoordination are insufficient to explain coordinationas it occurs in practice. First, because expertise is dis-tributed and work highly contextualized, expertisecoordination is required to manage knowledge andskill interdependencies. Second, to avoid error andto ensure that the patient remains on a recoveringtrajectory, fast-response cross-boundary coordinationpractices are enacted. Because of the epistemic dis-tance between specialists organized in communitiesofpractice,theselattercoordinationpracticesmagnifyknowledge differences and are partly contentious.

In this paper, we focus on the collective perfor-manceaspectofcoordinationandemphasizethetem-poral unfolding and situated nature of coordinativeaction. We address how knowledge work is coor-dinated in organizations where decisions must bemade rapidly and where errors can be fatal.

Our empirical example also highlighted the value of achieving virtual temporal symmetry for members of a geographically dispersed community. As electronic me­dia become increasingly central to organizational life, in­dividuals may use asynchronous media in various ways to shape devices of virtual symmetry that help them co­ordinate across geographical distance and across multiple temporal structures. This suggests that when studying the use of electronic media, researchers should pay attention to the conditions in which virtual temporal symmetry may be enacted to coordinate distributed activities, and with what consequences. Interesting questions for empirical research include the following. As work groups in orga­nizations become more geographically dispersed and/or more dependent on electronic media, do members enact virtual temporal symmetry for certain purposes? If so, for which types of purposes? And how? If not, how do such work groups achieve temporal coordination?

Leaming viewed as situated activity has as its central defining characteristic a process that we call legitimate peripheral par­ticipation. By this we mean to draw attention to the point that learners inevitably participate in communities of practitioners and that the mastery of knowledge and skill requires newcom­ers to move toward full participation in the sociocultural practices of a community.

"Legitimate peripheral participation" provides a way to speak about the relations between newcom­ers and old-timers, and about activities, identities, artifacts, and communities of knowledge and practice. It concerns the process by which newcomers become part of a community of practice. A person's intentions to learn are engaged and the meaning of learning is configured through the process of be­coming a full participant in a sociocultural practice. This so­cial process includes, indeed it subsumes, the learning of knowledgeable skills.

Thus, an enacted environment has both a public and a private face. Publicly, it is a construction that is usually visible to observers other than the actor. Privately, it is a map of if-then assertions in which actions are related to out- comes. These assertions serve as expectations about what will happen in the future.

At the heart of enactment is the idea that cognition lies in the path of the action. Action precedes cognition and focuses cognition. The sensemaking sequence implied in the phrase, ‘How can I know what I think until I see what I say?’ involves the action of talking, which lays down traces that are examined, so that cognitions can be inferred. These inferred cognitions then become pre- conceptions which partially affect the next episode of talk, which means the next set of traces deposited by talk are affected partially by previous labels and partially by current context. These earlier inferences also affect how the next episode of talk is examined and what is seen.

Another way to use a classification system is to consider if there are other possible values that could be used for a given dimension.

For researchers looking for new avenues within human computation, a starting point would be to pick two dimensions and list all possible combinations of values.

These properties formed three of our dimensions: motivation, human skill, and aggregation.

A subtle distinction among human computation systems is the order in which these three roles are performed. We consider the computer to be active only when it is playing an active role in solving the problem, as opposed to simply aggregating results or acting as an information channel. Many permutations are possible.

The classification system we are presenting is based on six of the most salient distinguishing factors. These are summarized in Figure 3.

"... groups of individuals doing things collectively that seem intelligent.” [41]

Data mining can be defined broadly as: “the application of specific algorithms for extracting patterns from data.” [17]

“... applications and services that facilitate collective action and social interaction online with rich exchange of multimedia information and evolution of aggregate knowledge...” [48]

The intersection of crowdsourcing with human computation in Figure 1 represents applications that could reasonably be considered as replacements for either traditional human roles or computer roles.

“Crowdsourcing is the act of taking a job traditionally performed by a designated agent (usually an employee) and outsourcing it to an undefined, generally large group of people in the form of an open call.” [24

modern usage was inspired by von Ahn’s 2005 dissertation titled "Human Computation" [64] and the work leading to it. That thesis defines the term as: “...a paradigm for utilizing human processing power to solve problems that computers cannot yet solve.”

When classifying an artifact, we consider not what it aspires to be, but what it is in its present state.

human computation does not encompass online discussions or creative projects where the initiative and flow of activity are directed primarily by the participants’ inspiration, as opposed to a predetermined plan designed to solve a computational problem.

Human computation is related to, but not synonymous with terms such as collective intelligence, crowdsourcing, and social computing, though all are important to understanding the landscape in which human computation is situated.

