The Dynamics of Material Artifacts in Collaborative Research Teams

The Dynamics of Material Artifacts in CollaborativeResearch Teams

Deana D. PenningtonDepartment of Biology, University of New Mexico, MSC03 2020, Albuquerque, NM 87131-0001,USA (Phone: +1-505-2772595; Fax: +1-505-2772541; E-mail: [email protected])

Abstract. Boundary objects are material artifacts that mediate the relationship between two ormore disparate perspectives. The concept of boundary objects has been demonstrably useful in avariety of research areas; however, the meaning and function of boundary objects is contested. Atissue is the relationship between boundary objects that negotiate between perspectives and thosethat specify across perspectives. In this study the changing nature of boundary objects incooperative work is related to the dynamics of evolving problem conceptualization, system design,and enactment within cooperative work settings. Design based research on material artifactsproduced by an incipient cross-disciplinary research team during their efforts towards negotiatingintegrated conceptualizations and specifying shared research agendas is used to generate a morecomprehensive model of boundary objects through the life of a project.

Key words: eScience, eResearch, cross-disciplinary collaboration, boundary objects, materialartifacts, design based research

1. Introduction

Star and Griesemer (1989) proposed two central activities for translating betweendifferent viewpoints in cooperative scientific work: methods standardization andboundary object creation. The first is critical in that established guidelines form amanagerial system to which diverse allies can contribute concurrently. Thesecond arises after the first. Once guidelines are established, boundary objects canbe constructed. They conceived of boundary objects as an “analytical concept ofthose scientific objects which both inhabit several intersecting social worlds andsatisfy the informational requirements of each of them.” Boundary objects werethe point of intersection between viewpoints, and the process of creating andmanaging them was critical for maintaining coherence of the heterogeneous worksystem. Star and Griesemer (ibid.) found four types of boundary objects in theirstudy of a natural history museum, which they recognized was not exhaustive: 1)repositories, 2) ideal types, 3) coincident boundaries, and 4) standardized forms.

Subsequent work has demonstrated both the utility of the boundary objectconceptualization, and its shortcomings for explaining the full range of usage ofartifacts in practice. Case studies have further articulated characteristics of thesefour types, and identified new types (Fujimura 1992; Perry and Sanderson 1998;

Computer Supported Cooperative Work (2010) 19:175–199 © Springer 2010DOI 10.1007/s10606-010-9108-9

Roth and McGinn 1998; Bowker and Star 1999). Boundary objects as theoreticalconstructs have become common in social science studies in general, and inCSCW studies in particular. The critical role of material artifacts in coordinatingwork practices is undisputed. However, Lee (2007), based on an ethnographicstudy of incipient collaborative work, implicated an impoverished view ofboundary objects as a barrier to more comprehensive understanding of the role ofmaterial artifacts in collaborative work. She suggested material artifacts play analternative role as ‘boundary negotiating objects’. Boundary negotiating artifactsare not constructed from standardized processes; rather, they assist in theestablishment of standard processes. They are used to “record, organize, exploreand share ideas; introduce concepts and techniques; create alliances; create avenue for the exchange of information; augment brokering activities; and createshared understanding” (p. 333). She suggested that boundary objects may befound primarily in fairly routine or fairly simple work projects, while boundarynegotiating objects may be more prevalent in projects that are fairly non-routineand fairly complex, and that more studies of incipient collaborations might helpresolve these issues.

This paper presents a study of artifacts in incipient collaborative work in across-disciplinary science and technology research team. The goal is to develop abetter understanding of the relationship between boundary negotiating objects(sensu Lee 2007) and boundary objects (sensu Star and Griesmer 1989). In thisarticle the latter will be referred to as “boundary specifying objects” for clarity,and both types will be referred to as boundary objects (Figure 1). The questionsof interest are “What is the relationship between boundary objects that negotiateand those that span viewpoints?”; “What is their relationship with methodsstandardization?”; and “How do each enable collaborative work?” The paperdevelops a fuller, dynamic system account of the role of different kinds ofartifacts in different phases of collaborative work. This framework provides abasis for integrating the diverse empirical observations of this and prior work.

Figure 1. Terminology used in this paper. The set of all material artifacts contains a set ofartifacts that constitute boundary objects—artifacts that bridge different viewpoints. Twoclasses of boundary objects have been recognized: 1) those that specify viewpoints and fullymediate their interaction, and 2) those that negotiate interaction between viewpoints. Thequestion of interest in this article is, “What is the relationship between boundary negotiatingand boundary specifying objects?”.

176 Deana D. Pennington

2. Methods

2.1. Study context

This study is based on a National Science Foundation Cyberinfrastructure (CI)Team project conducted from 2006 to 2009. The goal of the CI-Team program isdevelopment of cross-trained scientists, engineers, and computer scientists able todesign, develop, and use emerging technologies. This CI-Team project (http://scidesign.org) was designed to explore potential mechanisms for enablingcollaborative research between scientists and computer scientists (eResearch oreScience). The overarching goal of such research is not just to support the workof scientists, but rather to collaboratively conceptualize and design innovativeapproaches in both science and technology (Lawrence 2006). Problem concep-tualization must emerge during interactions between participants. The process forachieving this must include mechanisms for evolving individual ideas intocollectively constructed agendas; advancing vague, ill-defined issues into welldefined, tractable problems; and migrating independent researchers into unifiedcognitive systems. These must happen more or less simultaneously, as each isdependent on the others.

A key factor known to influence the outcome of such collaborations is thedevelopment of common ground between participants (Olson and Olson 2000).Common ground is the mutual knowledge, beliefs and assumptions that form thebasis for meaningful communication (Clark and Brennan 1991). The CI-Teamproject focused on building common conceptual ground that could lead to betterintegrated research efforts between diverse scientists researching broad-scaleecological problems, and computer scientists interested in developing distributedsystems. Specifically, participants had to develop an understanding of eachother’s research interests, which depended on effectively learning relevantvocabulary and concepts across disciplines. Based on that understanding, theyhad to formulate conceptual linkages between each other’s research interests,develop integrated conceptual frameworks, and conceive of innovative collabo-rative research.

Participants in the group were recruited based on potential researchcontribution to the general problem area of climate change and vegetationimpacts in the American Southwest, and included twenty five participantsroughly equally divided between several scientific (ecology, geography, anthro-pology) and computer science (visualization, data mining, human factors,distributed computing) disciplines. Participants were a mixture of faculty, theirstudents, and researchers. Common ground was built across these diverseperspectives using a set of group activities over the 3 year project periodmediated by the principal investigator. All activities included creation of materialartifacts. This article focuses on the changing nature of the material artifacts;however a brief description of the activities follows to depict the process thatproduced the artifacts.