Because of both the content that people upload and the behavioral traces that they leavebehind, social network sites have unprecedented quantities of data concerning humaninteraction. This presents unique opportunities and challenges. On one hand, SNSs offera vibrant “living lab” and access to behavioral data at a scale inconceivable to manysocial scientists. On the other, the data that are available present serious research ethicsquestions and introduce new types of biases that must be examined (boyd and Crawford2012)

For those of us who believe that social network sites are socio-technical systems, in whichsocial and technical factors shape one another, failing to describe the site under studyignores the fact that the technological constraints and affordances of a site will shapeuser practices and that social norms will emerge over time. Not including informationabout what the feature set was at the time of data collection forecloses the possibility ofidentifying patterns that emerge over time and through the accumulated scholarshipacross a range of sites and user samples. Unfortunately, because they have no knowledgeabout how things will continue to evolve and which features will becomeimportant to track, researchers may not be able to identify the salient features to reportand may struggle with devoting scarce publication space to these details, but this doesn’tundermine the importance of conscientious consideration towards describing the artifactbeing analyzed.

In order to produce scholarship that will be enduring, the onus is on social mediaresearchers to describe the technological artifact that they are analyzing with as muchcare as survey researchers take in describing the population sampled, and with as muchdetail as ethnographers use when describing their field site. This is not to say thatresearchers must continue to describe technologies as if no one knows what they are—weare beyond the point where researchers must explain how electronic mail or “email” islike or unlike postal mail. But, rather, researchers must clearly describe the socio-technical context of the particular site, service, or application their scholarship isaddressing. In addition to attending to the technology itself, and the interchange betweentechnical and social processes, we believe SNS researchers should make a concertedeffort to include the date of data collection and to describe the site at the moment of datacollection and the relevant practices of its users. These descriptions will enable laterresearchers to synthesize across studies to identify patterns, much in the same wayreporting exact effect sizes allows for future meta-analyses

One key challenge of studying social media is that designers of these tools are innovatingat a very rapid timeframe and often with little advance notice. Given the rapidly changinginfrastructure and the timeframe of academic publishing, the site at the time of datacollection is likely to be very different from its incarnation at the point of publication.

Because of howpeople's position within the SNS shapes their experiences of it, activity-centric analysesrequire contextualization and translation, not unlike what social scientists studyingdiffering cultural practices have had to do for decades.

What oneexperiences on SNSs and the content to which one is exposed differs depending on thestructure of one's network, a user's individual preferences and history, and her activitiesat that moment.

By far the most pressing challengefor SNS scholars lies in the rapid pace at which innovations and technical changes areimplemented in this space. For scholarship in this arena to develop, SNS researchersneed to be mindful of the ways in which these sites evolve over time and the effects thismay have on the interpersonal, psychological, and sociological processes they arestudying.

What makes “social media” significant as a category is not the technology, butrather the socio-technical dynamics that unfolded as millions of people embraced thetechnology and used it to collaborate, share information, and socialize. Popular genres ofsocial media integrated the public nature of interest-driven CMC with the more intimatedynamics of interpersonal CMC.

All SNSs support multiple modes of communication: one-to-many and one-to-one,synchronous and asynchronous, textual and media-based

Many of the weak tie relationships articulatedon SNSs would fade away were it not for the ease with which people can communicate,share, and maintain simple connections. For this reason, this new definition positionssocial network sites first and foremost as a communication platform, while alsohighlighting the importance of sharing content, typically consumed through a stream.

A social network site is anetworked communication platformin which participants1) haveuniquely identifiable profilesthat consist of user-supplied content, contentprovided by other users, and/or system-level data; 2) canpublicly articulateconnectionsthat can be viewed and traversed by others; and 3) can consume,produce, and/or interact withstreams of user-generated contentprovided by theirconnections on the site.

As social network sites have become mainstream, traversing the connections betweenpeople to view profiles is no longer the sole—or, even primary—way of participation.Content is surfaced through streams, and each piece of content is embedded withnumerous links to other content nuggets.

Yet, one significant shift has unfolded: the traversability ofconnections has become more important for machines than users. As APIs make thesocial graph available to broader audiences, algorithms are being designed to traversethe graph and learn about the individual nodes’ relationship to one another.

The ability to see—andtraverse—others’ contact lists was innovative and important in several ways. From anadoption perspective, it enabled users to find shared contacts easily, thus lowering thebarriers to initiating contact with other users and enabling users to harness networkeffects more easily. From a social perspective, it allowed people to easily see therelationships between others, to reconnect with old friends and acquaintances, and totravel through the network in a way that enhanced social interactions.

The rise of open APIs and developer platforms meant that these collections of articulatedcontacts became valuable in contexts outside that particular SNS. Engineers andentrepreneurs alike began talking about the “social graph”—the global network oflinkages between all individuals within a system (Fitzpatrick and Recordon2007). Thislanguage emerged at a time when commercial entities began to believe that the socialgraph hadvalue beyond the individual's relationship with a given social networksite.

As SNSs became more popular with a wider range ofindividuals, many individuals’ contact lists became more diverse as these users Friendedpeople representing a range of contexts (family, professional contacts, church members,etc.). This growing diversity has contributed to cases of “context collapse,” whichdescribes the ways in which individuals that we know from different social contexts cometogether in SNSs in potentially uncomfortable ways (Marwick and boyd2011)

For users, these connections represent what sociologistsrefer to as a person'ssocial network—the collection of social relations of varyingstrengths and importance that a person maintains

Earlier communication tools enabled individuals to create a private list ofcontacts (for instance a buddy list on instant messaging), to establish a group of contactsthat were shared by others (such as a listserv membership list), or to publish a list ofrelated links (such as a blogroll), but SNSs extended the practice of creating a publiclyvisible, personally curated list of contacts and made it a mainstream practice.

Streams of quotidian,ephemeral content encourage people to participate more in that they provide an initialartifact around which others can engage. Features that support actions associated withstatus updates—the ability to post comments to, share, or register interest in an update—also encourage a stream of activity that is prompted by an update but often takes on a lifeof its own in the central stream. Today's SNSs are more like news aggregators than theyare like profile-based contexts, even if the algorithm for displaying content is quiteobfuscated.

In boyd and Ellison (2007), we attempted to stabilize the discussion by offeringa definition of social network sites:web-based services that allow individuals to (1) construct a public or semi-publicprofile within a bounded system, (2) articulate a list of other users with whom theyshare a connection, and (3) view and traverse their list of connections and thosemade by others within the system.

hus it becomes possible to see how ques-tions around data use need to shift from asking what is in the data, to include discussions of how the data is structured, and how this structure codifies value systems and social practices, subject positions and forms of visibility and invisi-bility (and thus forms of surveillance), along with the very ideas of crisis, risk governance and preparedness. Practices around big data produce and perpetuate specific forms of social engagement as well as understandings of the areas affected and the people being served.

Big data is not just about knowing more. It could be – and should be – about knowing better or about changing what knowing means. It is an ethico- episteme-ontological- political matter. The ‘needle in the haystack’ metaphor conceals the fact that there is no such thing as one reality that can be revealed. But multiple, lived are made through mediations and human and technological assemblages. Refugees’ realities of intersecting intelligences are shaped by the ethico- episteme-ontological politics of big data.

Burns (2015) builds on this to investigate how within digital humanitarianism discourses, big data produce and perform subjects ‘in need’ (individuals or com-munities affected by crises) and a humanitarian ‘saviour’ community that, in turn, seeks answers through big data

Crises are often not a crisis of information. It is often not a lack of data or capacity to analyse it that prevents ‘us’ from pre-venting disasters or responding effectively. Risk management fails because there is a lack of a relational sense of responsibility. But this does not have to be the case. Technologies that are designed to support collaboration, such as what Jasanoff (2007) terms ‘technologies of humility’, can be better explored to find ways of framing data and correlations that elicit a greater sense of relational responsibility and commitment.

In some ways this constitutes the production of ‘liquid resilience’ – a deflection of risk to the individuals and communities affected which moves us from the idea of an all-powerful and knowing state to that of a ‘plethora of partial projects and initiatives that are seeking to harness ICTs in the service of better knowing and governing individuals and populations’ (Ruppert 2012: 118)

Indeed, as Chandler (2015: 9) also argues, crowdsourcing of big data does not equate to a democratisation of risk assessment or risk governance:

ut Burns finds that humanitarian staff often describe the local communities and ‘crowds’ as the ‘eyes, ears and sensors’ of UN staff, which does not index a genuine collaborative relationship. He states: ‘In all these cases, the discourse talks of putting local people “in the driving seat” when in reality the direction of the journey has already been decided’ (Burns 2015: 48). Burns (2015: 42) also notes that this leads to a transformation of social responsibility into individual responsibility.Neoliberalism’s promotion of free market norms is therefore much more than the simple ideology of free market economics. It is a specific form of social rule that institutionalises a rationality of competition, enterprise indi-vidualised responsibility. Although the state ‘steps back’ and encourages the free conduct of individuals, this is achieved through active intervention into civil society and the opening up of new areas to the logic of private enter-prise and individual initiative. This is the logic behind the rise of resilience

The UNHCR has even called for the refugees themselves to also develop their own data solutions and ideas (see Palmer 2014) as a way to help build their ideologies into the data infrastructures and thus bring their prisms into view. This could create a richer situational awareness and a better ability to understand and deal with unfolding and future crises by supporting resilient communities through giving them the means of data producing and sharing

Doing so switches the discourse from vulnerability, where there is a need for external protection mobilised from above to come in and rescue the refugees, to one of resilience, where self- sufficiency and autonomy are part of the equation (Meier 2013).