177Collaborative Research Teams

http://scidesign.org


Four consecutive group activities were conducted: 1) CI-Seminar, 2) CI-VisionWorkshop, 3) CI-Strategy Workshop, and 4) CI-Design Workshop. The goal ofthe CI-Seminar was to expose the scientists to ongoing technical work in manyexisting eScience projects, reflect on how these emerging technologies could berelevant for their own research efforts, and develop relevant technical vocabulary.The CI-Vision Workshop brought the scientists and computer scientists togetherto initiate learning about each other’s research interests and begin to develop ashared research vision. The CI-Strategy Workshop further developed that visionby identifying strategic research directions. Lastly, multiple CI-Design Work-shops targeted research proposal development in each strategic area. Although theoverall, four step process was linear, each workshop consisted of complexinteractions and iterative activities. Artifact creation was only one mechanismused to enable cross-disciplinary communication, but it was the most continu-ously used intervention throughout the project. A design-based research approachwas used to study the role of artifacts in this incipient collaborative team.

2.2. Design based research (DBR) approach

DBR is an extension of standard design and evaluation methodologies thatexplicitly maps interventions and their evaluation to extant theory, for the purposeof theory development (Figure 2). DBR was developed in classroom settings byeducation researchers in collaboration with teaching practitioners. For a completereview of DBR and its application in education, see the Special Issue of theEducational Psychologist devoted to DBR (Sandoval and Bell 2004). As Edelson

Figure 2. Logic model showing design based research approach and its application in thisstudy. Theories, with or without formative evaluation, are used to design an intervention.During enactment the designed intervention interacts with exogenous factors to produceoutcomes, some of which may have been expected based on theory and formative evaluation,others of which are unexpected. Summative evaluation takes place. Outcomes are explicitlymapped back to the theories that informed the design. The explicit link between scientifictheory and summative evaluation is what sets this approach apart from standard evaluationpractices.


(2002) stated, DBR “explicitly exploits the design process as an opportunity toadvance the researchers understanding.” Some CSCW projects already tacitly usethe design process in this way. DBR formalizes the approach by making explicitthe mapping between theories being tested, designed interventions andintervention outcomes.

DBR is a synergistic methodology with case studies and controlledexperimentation, which provide a wealth of valuable information about factorsthat must be considered in real situations and the processes they influence.However, they do not provide practical guidance about how the knowledgegained can be integrated in a real situation in order to achieve specific outcomes.As Sandoval (2004) states, “there is a tension between the desire for locallyusable knowledge on the one hand and scientifically sound, generalizableknowledge on the other.” DBR integrates findings from case studies andcontrolled experiments into contrived activities in real settings with two goals:1) development of theory-based conjectures that are used to manipulate groupactivities to produce targeted outcomes; and 2) refinement of theory based onevaluation of actual outcomes of those activities. Theory and practice are tightlycoupled in DBR. DBR experiments with group interventions, recognizing thatmany factors in live settings cannot be controlled and that important,unanticipated outcomes may emerge as a result of the intervention. Theseunexpected outcomes provide evidence that can be used in refinement of thetheory and concepts on which the original conjecture was based.

Construction of material artifacts is known to be an important part of enablinggroup interaction and therefore was incorporated as part of the design of the CI-Team intervention. The changing nature of the constructed artifacts, however,emerged during the process and was recognized as an important, unanticipatedoutcome. This paper focuses on that outcome and uses it to generate new ways ofconceptualizing the role of material artifacts. At the same time, it illustrates theDBR approach and provides an example of its utility in CSCW contexts.

2.3. Material artifacts: theory, concepts and conjecture

Frequently in DBR, theories and concepts from a variety of disciplinaryperspectives are applied to generate the conjecture from which an interventionmay be designed. There is a rich history of research on material artifacts acrossmany disciplines. Two, in particular, informed the design of the CI-Teamintervention: 1) material artifacts as boundary objects, and 2) material artifacts aslearning scaffolds.

Material artifacts as boundary objects. Voluminous literature exists on boundaryobjects as theoretical constructs and the application of boundary objects in a widevariety of settings. As mentioned in the introduction, boundary objects areconceived of as artifacts that capture and relate elements from more than one


perspective. In the original conceptualization of Star and Griesemer (1989),boundary objects align elements such that individuals can make use of them withno knowledge of any perspective other than their own. Conversely, Lee (2007)suggested that an alternative use of boundary objects is to negotiate and developunderstanding between perspectives. Both perspectives, and others, havedemonstrated the critical role of boundary objects in enabling informationprocessing in a group, and are needed in incipient research teams. On the onehand, developing cross-disciplinary understanding is a critical prerequisite todeveloping integrated conceptual frameworks, from which co-created researchideas can emerge. On the other hand, cross-disciplinary efforts need to proceed asmuch as possible with only limited understanding of other disciplinaryperspectives. Boundary negotiation must lead to boundary specification thateveryone can use to align their research.

Boundary negotiation and specification represent two different general kinds ofcognitive processes: divergent and convergent thinking. Divergent and conver-gent thinking processes were initially identified by Guilford (1967) as two of sixoperational intellectual processes. Divergent thinking is the ability to generate avariety of solutions to a problem. Convergent thinking is the ability to deduce asingle solution to a problem. In cross-disciplinary research, divergent thinking isneeded to explore the problem space in search of potential conceptualconnections. Convergent thinking is needed to evaluate those exploratoryconnections. Material artifacts can enable both kinds of thinking, but it isreasonable to expect that different kinds of material artifacts enable these differentkinds of processes.

Within CSCW, a substantial research agenda on knowledge cartography exists(Okada et al. 2008). These tools strive to map knowledge in order to betterunderstand and reason about information. Okada et al. (2008) categorized theseas concept mapping, argument and evidence mapping, issue mapping, webmapping, and thinking maps. These tools are similar in that they all enableexplication of thoughts by iconic representation of terms and connectionsbetween terms. They differ in the kind of reasoning or audience that they target.For example, concept maps enable representation of the concepts a map creatorconsiders relevant to a topic, and how he structures his knowledge about thattopic. Conversely, argument and evidence maps enable explicit representation ofthe pros, cons, and evidence related to an issue. Either of these could be useful fordivergent or convergent thinking activities, depending on how they are used. Forinstance, concept maps could be used to explore diverse perspectives on how toorganize the problem space, a divergent thinking activity. Conversely, conceptmaps could be used to specify how different perspectives are going to be relatedin a particular project, a convergent thinking activity. Artifacts produced fromeither of these approaches, if used within a cooperative group to mediateperspectives, are boundary objects. The kind of boundary object that theyrepresent (negotiating or specifying) depends on how they are used by the group.