Yet, at the same time as power is exercised by both the state and corporations, power is gathering from the bottom up in new ways. In disaster response, a dynamic interplay between publics and experts is captured by the concept of social collective intelligence (Büscher et al. 2014); a disruptive innovative force that is challenging the social, economic, political and organisational practices that shape disaster response.

Since the data is already being collected on a regular basis by ubiquitous private firms, it is thought to contain information that will increase opportunities for intelligence gathering and thereby security. This marks a shift from surveillance to ‘dataveillance’ (van Dijck 2014), where the impetus for data processing is no longer motivated by specific purposes or suspicions, but opportunistic discovery of anomalies that can be investigated. For crisis management this could mean benefits such as richer situation awareness, increased capacity for risk assess-ment, anticipation and prediction, as well as more agile response

lthough Snowden’s revelations shocked the world and prompted calls for a public debate on issues of privacy and transparency

While seemingly avoiding the traps of exerting top- down power over people the state does not yet have formal control over, and simultaneously providing support for self- determination and choice to empower individuals for self- sufficiency rather than defining them as vulnerable and passive recipients of top- down protection (Meier 2013), tying individual aid to mobile tracking puts refugees in a situation where their security is dependent upon individual choice and the private sector. Apart from disrupting traditional dynamics of responsibility for aid and protection, public–private sharing of intel-ligence brings new forms of dataveillance

However, with these big data collections, the focus becomes not the individu-al’s behaviour but social and economic insecurities, vulnerabilities and resilience in relation to the movement of such people. The shift acknowledges that what is surveilled is more complex than an individual person’s movements, communica-tions and actions over time.

Andrejevic and Gates (2014: 190) suggest that ‘the target becomes the hidden patterns in the data, rather than particular individuals or events’. National and local authorities are not seeking to monitor individuals and discipline their behaviour but to see how many people will reach the country and when, so that they can accommodate them, secure borders, and identify long- term social out-looks such as education, civil services, and impacts upon the host community (Pham et al. 2015).

as boyd and Crawford argue, ‘without taking into account the sample of a data set, the size of the data set is meaningless’ (2012: 669). Furthermore, many tech-niques used by the state and corporations in big data analysis are based on probabilistic prediction which, some experts argue, is alien to, and even incom-prehensible for, human reasoning (Heaven 2013). As Mayer-Schönberger stresses, we should be ‘less worried about privacy and more worried about the abuse of probabilistic prediction’ as these processes confront us with ‘profound ethical dilemmas’ (in Heaven 2013: 35).

Second, this tracking and tracing of refugees has become a deeply ambiguous process in a world riven by political conflict, where ‘migration’ increasingly comes to be discussed in co- location with terrorism.

Surveillance studies have tracked a shift from discipline to control (Deleuze 1992; Haggerty and Ericson 2000; Lyon 2014) exemplified by the shift from monitoring confined populations (through technologies such as the panopticon) to using new technologies to keep track of mobile populations.

As Coyle and Meier (2009) argue, disasters are often seen as crises of information where it is vital to make sure that people know where to find potable water, how to ask for help, where their relatives are, or if their home is at risk; as well as providing emergency response and human-itarian agencies with information about affected populations. Such a quest for information for ‘security’, in turn, provides fertile ground for a quest for technological solutions, such as big data, which open up opportunities for the extended surveillance of everyday life. The assumption is that if only enough information could be gathered and exchanged, preparedness, resilience and control would follow. This is particularly pertinent with regard to mobile pop-ulations (Adey and Kirby 2016)

hird, at this juncture, control is being equated with visibility and visibility with personal security. But how these individuals are made visible matters for both privacy and security, let alone the politics of conflating refugees, migration and terrorism. Indeed, working with specific data framing mechanisms affects how the causes and effects of disasters are identified and what elements and people are considered (Frickel 2008

First, there is a double dynamic to the generation of data in the refugee crisis.

Datafication refers to the fact that ‘we can now capture and calculate at a much more comprehensive scale the physical and intangible aspects of existence and act on them’ (Mayer- Schönberger and Cukier 2013: 97

There is an uneasy coming together of diverse computational and human intelligences in these intersections, and the ambiguous nature of intelligence – understood, on the one hand, as a capacity for perceiving, learning and under-standing and, on the other, as information obtained for strategic purposes – marks complex relationships between ‘good’ and ‘dark’ aspects of big data, surveil-lance and crisis management.

The goal of this framework is to envision afuture of crowd work that cansupport more complex, creative, and highly valued work. At the highest level, a platformis needed for managing pools of tasksandworkers. Complex tasks must be decomposed into smaller subtasks, each designed with particular needs and characteristics which must be assignedto appropriate groups of workerswho themselves must be properly motivated, selected(e.g., through reputation), and organized (e.g., through hierarchy). Tasks may be structured through multi-stage workflowsin which workers may collaborateeither synchronously or asynchronously. As part of this, AImay guide (and be guided by) crowd workers. Finally, quality assuranceis needed to ensure each worker’s output is of high quality and fits together.

n human computation, people act as computational components and perform the work that AI systemslack the skillstocomplete

A promising approach that addresses some worker output issues examines the way that workers do their work rather than the output itself, using machine learning and/or visualization to predict the quality of a worker’s output from their behavior [119,120]

Of the research foci, quality control has arguably received the most attention so far. Approaches for quality control largely fall into two camps: up-front task design and post-hoc result analysis. Task design aimsto design tasks that are resistant to low-quality work.

Many tasks worth completing require cooperation –yet crowdsourcing has largely focused on independent work. Distributed teams have always facedchallenges in cultural differences and coordination[60], but crowd collaboration now must createrapport over much shorter timescales(e.g., one hour) and possibly wider culturalor socioeconomic gaps

Finally, it will be amajor research undertaking to invent and describe the tasks and techniques that succeed with synchronous collaboration

The two core challenges for realtime crowdsourcing will be 1) scaling upto increased demand for realtime workers, and 2) making workers efficient enough to collectively generateresultsahead of time deadlines.

Volunteer crowdsourcing platforms have evolved their own hierarchies and decision-making processes [104,156], appropriating techniques from other online communities where appropriate [101]. Most paid approaches have workers make hierarchical decisions collectively: for example, task decomposition and integration[75,80],quality oversight of each others’ contributions[78,100], and leader elections to represent collective opinions[83].

Complex tasks have dependencies, changing requirements, and require multiple types of expertise.

n the sections below, we survey and analyze the 12research focithat comprise our model. First, we consider the future of the work processesand how the workis organized and accomplished. Second, we consider the integration of crowd work and computation, including the symbiosis between human cognition, artificial intelligence(AI), and computationally-enabled crowd platforms. Finally, we consider crowd workers and how we can develop jobs, reputation systems, motivations, and incentives that will benefit them.

Unlike traditional organizations in which workers possess job security and managers can closely supervise and appropriately reward or sanction workers, or distributed computing systems in which processors are usually highly reliable, crowd work poses uniquechallenges for both workers and requestersranging fromjob satisfactiontodirection-setting, coordination, and quality control.

These same requirements exist in distributed computing, in which tasks need to be scheduled so that they can be completed in the correct sequence and in a timely manner, with data being transferred between computing elements appropriately.

However, crowd work can bea double-edged sword,equally capable of enhancing or diminishing the quality of workers’ lives.We maysee echoes of past labor abuses in globally distributed crowd work: extremely low pay for labor,with marketplaces such as Amazon’s Mechanical

We focus this paper on paid,onlinecrowd work, which we define here as the performance of tasks online by distributed crowd workers who are financially compensated by requesters(individuals, groups, or organizations). In this sense, crowd work is a socio-technical work systemconstituted through a set of relationships that connect organizations, individuals, technologies and work activities

A variety of terminology is currently used in regard to crowds, e.g. crowdsourcing, collective intelligence, human computation, serious games,peer production, and citizen science

A related series of studies have sought to unpack the dynamics of collabo-ration and to understand which features of peer productions support the cre-ation of higher quality content. This topic has been studied especially closelyin the case of Wikipedia, where particular organizational attributes, routines,norms, and technical features impact the quality of individual contributionsas well as the final, collaborative product.

A more fruitfulapproach considers variation in peer production success to understand whenand where it works better and worse.

Just as peer produced goods vary in their nature and form, there is alsoenormous variation between and within projects in terms of the dimensionsalong which quality might be evaluated. In the case of Wikipedia, scholarshave assessed the encyclopedia in terms of factual accuracy, scope of coverage,political bias, expert evaluation, and peer evaluation – often drawing differentconclusions about the quality of Wikipedia or particular articles.

Although this hasconstituted an inconvenient fact in peer production practice, it also reflects animportant opportunity for future research. By focusing only on the projectsthat successfully mobilize contributors, researchers interested inwhenpeerproduction occurs or the reasonswhyit succeeds at producing high qualityoutputs have systematically selected on their dependent variables. An impor-tant direction for peer production research will be to study these failures.

Infollowing these three paths, scholars have begun to consider variation withinpeer production projects to understand when and why peer production leadsto different kinds of high quality outputs

Both for-profit andnon-profit organizations that have incorporated peer production models havethrived in the networked environment, often overcoming competition frommore traditional, market- and firm-based models.

Recent work hasbegun to probe more deeply into different dimensions along which qualitycan be conceptualized and measured. This new scholarship has given rise toa more nuanced understanding of the different mechanisms through whichhigh quality resources arise, and founder, in peer production.