Material artifacts as learning scaffolds. In cross-disciplinary research teams a setof individuals representing diverse perspectives must learn each other’s mentalmodels, learn how to fuse those differences into an integrated conceptualframework, and learn how to use that conceptual framework as a springboard tocollaborative problem-solving (Pennington 2008). This depends on exchange ofknowledge in ways that are conducive to making sense of a subject withoutrequiring depth of understanding. A fairly limited understanding is sufficient.However, it is impossible to know in advance which aspects will be relevant,requiring an exploratory phase of high-level learning that helps frame problemconceptualization. Unfortunately, disciplinary concepts are rarely accessible informs digestible by those from other disciplines lacking appropriate background(Jeffrey 2003). Some research has suggested that experts have specific cognitiveissues that limit their ability to convey their knowledge to novices (Hinds andPfeffer 2003). In particular, as depth of knowledge on a topic increases, expertsconceive of the content in increasingly abstract and simplified ways and areunable to revert to more concrete, detailed descriptions required by novices(Hinds and Pfeffer ibid.). Research in the learning sciences has demonstrated thatrepresentations, technological tools, activities, and/or physical artifacts are usedby more expert individuals in one-on-one learning situations as temporarysupports (scaffolds) that improve novice learning (Davis and Miyake 2004). Oncethe learner has grasped the target concepts, the scaffold is no longer necessary—differing from the permanent artifacts described as boundary objects. Whilematerial artifacts are not the only scaffolding mechanism, they play a critical rolein enabling the flow of information between individuals. There are two effects: 1)the effect on the recipient, for whom information is more easily grasped, and 2)the effect on the artifact creator, who must strive to construct an artifact that iseasily grasped. In striving to construct an easily grasped artifact, an individualmust organize his thoughts and conceptual frameworks in a way that he believeswill make sense to the recipient. This requires active thought about whatframeworks, other than his own, might make sense. The artifact creator reframeshis own mental models in ways that he thinks might resonate with the recipient.In so doing, he is engaging in his own internal manipulation activity that partiallyincorporates the recipient’s perspective. Hence, artifact construction entailslearning by both the creator and the recipient, and enables not just the flow ofinformation but also the dynamic creation of new mental models that containlinkages between participants. Artifact construction dynamically affordscollective learning (sensu Cook and Brown 1999).

Given the conceptual frameworks provided by boundary objects andscaffolding accounts of material artifacts, the conjecture was made that materialartifact construction could be a primary mechanism for enabling cross-disciplinary learning and discourse on eScience teams if used in structured,facilitated ways. Typically on these teams material artifacts are constructed inadvance (such as presentation slides) or occasionally on the fly (such as a white


board illustration). The conjecture was that centering all group activities on thecollective construction of specific kinds of material artifacts and facilitatingdiscourse around those artifacts would better enable the emergence of integratedresearch ideas. Construction of material artifacts during discourse has theconcomitant benefit of capturing the content being discussed, which can besubjected to content analysis and provide a mechanism for measuring the changein conceptual integration across disciplines. Therefore, if material artifacts wereindeed playing a central role in the exchange of knowledge and ideas then theircontent should reflect the transformation from separate, disciplinary perspectivesto more integrated cross-disciplinary frameworks through time.

2.4. Designing theory-based team interactions

Based on the above concepts, an intervention was designed to enable cross-disciplinary discourse through the creation of material artifacts. Several keystrategies were incorporated:

& All interactions between participants were centered around the creation ofmaterial artifacts.

& All material artifacts produced during the project were captured digitally andincorporated into a web accessible archive. The web archive both containsmaterial artifacts and itself acts as a material artifact.

& The CI-Seminar incorporated numerous examples of ways emergingtechnologies are being used in science. These were made explicit throughdigitally recording of presentations, and incorporation of slide presentationsinto the web archive.

& Subsequent interactions between scientists and computer scientists iteratedbetween 1) divergent thinking activities that supported exploration andlearning across the problem space and 2) convergent thinking activities thatstrove to scope specific research areas of interest. Artifact creation was anintegral part of both types of activities, and different types of artifacts werenecessary for these two different kinds of activities.

& Divergent thinking activities, e.g. exploratory learning, were enabled byconcept mapping, a commonly used scaffolding mechanism in classroomsettings (Novak and Wurst 2005). Concept maps are diagrams that captureassociations between concepts (Figure 3). The utility of concept maps as amechanism for enabling interdisciplinary discussion has been demonstrated(Heemskerk et al. 2003; Jeffrey 2003).

& Convergent thinking activities, e.g. scoping activities, targeted makingconceptual linkages explicit, identification of potential nexus points betweenresearchers, and progressive narrowing of linked research interests. Use of adiverse array of material artifacts (wikis, issue visualization, and white-boards) were explored.


2.5. Enactment and data collection

A wiki-based project website (http://scidesign.org) was designed using the opensource content management platform Drupal (http://drupal.org). This was linkedto the open source course management system Moodle (http://moodle.org). Thewebsite was used for participants to collectively edit shared documents of apublic nature while Moodle was used to manage the CI-Seminar and to organizeartifacts from topic specific research areas of interest. Moodle provides a widevariety of tools for group work, including document archives, links to sharedresources, forums, wikis and chat.

The CI-Seminar was held synchronously at three institutions in a videocon-ferencing environment (Macromedia Breeze; http://www.adobe.com): Universityof New Mexico, University of Arizona, and Northern Arizona University.Macromedia Breeze was selected because of its digital recording capabilities,lacking in available open source solutions such as AccessGrid (http://www.accessgrid.org). All scientists involved in the project attended the seminar; onlystudents were required to complete assignments. Each session was conducted byremote speakers from across the US and recorded in its entirety. Each remotespeaker provided an electronic copy of his presentation that was placed on the

Figure 3. Example concept map from a portion of a map created by a scientist.



http://drupal.org

http://moodle.org

http://www.adobe.com

http://www.accessgrid.org

http://www.accessgrid.org

website. Relevant papers were provided in Moodle prior to each session. Studentswere routinely assigned the task of writing about the content. They participated ina written online forum, responded to questions about technical developments intheir own field, and wrote about prospective applications of technologiesintroduced in the seminar in their own fields.

The CI-Vision Workshop was held shortly after the seminar concluded. Bothscientists and IT experts were involved, and the meeting consisted of a divergentthinking activity designed to learn about each other’s research interests andexplore possible connections followed by a convergent thinking activity to createa shared vision of integrated research. During this workshop concept maps wereconstructed by each participant regarding their research interests (Figure 3), afterbrief training on the approach and the open source CMap Tools software (http://cmap.imhc.us). Each participant talked about their own research using theirconcept map as a scaffolding device. Participants were asked to keep a runninglog of potential pairwise linkages between themselves and others, a convergentactivity. When every participant had presented and exchanged ideas, eachindividual posted their list of potential linkages on a shared wiki, from which acollective list of potentially interesting research topics was created. That list, andthe activity of generating the list, was used to initiate discussion of a broadresearch vision that encompassed those linkages. The final outcome of thisworkshop was a written shared vision that was rather ill-defined, but which wascollectively generated and incorporated the individual research interests of mostparticipants. Artifacts produced by the CI-Vision Workshop included a repositoryof individual concept maps, a shared wiki that included individually identifiedlinkages, and a short statement of the shared research vision.