Peer production successfully elicits contributions from diverse individu-als with diverse motivations – a quality that continues to distinguish it fromsimilar forms of collective intelligence

Resolving the tensions between different motivations and incentives presentsa design challenge for peer production systems and other collective intelli-gence platforms. The complex interdependence of motivations, incentive sys-tems, and the social behaviors that distinct system designs elicit has led Krautand Resnick (2012) to call for evidence-based social design and Benkler (2009,2011) for cooperative human system design.

Evidence from this newer body of research shows that motivations are di-versewithincontributors and that different contributors have different mixesof motivations.

The most important insight provided by some of this newer workis that contributors act for different reasons, and that theories based on a sin-gle uniform motivational model are likely to mischaracterize the motivationaldynamics.

In particular, these neweraccounts have focused on social status, peer effects, prosocial altruism, groupidentification, and related social psychological dimensions of group behavior

hat said, a growing number of stud-ies also suggest that these motives interact with each other in unpredictableways and, as a result, are vulnerable to “crowding out” when the introduc-tion of extrinsic incentives undermines intrinsic motivation

Other foundational research on motivation in peer production by Lernerand Schankerman (2010) and others has explored why organizations, firmsand governments, rather than individual users, choose to participate in opensource software.

Despite their differences in emphasis,scope, and genre, all of these surveys support the claim that motivations inpeer production are diverse and heterogeneous.

Frequently cited motivations in foundational work by theseauthors, von Hippel and von Krogh (2003), von Krogh (2003), and others in-cluded: the use value of the software to the contributing developer; the hedo-nic pleasure of building software; the increased human capital, reputation,or employment prospects; and social status within a community of peers.Other early accounts analyzing examples of peer production beyond FLOSSsuggested additional motivations. For example, Kollock (1999) emphasizedreciprocity, reputation, a sense of efficacy, and collective identity as salientsocial psychological drivers of contribution to online communities and fo-rums.

A second quality of peer production that challenged conventional economictheories of motivation and cooperation was the absence of clear extrinsic in-centives like monetary rewards. Traditional economic explanations of behav-ior rely on the assumption of a fundamentally self-interested actor mobilizedthrough financial or other incentives. In seeking to explain how peer produc-tion projects attract highly skilled contributors without money, much of theliterature on peer production has focused on questions of participant moti-vation.

A newer wave of work has stepped back from this approachand sought to explain how multiple motivational “vectors” figure in the cre-ation of common pool resources online – an approach that underscores a coreadvantage of peer production in its capacity to enable action without requir-ing translation into a system of formalized, extrinsic, carrots and sticks.

Future workcan also begin to address questions of whether, how, and why peer produc-tion systems have transformed some existing organizational fields more pro-foundly than others. An empirically-informed understanding of when andwhere organizational practices drawn from peer production provide efficientand equitable means to produce, disseminate, and access information can pro-vide social impact beyond the insights available through the study of any in-dividual community

By structuring design changes as exper-iments, these studies make credible causal claims about the relationship oforganizational structure and project outcomes that previous work struggledto establish. By intervening in real communities, these efforts achieve a levelof external validity that lab-based experiments cannot

Similar cross-organization studies in other areas of peer production, orstudies comparing across differenttypesof peer production, have remainedchallenging and rare. One difficulty with comparative work across organiza-tions, in general, is designing research capable of supporting inference intothe causes of organizational success and failure

Although some of the earliest theories of the organization of peer pro-duction celebrated the phenomena as non-hierarchical, more recent work hasquestioned both the putative lack of hierarchy and its purported benefits (e.g.,Kreiss et al., 2011).

More recent work on organizational aspects of peer production has begunto question the “stylized facts” that prevailed in earlier research.

This research built on earlier work showing that contrib-utors to bulletin boards, newsgroups, forums, and related systems adopteddurable “social roles” through their patterns of contribution (e.g., Fisher et al.,2006) – an approach that was subsequently applied to Wikipedia by Welseret al. (2011) and McDonald et al. (2011)

A large body of descriptive work has also sought to characterize organi-zational dimensions of Wikipedia and the practices of its contributors.

Much of the earliest empirical research inductively sought to describe thestructure and organization of FLOSS communities

Other foundational accounts, like Moglen (1999) and Weber (2004),attended to the emergence of informal hierarchies and governance arrange-ments within communities.

Peer productioncould, Benkler argued, outperform traditional organizational forms underconditions of widespread access to networked communications technologies,a multitude of motivations driving contributions, and non-rival informationcapable of being broken down into granular, modular, and easy-to-integrate

Initial scholarship on the organization of peer production emphasized the-oretical distinctions between peer production, bureaucracy, and transactioncost explanations of firms and markets.

Indeed, peer production com-munities perform all of the “classical” organizational functions like coordina-tion, division of labor, recruitment, training, norm creation and enforcement,conflict resolution, and boundary maintenance – but do so in the absence ofmany of the institutions associated with more traditional organizations.

In the rest of this chapter, we describe the development of the academicliterature on peer production and collective intelligence in three areas – or-ganization, motivation, and quality. In each area, we introducefoundationalworkconsisting primarily of earlier scholarship that sought to describe peerproduction and establish its legitimacy. Subsequently, we characterize work,usually more recent, that seeks to pursuenew directionsand to derive morenuanced analytical insights. Because FLOSS and Wikipedia have generatedthe majority of the peer production research to date, we focus on researchanalyzing those efforts. For each theoretical area, we briefly synthesize bothfoundational work and new directions and describe some of the challenges forfuture scholarship. Neither our themes nor our periodization are intended toencompass the complete literature on peer production. Instead, they reflectcore areas of research that speak most directly to the literature on collectiveintelligence and locate peer production within that broader phenomenon. Weconclude with a discussion of several issues that traverse our themes and impli-cations of peer production scholarship for research on collective intelligencemore broadly.

Although peer production is central to social scientific and legal researchon collective intelligence, not all examples of collective intelligence created inonline systems are peer production. First, (1) collective intelligence can in-volve centralized control over goal-setting and execution of tasks.

Peer production is the most significant organizational in-novation that has emerged from Internet-mediated social practice, among themost visible and important examples of collective intelligence, and a centraltheoretical frame used by social scientists and legal scholars of collective intel-ligence.

Following Benkler (2013), we define peer produc-tion as a form of open creation and sharing performed by groups online that:set and execute goals in a decentralized manner; harness a diverse range ofparticipant motivations, particularly non-monetary motivations; and sepa-rate governance and management relations from exclusive forms of propertyand relational contracts (i.e., projects are governed as open commons or com-mon property regimes and organizational governance utilizes combinationsof participatory, meritocratic and charismatic, rather than proprietary or con-

Consistent with this exam-ple, foundational social scientific research relevant to understanding collec-tive intelligence has focused on three central concerns: (1) explaining the or-ganization and governance of decentralized projects, (2) understanding themotivation of contributors in the absence of financial incentives or coerciveobligations, and (3) evaluating the quality of the products generated throughcollective intelligence systems.

Historically,researchers in diverse fields such as communication, sociology, law, and eco-nomics have argued that effective human systems organize people through acombination of hierarchical structures (e.g., bureaucracies), completely dis-tributed coordination mechanisms (e.g., markets), and social institutions ofvarious kinds (e.g., cultural norms). However, the rise of networked systemsand online platforms for collective intelligence has upended many of the as-sumptions and findings from this earlier research.

mporal features. We have then to consider how organizational participants are affected by situations containing temporal features, but also how these actors shape, by their behavior and beliefs, local context according to their needs.

s a means for coordinating action among groups of users (e.g., Bardram, 2005, this volume)

The hierarchical structure of the Boer et al. adaptation of the Activity Theory model can help to reconcile the differences in granularity and the difficulties of supporting collaboration identified in our work; future activity-centered user interfaces might take advantage of the zoomable user interface paradigm or feature control over the level of detail (LOD) represented in the interface to more accurately reflect the depth at which a given user conceptualizes their own tasks or the tasks of their colleagues.

Activities also span place; that is, it is common for work to take place outside of the immediate office environment. However, current office technologies sometimes present a very different view of information across different physical and virtual settings.

The idea that activities may exist at different levels of granularity is not a new one. Boer, van Baalen & Kumar (2002) provide a model explaining how an activity at one level of analysis may be modeled as an action—a component of an activity—at another. This holds true for individual users, as in the example provided above, but is even more pronounced when a single activity is viewed from multiple participants’ perspectives.

Additionally, activities need to be represented in such a way that their contents can be shared, with the caveats that individual participants in an activity may have very different perceptions of the activity, they may bring different resources to play over the course of the activity, and, particularly for large activities in which many individual users participate, users themselves may come and go over the life of the activity.

Recognizing the mediating role of the digital work environment in enabling users to meaningfully collaborate is a critical step to ensuring the success of these systems.

At present, information and communication technologies con-tinue to reshape temporal experience and collective time consciousness (Nowotny, 1989b)

Despite the apparent diversity of themes, certain common patterns can be discerned in empirical studies dealing with time. They bear the imprint of the ups and downs of research fashions as well as the waxing and waning of influences from neighbouring disciplines. But they all acknowledge 'time as a problem' in 'time-compact' societies (Lenntorp, 1978), imbued with the pressures of time that come from time being a scarce resource.