The CI-Strategy Workshop targeted further refinement of the shared vision.This workshop began with a restatement of the shared vision and discussion ofpotential nexus points, which in this case revolved around methods for dataanalysis and visualization. Participants were asked to draw concept maps of theirview of data, analysis and visualization in the context of the shared vision, usingseed terms that were suggested by one of the participants (a divergent thinkingactivity). Eight seed terms were given: distributed data, data analysis, dataproducts, online modeling, modeling processes, decision support, visualizationtools, and animation tools. Participants could elect to use any or none of these.Each concept map was collectively viewed and discussed, and an integratedconceptual workflow was generated (a convergent thinking activity). Dialogueand interactions around these concept maps led to identification of three separatecross-disciplinary research areas of interest; identified content for two new virtualseminars to develop better cross-disciplinary understanding in those areas;considered specific training that would enhance the ability of the scientists toenvision usefulness of certain technical approaches; and creation of a noveleducational approach for engaging computer science and science studentstogether in a collaborative classroom setting. Material artifacts from this


http://cmap.imhc.us

http://cmap.imhc.us

workshop included the concept maps, an integrated conceptual workflow, andwiki notes about all of the areas of interest identified above.

Several months later, a CI-Design Workshop was held to design and develop aspecific research proposal around one of the identified research areas. Thismeeting involved a smaller subset of participants (10). The meeting resulted in across-disciplinary research design that was collectively generated and engaged allof the participants. During the CI-Design meeting participants sometimes referredto shared artifacts from prior meetings, and created new ones as needed. Theseartifacts were primarily freeform drawings on whiteboard that were either copiedinto electronic drawing software or were photographed with a digital camera.Near the end of the meeting a draft proposal outline was collectively generatedthrough one person editing a wiki and projection of the text onto a screen. Thefinal draft was individually edited until an agreed upon draft emerged.

2.6. Data analysis

Artifacts were classified as individually or group constructed. The number ofindividual and group produced artifacts was counted. Artifacts from the CI-Seminar were excluded because only the students created them. Each remainingmaterial artifact was visually inspected and categorized as: 1) boundarynegotiating object, or 2) boundary specifying object. This categorization wasbased on how the artifacts were used by the group, following the distinctionidentified by Lee (2007). Boundary negotiating objects were used to “record,organize, explore and share ideas; introduce concepts and techniques; createalliances; create a venue for the exchange of information; augment brokeringactivities; and create shared understanding.” Conversely, boundary specifyingobjects merged different perspectives in ways that removed the need to interactacross disciplines. Artifacts collected were assessed for analytical compatibility.Concept maps from the CI-Vision and CI-Strategy Workshop were matched forcomparative analysis of individually produced artifacts. The vision statementfrom the CI-Vision Workshop was matched to the CI-Design proposal forcomparative analysis of group produced artifacts.

Artifact content was assessed through semantic analysis as a surrogate measureof cross-disciplinary conceptual integration. Terms were extracted from eachartifact and classified as 1) unique to science language, 2) unique to informationtechnology language, or 3) shared by both. The percentage of terms in each ofthese three classes for each artifact was tabulated.

Pairwise comparisons of individual concept maps collected during the CI-Vision and CI-Strategy Workshops were conducted using functionality providedby CMap Tools. Only concepts in each map (nodes) were compared; connections(links) were excluded because many participants did not specify connectionsbetween concepts, or used semantically-neutral phrases such as “leads to”.Matches were determined through four mechanisms: 1) full text match, 2) partial


text match, 3) keyword, and 4) synonym. Keyword and synonym matches weredetermined based on thesauri and other resources internal to the software. Allpotential matches identified by the software were manually checked; erroneousmatches were excluded. Matches based on seed terms provided during the CI-Strategy workshop were excluded. Four concept maps were discarded from theanalysis because it was obvious that they had collaborated on parts—sections oftheir maps were duplicated. A fifth—the one that was the most comprehensive—was retained. Because the number of terms used in each concept map varied,counts for a given pair differed depending on which concept map was consideredthe source and which the target. Therefore counts of matches were made in bothdirections.

Pairwise comparisons of terms used in the concept maps were classified as acomparison between 1) two scientists, 2) two information technologists, or 3) oneof each. This classification was further generalized as 1) within disciplinecomparison and 2) between discipline comparison. Basic statistics for each ofthese were calculated: number of samples, mean count of terms in common,median count, variance and standard deviation. The data were not normallydistributed, contained unequal numbers of samples, and the occurrence of valuesof zero prevented log transformation. Therefore the non-parametric statisticalrank sum test was used to analyze differences between groupings. Differencesbetween groups through time were compared.

Artifacts were again visually checked to verify that the findings highlighted bythe statistical analyses were valid, and to qualitatively assess the ways anddegrees to which artifacts not analyzed diverged from those findings.

3. Results

The total number of artifacts produced during the project period was 111. 20 ofthese were produced by the group and 91 by individuals (Figure 4). Ignoring theCI-Seminar artifacts which were all produced by individuals by design, thenumber of artifacts produced by individuals decreased in number through timefrom 34 during the CI-Vision Workshop to 0 during the CI-Design workingmeeting. The number of artifacts produced by the groups varied through time,peaking mid-project.

Basic statistical analysis of terms used in concept maps indicates that thenumber of overlapping terms between paired concept maps was higher for alldisciplinary combinations during the CI-Strategy Workshop than during theearlier CI-Vision Workshop (Table 1). During both the CI-Vision and CI-StrategyWorkshops, the highest observed number of terms in common was between pairsof scientists (CI-Vision mean = 1.6, median = 1 term in common; CI-Strategymean = 5.7, median = 5 terms in common). Pairs of information technologyspecialists had fewer terms in common than pairs of scientists (CI-VisionWorkshop mean = 1.2, median = 1 term in common; CI-Strategy Workshop


mean = 3.8, median = 4 terms in common). Interdisciplinary pairs had the lowestnumber of terms in common during the CI-Vision Workshop (mean = 0.3 and 0.6,median = 0 terms in common). During the CI-Strategy Workshop they had fewerterms in common that pairs of scientists but more terms in common than pairs of

Figure 4. Counts of the number of material artifacts produced by individuals and by thegroup in four consecutive activities.

Table 1. Basic statistics from semantic analysis of concept maps created during two activities.

Withindiscipline

Betweendiscipline

Comparison of withinand between disciplines

SCI/SCI IT/IT SCI/IT IT/SCI

CI-Vision WorkshopSample size

90 20 50 50Significant differencez=−6.25 P << 0.0002

Mean number of termsn common 1.6 1.2 0.3 0.6

Median number of termsin common 1 1 0 0

Variance2.6 1.9 0.4 1.5

CI-Strategy WorkshopSample size

30 6 18 18No significant differencez=−1.22 P=0.11

Mean number of termsin common 5.7 3.8 4.6 4.4

Median number of termsin common 5 4 4 4

Variance17.1 1.8 8.4 4.3


information technology researchers (mean = 4.6 and 4.4, median = 4 terms incommon). Concept maps from the CI-Vision Workshop showed a high significantdifference in the number of overlapping terms used by individuals from withinthe same discipline (two scientists or two information technology specialists)compared with the number of overlapping terms used by individuals fromdiffering disciplines (z=−6.25, P << 0.0002; Table 1). In contrast, there was nosignificant difference between the number of overlapping terms used byindividuals within the same discipline or differing disciplines at the later CI-Strategy Workshop (z=−1.22, P=0.11; Table 1). These statistical results indicatethat during the CI-Vision Workshop the terms used by two people from the samediscipline were more similar that the terms used by two people from differentdisciplines. During the CI-Strategy Workshop there was no difference in the termsused by two people from the same discipline and two people from differentdisciplines. The larger degree of overlap in terms overall, and increasingsimilarity in terms used by cross-disciplinary pairs, suggests that more integratedconceptual frameworks did evolve through the time period represented by thesetwo workshops. There was no reported interaction between participants duringthis time frame outside of the workshops; hence one can conclude that theactivities conducted within the workshops, including creation of materialartifacts, were responsible for enabling the construction of shared conceptualframeworks.

The group-produced vision statement from the CI-Vision workshop contained17 terms unique to either science (7 terms, 41%) or information technology (10terms, 59%). The group produced proposal contained 58 terms unique to eitherscience (32 terms, 55%) or information technology (26 terms, 45%). 3 newlycreated compound terms were used in the proposal to represent complex contentthat integrated concepts from both science and technology. These terms werederived from a whiteboard artifact created during the CI-Design working meeting(Figure 5). The phrase “Strategic Data Analysis” was created to refer to areasoning and questioning process that was used to align relevant methodologiesfrom science and technology, “Literate Workflows” was created to refer to thecombination of methods (both scientific and technical) that the research targeted,and the phrase “Rasters to Reason (R2R) was created to refer to the combinedquestioning and analytical process.

This diagram (Figure 5) was the first boundary specifying object to emerge; allprevious artifacts were used for boundary negotiation. Any of the participantscould refer to the diagram and understand their role in the integrated researchwithout understanding the other disciplinary perspective. However, those whoparticipated in the creation of this diagram did have to understand each others’perspectives. The diagram not only captured the outcome from all of the priornegotiation that led the group to the point where they were able to align theirmethods in this way, the construction of this artifact itself dynamically affordedthe alignment of methods. Hence, from the perspective of the artifact creators,


this diagram served as both a boundary negotiating object (during its creation)and also a boundary specifying object (afterwards). In the same way, the proposalalso served as both a boundary negotiating and boundary specifying object. Noneof the artifacts created by the team were used solely as boundary specifyingobjects. However, from the perspective of technology developers and/or scientistsnot involved in the research design process, the diagram and proposal couldpotentially act solely as boundary specifying objects. These individuals aretypically brought into a research project after it has been funded. They are notinvolved in the production of the proposal or the artifacts leading up to theproposal. Hence, the possibility exists that a piece of research or softwaredevelopment identified in the diagram and proposal could be worked on by anindividual without any further negotiation across disciplines. They would simplyrefer to the proposal or diagram to know how their piece fit into the bigger cross-disciplinary project. To them, the proposal and diagram would be solely boundaryspecifying objects.

4. Conceptual reformulation and discussion

4.1. Role of boundary objects with respect to methods standardization

Star and Griesemer (1989) initially conceived of methods standardization asessential to collaboration. Boundary objects were artifacts produced from thenegotiation of those standards, enabling diverse groups to interact via the artifactwith little or no knowledge of each other’s use of the artifact. Standards wereessential for achieving this outcome. Yet substantial collaborative work mustoccur before methods can be standardized. It is in that situation that Lee (2007)

Figure 5. Conceptual model of a system enabling interactions between analytical reasoningprocesses, scientific methods, and information technology tools. This diagram became theprimary mechanism for aligning scientific and technical methodological approaches.


conceived of artifacts as boundary negotiating objects. Many of the empiricalstudies of boundary objects in the CSCW literature describe artifacts that are usedfor boundary negotiation rather than specification. For instance, Bodker andChristiansen (2006) identified use case scenarios as boundary objects. Eckert andBoujut (2003) explored the use of design artifacts as boundary objects. Herrmannand Hoffmann (2005) described workflow description as boundary objects.Boujut and Blanco (2003) conceived of intermediate artifacts as boundary objectsthat mediate, transform and re-represent information from one perspective into aform that is understandable by another.

The progressive alignment and standardization of methods depends onprogressively integrating conceptualizations of the target problem (Figure 6). Inthe case of eScience research agenda setting, for example, scientists may have aparticular analysis they frequently conduct that requires certain kinds of effort.Computer scientists may have advanced techniques that could replace parts ofthat effort. There is, potentially, an alignment of methods. In order to make thatmatch, scientists and technology experts must articulate their research methods ina way that allows at least one participant to identify the conceptual linkagebetween methods. Collaborative problem conceptualization is a process ofiteratively searching and discovering methods that potentially align, and thearticulation of those methods in a format that all collaborators understand and canmake use of. Hence, the role of boundary negotiating objects during incipientcollaboration is to help generate conceptual linkages from which aligned andstandardized methods can emerge.

Figure 6. Conceptual model of interaction between concepts, method alignment, andboundary objects. Participants iteratively explore concepts from different perspectives usingboundary negotiating objects until they generate enough conceptual linkages that they areable to align their methods. Once methods are aligned and integrative concepts are available,boundary specifying objects can be produced. This process may also generate new boundaryconcepts that represent linkages between disciplines, which can potentially evolve into newinterdisciplinary knowledge domains.


Method alignment and standardization can itself be captured in an artifact. Forinstance, the whiteboard diagram (Figure 5) represented the alignment of methodsacross disciplines. These artifacts simultaneously represent the outcome fromboundary negotiation (boundary negotiating object) and a source to which eachparticipant can refer without knowledge of other perspectives (boundaryspecifying object).

4.2. Emergence of boundary concepts

As conceptual linkages are made and methods are aligned, new boundaryconcepts emerge (Figure 6). These concepts reflect complex integratedconceptualizations drawn from both perspectives, and new terms may be createdthat become a shorthand way of referring to them. These boundary concepts are anecessary predecessor of method alignment. However, they can potentiallyfurther evolve into a new knowledge domain—a transformational learningoutcome. As the number of conceptual linkages increase, integrated conceptualframeworks evolve and the likelihood of emergence of novel, integrated problemconceptualizations increases as well. In that case, rather than participantsinteracting without knowledge of each others’ perspective, a new, integratedperspective is established. Duncker (2001) proposed symbolic coupling andtransformation mechanisms that lead to the emergence of boundary concepts andassociated terminology and artifacts in cross-disciplinary settings, noting thathybrid concepts that juxtapose specialties and are understood by all participants,if used over a long period of time, can lead to the emergence of new symbolicregimes. Therefore, there are two potential paths that this process may take: 1)conceptual links are developed that enable the alignment of methods, from whichboundary specifying objects can be created that enable group members to workindependently; and 2) conceptual links are developed that enable the alignment ofmethods, from which a new, integrated perspective emerges that enables groupmembers to work synergistically.