Another strategy in dealing with sui generis time consists in juxtaposing clock time to the various forms of 'social time' and considers the latter as the more 'natural' ones, i.e. closer to subjective perceptions of time, or to the temporality that results from adaptations to seasons or other kinds of natural (biological, environmental) rhythm. This strategy, often couched also in terms of an opposition between 'linear' clock time and 'cyclical' time of natural and social rhythms devalues, or at least ques-tions, the temporality of formal organizations which rely heavily on clock time in fulfilling their coordinative and integrative and controlling functions (Young, 1988; Elchardus, 1988).

Sorokin and Merton (1937) may be said to have provided the 'definitive' classic statement on the distinction between social and natural time. They associate the physical time of diurnal and seasonal cycles with clock time and define this time as 'purely quantitative, shorn of qualitative variation' (p. 621). 'All time systems', Sorokin and Merton suggest further, 'may be reduced to the need of providing means for synchronising and co-ordinating the activities and observations of the constituents of groups' (p. 627).

Brown and Eisenhardt studied change and project management in com­puter firms and found that firms with less successful project portfolios demon­strated very low amounts of communication across projects. This was part of the context in which projects were planned, divided into small tasks, and then executed in a “structured sequence of steps” (1997, p. 14). A structured sequence is, of course, a monochronic strategy, and the low amount of communication is consistent with the proposition that monochronic strategies generate less awareness of other activities and tasks. One of the managers in their study re­marked, “Most people only look at their part” (p. 14); another, “The work of everyone else doesn’t really affect my work” (p. 14). These responses contrasted with the pattern of work in the companies that managed their portfolios of projects more successfully, which Brown and Eisenhardt characterized as “it­erative” (p. 14). Iterative (repetitive) patterns are suggestive of the back and forth flow of polychronic strategies

in a more polychronic culture, people would stand closer to each other while talking. So time and space are related in the social as well as the physical world.

Leshed and Sengers’s research reminds us that calendars are not just tools for the management of time, but are also sites of identity work where people can project to themselves and others the density of their days and apparent ‘success’ at doing it all[26]. These seemingly innocuous artifacts can thus perpetuate deeper normative logics a

. When creating tools for schedulingandcoordination, it is crucial to provide ways for people to take into account not just the multiplicity of (potentially dissonant) rhythms [22, 46], but also the differential affective experiencesof time rendered by such rhythms.

This point ties into the conceptualisation of time as collec-tive [29] and entangled [43]. The infrastructure that sup-ports a 24/7 society is one that relies on people as well as technologies, the conventional nine-to-five work rhythm, for example, being underpinned by people working shifts outside of these hours.

Notably, a practice-oriented treatment of digital time does open up avenues for research and design, one that resonates with Kuutti and Bannon’s [23] recent account of a practice perspective forming a new paradigm for HCI. They propose that a central issue in a practice-based research agenda is the need to develop the capability to transform practices through technology. Essential to this is understanding the role of computer artefacts in the emergence and transfor-mation of practices, and the possibilities for influencing these by changing the artefacts themselves.

How can we design for time as collective and interdependent, rather than individualised on the one hand, or explicitly scheduled on the other? What does it mean to position collective time not as something that is achieved when people come together, but as a set of relationships through which they are connected? Both Sharma and Mazmanian and Erickson raise this challenge while highlighting the difficulty in addressing it; neither offer a solution.

Designing for an alternative temporal experience means understanding the ways in which multiple temporali-ties intersect, whether these frame a person’s working day, or allow a family to spend time together. While scheduling technologies do of course have a role to play here [see e.g. 31], many of the temporal structures that frame everyday life are not so much scheduled as unfold in a way that isunremarkable [54], or are so firmly established that they are no longer seen as alterable.

In his analysis of the concept, Southerton [45] identifies quality time as a contemporary concern, and uses it alongside an analysis of diaries written in 1937 and 2000 to examine the impression that everyday life is speeding up. His findings lead him to argue that the feeling of time pressure that seems inherent to modern life is due to difficulties in coor-dinating practices, rather than the sheer density of events that need to be accomplished.

While the Printer Clock focused on emphasising the embodied and situated nature of time, pointing to the mesh of activities and characters that come together to create time, the TimeBots drew attention to personal rhythms and how they played out within the context of the classroom

Thus in developing a theoretical framework which could support an understanding of time as multiple, heterogeneous and deeply entangled within various social formations (which may be discrete or overlapping), work in the social sciences, particularly anthropology and sociology, has proven to be more useful. Such approaches enable us to ask different questions about what time is and how it works. Rather than seeing time as a flow between past, present and future (whether this be linear or nonlinear), it becomes possible to ask how time operates as a system for social collaboration (Sorokin and Merton 1937), how it legitimates some and ‘manages’ others (Greenhouse 1996), or how it works within systems of exclusion (Fabian 1983). We thus move from time as flow to time as s