Such transformation has been observed in numerous case studies. For instance,shortly after Star and Griesemer (1989) initially proposed the concept ofboundary objects, Bud (1991) observed that the term “biotechnology” serves asa boundary object between biology and engineering domains. He noted that whilethe meaning of the term was highly ambiguous, the innovative conceptuallinkages that emerged surrounding it led to a range of new interdisciplinaryapproaches. Bud conceived of the term as a boundary specifying object unitingmany different, but related, interdisciplinary conceptualizations. Upham (2000)found nearly identical results in a study of The Natural Step (TNS), aninternational approach to achieving consensus on sustainable development.TNS itself was found to be highly misunderstood by the professional community,but effective in uniting a number of different organizations involved inenvironmental impact reduction. In both of these studies, the boundary


negotiating process led to the emergence of new conceptualizations (boundaryconcepts), new terms representing those conceptualizations (boundary terms) thatthemselves act as boundary objects, and new artifacts explicating those terms andconceptualizations (boundary objects). The conceptualizations themselves con-tinued to evolve in many different ways, transcending the original meaning of theterms and artifacts. Indeed, this is exactly the path that the term “boundaryobject” has taken. The conceptualization of boundary objects has evolved past theoriginal meaning, yet the term “boundary object” is itself a boundary object thatunites many different but related conceptualizations!

I agree with Lee (2007) that the original conceptualization of boundary objectfalls far short of describing the range of material artifacts that are used in settingsthat merge different communities. Lee chose to constrain the term “boundaryobject” to its original conceptualization and create new terms for other classes ofartifacts. I think it makes more sense to recognize that the conceptualization hasgrown and the meaning of the term with it, given that analogous situationssuggest it is not possible to restrain that evolution. I believe that the term“boundary object” should refer to any artifact that is used to cross communityboundaries, whether it is used for negotiation, for specification, or for any otherboundary crossing process. Regardless of how it is used, it is an artifact at theboundary between communities. The artifacts described by Star and Griesemer(1989) were merely a subset of the class which I have referred to as “boundaryspecifying objects” in this article. The boundary negotiating artifacts of Lee areanother subclass (Figure 1). There are potentially other subclasses that should beincluded. For instance, is a term an artifact? If so, then perhaps there should be asubclass of boundary objects for boundary terms. More explicit understanding ofthe dynamic interactions between concepts, terms, artifacts and cooperative workthrough time is needed, rather than focusing just on the artifacts.

4.3. Dynamic model of boundary objects

Hutchins (1995, following Marr 1982: 24–27) distinguished three levels ofdescription of an information processing system. He referred to these as thecomputational, algorithmic, and implementation levels of description. Since theseterms have many meanings in different communities, I will refer to these as thetask, approach, and enactment levels, respectively. The highest level, task, is thefunctional definition, or what the system is supposed to do (Magnus 2007). Thisis the level at which the CI-Team group began. It is also the initial level of thedesign team described by Lee (2007). In incipient collaborations, this task levelmust first be defined, and conceptual linkages that support the task must bedeveloped and integrated. The second level, approach, describes how the task isgoing to be accomplished. There may be many different strategic and tacticalapproaches that could be used for any given task. The third level, enactment, isthe level at which the work is actually completed. Given a particular approach, is


must be instantiated in concrete ways. To illustrate, a task level statement mightbe that we are going to design a system for sharing mental models. The approachlevel might be that we are going to have participants construct and share conceptmaps. There are other strategic ways that mental models could be shared (such asa shared wiki, for instance), and other tactical mechanisms (such as issuevisualization, for instance). The enactment level is the creation of the specificconcept mapping software (such as IHMC CMap Tools, for instance) and themechanisms invoked to operationalize use of it. In an existing, stable systemthese three levels are already specified. In a nascent system, they must bespecified, and that can only occur in a top down manner. Enactment detailscannot be specified until the task is understood, and the approach to be used hasbeen defined. Therefore, in the course of a project, work on these three levels istemporally based.

Specification of these three levels is the ‘articulation work’ described bySchmidt and Simone (1996). They distinguished between cooperative work,which is “constituted by the interdependence of multiple actors who interactthrough changing the state of a common field of work” and articulation work,which is “constituted by the need to restrain the distributed nature of complexlyinterdependent activities.” This describes a process of boundary negotiation, inthat the goal is to articulate the relationships between distributed activities. Therelationships cannot be articulated until they are discovered and negotiated. Theirexample of cooperative work includes the use of boundary specifying objects thatenable participants to work independently. Conversely, they note that articulationwork is more complex, should be addressed as an issue in its own right, and isinherently recursive in nature. That is, within any articulation task there arenested articulation tasks—the recursive task, approach, and enactment levelsnoted above as well as recursive processes within each of these. Schmidt andSimone note that in complex articulation work settings everyday social andcommunication skills are insufficient, and “cooperating actors typically use aspecial category of artifacts which, in the context of a set of conventions andprocedures, stipulate and mediate articulation work and thereby are instrumentalin reducing its complexity and in alleviating the need for ad hoc communication.”This suggests that the articulation process, a boundary negotiation process, isenabled by artifacts that mediate articulation work. This describes boundarynegotiating objects. To summarize the discussion so far, group work can beseparated into two synergistic types: articulation work that is iterative andrecursive, and includes task identification, approach delineation, and enactment,and cooperative work that is achieved through prior articulation work (Figure 7).

Task and approach articulation work are accomplished by recursive mediationand negotiation processes nested within each of these (Figure 7). Boundarynegotiating, method alignment, method standardization, and boundary specifyinghave temporal contexts that are related to the system articulation or cooperativework that is occurring. The artifacts that are produced by the group will take on


characteristics of the processes that produced them. Once boundaries have beenspecified, artifacts will be more permanent taking on the characteristics describedby Star and Griesemer (1989), and will enable cooperative work in the absence ofcross-perspective understanding. Boundary specifying objects are not static,however. They are continually modified and reused as the system is improved andresponds to a changing external environment (Lutters and Ackerman 2007).Additionally, the complexity of the work being carried out will determine the needfor flexible boundary objects. Routine sequential collaborative work necessitatesrigid boundary objects that require no context or additional information for use.More complex, integrated work, such as scientific collaboration, necessitates moreflexible boundary objects.

At the other end of the continuum, collaborative groups in the problemconceptualization phase cannot construct boundary specifying objects becausethere is no standardized methodology, nor are these particularly needed duringthis phase. Rather, participants need ways to facilitate communication of theirindividual viewpoints and negotiate a shared conceptual space. Boundarynegotiating objects are useful for this purpose. Lee (2007) described twosubclasses of boundary negotiating objects that are particularly useful for this:self-explanation and inclusion artifacts. Self-explanation artifacts are used byindividuals for “learning, recording, organizing, remembering, and reflecting” ontheir own private perspectives. Examples given include private journals and

Figure 7. Diagram showing relationship between relevant processes at different scales andtool/artifact dynamics. Development of radically new systems depends on interactionsbetween IT experts and domain experts. These interactions can be mediated by boundarynegotiation objects that help define the tasks to be supported, and that integrate boundaryconcepts, method alignment, and method standardization. Boundary specifying objects maybe produced from these and incorporated into the system that is implemented. Normativechange requires rethinking of the system, which could be simple improvements, incrementalinnovation, or radical innovation. Incremental and radical innovation reset the mediationprocess back to appropriate points in the negotiation phase.


tables. Sometimes information captured in these are indirectly shared throughinclusion artifacts, which are used to propose new concepts and forms from oneperspective to other perspectives. In the CI-Team project, all divergent thinkingactivities involved individuals creating self-explanation artifacts that were thenused as inclusion artifacts for the purpose of learning scaffolding.

During boundary negotiation, artifacts may be temporary, partial, may or maynot represent all perspectives, and are used to articulate the cooperative task andapproach. They serve a critical purpose, as scaffolds for learning and negotiation,at that point in time. For instance, in the CI-Team project, concept maps producedby individuals represented their perspective only, captured only a few terms thatdescribed their research interest, and were used to provide high level conceptualcontext that other individuals could understand and enable learning. They enableddiscussion and negotiation of conceptual linkages between participants. Boundarynegotiating artifacts may or may not be needed in later phases, and may or maynot evolve into other kinds of artifacts. In the CI-Team project participantscommonly referred back to earlier boundary negotiating artifacts as a reminder ofearlier discussions that had taken place and to renegotiate themes that wereinteresting but had not been followed up on by the group. Conversely, there wereartifacts that were never referred to again after their initial construction. Even ifthey were discarded, they certainly contributed to later artifacts in the sense thatthe conceptual linkages that enable construction of later artifacts were themselvesenabled by earlier artifacts. Boundary negotiating artifacts may eventually evolveinto later boundary specifying objects. The diagram (Figure 5) and proposal thatwere generated through negotiation across disciplines will be more fullydeveloped into a project plan that specifies the work to be carried out in moredetail. This project plan will enable some participants to work independentlywithout knowledge of work in other disciplines, though that work will be alignedthrough the project plan.

In between boundary negotiation and boundary specification, method alignmentand standardization processes may produce a set of artifacts that are both boundarynegotiating and boundary specifying, depending on the perspective of theindividual using the artifact. From the perspective of those producing the artifact,artifact creation dynamically affords negotiation of method alignment andstandardization. From the perspective of those who use the artifact later, it specifiesthe boundary. As described earlier, the diagram and proposal are examples ofmixed-use boundary objects. Construction of these artifacts requires context,understanding, and agreement between constructors, even if later collaborators canuse the artifact without any knowledge of other views. These artifacts provide“anchors” that stabilize and constrain subsequent artifact creation.

The role of artifacts is more complex than simply providing a path forinformation flow. Artifact construction dynamically affords the construction ofconceptual linkages. The process of constructing artifacts is as important, orperhaps more so, than the end product. It is the co-creation of agreed upon


artifacts that forces collaborators to engage each other, not simply to understandalternative views but to construct new mental models that integrate that view intothe listener’s own conceptual framework. Again, the diagram (Figure 5) andproposal illustrate this concept. Hence, one can conceive of the collective cognitivesystem growing through the creation of conceptual linkages between participantsfacilitated by creation of material artifacts. Once the conceptual linkages arepresent, the artifact is either no longer necessary, or it must change to affordadditional linkages. Dynamic linking of mixtures of mental concepts and artifactsis a scaffolding process, using temporary frameworks that support and provideaccess to meaning. Once the cognitive framework is in place, the scaffold is nolonger necessary. Artifact-mediated negotiations between collaborators throughoutall phases of system development are instances of scaffolding processes.

The position of a team along the continuum from problem finding to enactmentis not static, nor is it unidirectional. Rather, it is recursive. A new team begins atthe task level and migrates through the approach and implementation levels. Atany time, changes in understanding may require repositioning back to an earlierlevel, and renegotiation of boundary objects. The effect of internal and externalchanges on organizational processes and concomitantly on boundary objects wasnoted by Subrahmanian et al. (2003), who concluded that every structural orinformation flow change in an organization is potentially accompanied by adeterioration of common ground perspectives, which must then be renegotiated.For instance, the CI-Team project iterated multiple times between task andapproach as different strategies for integrating across disciplines were negotiatedand proposed. If and when funding is received, during enactment a much betterunderstanding of the problem will be obtained, which may require rethinking theapproach level, or perhaps even redefining some aspects of the task. This wouldreflect an internal change of organizational processes. Alternatively, working,enacted systems may encounter external changes in the work process that requireimprovements, such as a technical breakthrough that changes the most effectiveapproach. Innovation, in general, is the process of repositioning the system tohigher levels of description in order to achieve new system outcomes (Figure 7).Continuous or incremental innovation is repositioning to the approach level.Discontinuous, disruptive, or radical innovation is repositioning to the task level.Each of these requires renegotiation of common ground and reconstruction ofmaterial artifacts. The CI-Team project itself is a radical innovation based on anew vision of cross-disciplinary collaboration within eScience teams as a learningproblem with collective cognition issues in addition to previously recognizedsociotechnical issues.

5. Limitations and future work

The proposed model is tentative, based on study of one research team combinedwith empirical observations by another investigator from one ethnographic study


of a design team, and synthesized with empirical findings from a number of otherstudies. Additional data will hopefully validate, further refine, or contest themodel. The proposed model is simply a starting point for building more robusttheories of the dynamics of conceptualization, terminology, material artifacts andwork in collaborative groups. The model is biased towards groups in whichcooperative work is highly creative, autonomous, and targeted towards generationof new conceptualizations. It should be complementary with the many CSCWstudies of less creative situations, representing one end of the spectrum that iscurrently not well understood.

As with any design-based research, the methodology itself introducesuncertainties into the interpretation. For instance, it is difficult to nail down theeffect of the intervention. Did the unplanned designation of seed terms for the conceptmaps created during the CI-Strategy Workshop predetermine that the number ofoverlapping terms would increase, interpreted as representing the formation of newconceptual linkages? Alternatively, since these seed terms emerged from theparticipants themselves during discussion, perhaps the analytical methodology isstill valid. Would the same pattern of increasing overlap and boundary conceptemergence be observed without any intervention? Certainly, research teams withoutthe same intervention manage to develop conceptual linkages. However, perhapsthere are significant differences in the timing with and without intervention.Additional observations both with andwithout intervention are needed to disentanglethese issues.

6. Conclusions

The dynamics of changing external drivers, system design processes, andmediation level processes in cooperative groups provides a complex arena withinwhich artifacts may take on innumerable, changing forms. This study demonstratesthat artifacts not only assist in information processing, they dynamically afford thegeneration of new and integrated conceptual frameworks, and lead to progressiverefinement of the conceptualization of a problem from the task, to the approach, tothe enactment levels of system description. This evolution towards increasingspecification is reflected in an evolution in artifacts, from boundary negotiationobjects (sensu Lee 2007) towards boundary specification objects (sensu Star andGriesemer 1989). In incipient collaborations boundary negotiating objects enablethe construction of standard methods. In later collaborations with stable, enactedsystems, standardized methods allow the construction of boundary specifyingobjects. In between, artifacts that are created to capture and make explicitstandardized methods act as both boundary negotiating objects (during creation)and boundary specifying objects (afterwards).

Changes in the normative environment resets the cooperative work to a higherlevel of system description, from which new conceptualizations are renegotiated,new artifacts produced and a new evolutionary path is taken. Radical innovation


occurs when the system is reset to the task level, and interactions between people,mediation processes, and artifacts can effectively achieve new conceptualizations.CSCW studies that more explicitly address the dynamic linkages between humanthought and the role of artifacts in enabling collective creativity and negotiationin boundary situations are likely to be a productive line of inquiry leading toinnovative tools that can enable those processes. Technologies designed tosupport collaborative work generally target stable, enacted systems where thework can be described. There is a need for technologies that support the processof articulating the system, both in incipient collaborations and in the face ofchanging requirements.

Acknowledgements

This work was supported by National Science Foundation grant numbers0636317 and 0753336 for the CI-Team Demonstration and ImplementationProjects: Advancing Cyber-infrastructure Based Science Through Education,Training, and Mentoring of Science Communitie. The author gratefully acknowl-edges the many collaborators involved in these projects, whose comments andinsights have been useful.

1. References

Bodker, S., & Christiansen, E. (2006). Computer support for social awareness in flexible work.Computer Supported Cooperative Work, 15, 1–28.

Boujut, J.-F., & Blanco, E. (2003). Intermediary objects as a means to foster co-operation inengineering design. Computer Supported Cooperative Work, 12, 205–219.

Bowker, G. C., & Star, S. L. (1999): Sorting things out (pp 389). MIT.Bud, R. (1991). Biotechnology in the Twentieth Century. Social Studies of Science, 21(3), 415–457.Clark, H. H., & Brennan, S. E. (1991). Grounding in communication. In L. Resnick, J. Levine, & S.

Teasley (Eds.), Perspectives on socially shared cognition (pp. 127–149). Hyattsville, MD:American Psychological Association.

Cook, S. D. N., & Brown, J. S. (1999). Bridging epistemologies: the generative dance betweenorganizational knowledge and organizational knowing. Organization Science, 10(4), 381.

Davis, E. A., & Miyake, N. (2004). Explorations of scaffolding in complex classroom systems.Journal of the Learning Sciences, 13(3), 265–272.

Duncker, E. (2001). Symbolic communication in multidisciplinary cooperation. Science, Technol-ogy, & Human Values, 26(3), 349–386.

Eckert, C., & Boujut, J.-F. (2003). The role of objects in design co-operation: communicationthrough physical or virtual objects. Computer Supported Cooperative Work, 12, 145–151.

Edelson, D. C. (2002). Design research: what we learn when we engage in design. EducationalPsychologist, 39(4), 105–121.

Fujimura, J. H. (1992). Crafting science: Standardized packages, boundary objects, and“translation”. In A. Pickering (Ed.), Science as practice and culture (pp. 168–211). Chicago:University of Chicago Press.

Guilford, J. (1967). The nature of human intelligence. New York: McGraw-Hill.Heemskerk, M., Wilson, K., & Pavao-Zuckerman, M. (2003). Conceptual models as tools for

communication across disciplines. Conservation Ecology, 7(3), 8–17.


Herrmann, T., & Hoffmann, M. (2005). The metamorphoses of workflow projects in their earlystages. Computer Supported Cooperative Work, 14, 399–432.

Hinds, P. J., & Pfeffer, J. (2003). Why organizations don’t “know what they know”: Cognitive andmotivational factors affecting the transfer of expertise. In M. S. Ackerman, V. Pipek, & V. Wulf(Eds.), Sharing expertise: Beyond knowledge management (pp. 3–26). Cambridge: MIT.

Hutchins, E. (1995). Cognition in the wild (pp 381). MIT.Jeffrey, P. (2003). Smoothing the waters: observations on the process of cross-disciplinary research

collaboration. Social Studies of Science, 33(4), 539–562.Lawrence, K. A. (2006). Walking the tightrope: the balancing acts of a large e-Research project.

Computer Supported Cooperative Work, 15, 385–411.Lee, C. P. (2007). Boundary negotiating artifacts: unbinding the routine of boundary objects and

embracing chaos in collaborative work. Computer Supported Cooperative Work, 16, 307–339.Lutters, W. G., & Ackerman, M. S. (2007). Beyond boundary objects: collaborative reuse in aircraft

technical support. Computer Supported Cooperative Work, 16, 341–372.Magnus, P. D. (2007). Distributed cognition and the task of science. Social Studies of Science, 37

(2), 297–310.Marr, D. (1982). Vision: A Computational investigation into the human representation and

processing of visual information (pp. 397). Henry Holt and Co., Inc.Novak, J. D., & Wurst, M. (2005). Collaborative knowledge visualization for cross-community

learning. In: S. Tergan, & T. Keller (Eds.), Knowledge and information visualization. LectureNotes in Computer Science, 3426, 95–116.

Okada, A., Buckingham Shum, S., & Sherborne, T. (2008). Knowledge cartography (pp 400).London: Springer.

Olson, G. M., & Olson, J. S. (2000). Distance matters. Human-Computer Interaction, 15, 139–178.Pennington, D. (2008). Cross-disciplinary collaboration and learning. Ecology and Society, vol. 13,

no. 2, pp. 8 [online] URL: http://www.ecologyandsociety.org/vol13/iss2/art8/Perry, M., & Sanderson, D. (1998). Co-ordinating joint design work: the role of communication and

artefacts. Design Studies, 19(3), 273–288.Roth, W. M., & McGinn, M. K. (1998). Inscriptions: toward a theory of representing as social

practice. Review of Educational Research, 68(1), 35–59.Sandoval, W. A. (2004). Developing elarning theory by refining conjectures embodied in

educational designs. Educational Psychologist, 39(4), 213–223.Sandoval, W. A., & Bell, P. (2004). Design-based research methods for studying learning in

context: introduction. Educational Psychologist, 39(4), 199–201.Schmidt, K., & Simone, C. (1996). Coordination mechanisms: towards a conceptual foundation of

CSCW systems design. Computer Supported Cooperative Work, 5, 155–200.Star, S. L., & Griesemer, J. R. (1989). Institutional ecology, ‘translations’ and boundary objects:

amateurs and professionals in Berkeley’s Museum of Vertebrate Zoology, 1907–39. SocialStudies of Science, 19, 387–420.

Subrahmanian, E., Monarch, I., Konda, S., Granger, H., Milliken, R., Westerberg, A., et al. (2003).Boundary objects and prototypes at the interfaces of engineering design. Computer SupportedCooperative Work, 12, 185–203.

Upham, P. (2000). Scientific consensus on sustainability: the case of the natural step. SustainableDevelopment, 8, 180–190.


http://www.ecologyandsociety.org/vol13/iss2/art8/

The Dynamics of Material Artifacts in Collaborative Research Teams

Documents