using visual representations as boundary objects to resolve conflict in collaborative model-building...

■ Research Paper

Using Visual Representations as BoundaryObjects to Resolve Conflict in CollaborativeModel-Building Approaches

Laura J. Black1* and David F. Andersen21Montana State University, College of Business, PO Box 173040, Bozeman, MT, USA2Rockefeller College of Public Affairs and Policy, University at Albany-State University of New York, Albany,NY, USA

In the context of facilitated, technology-supported efforts to resolve complex problems, werecognize the critical role that visual representations can play in both the content and processof collaboration. How these representations are wielded by facilitators and interpreted byparticipants determines whether they help resolve conflicts or close down conversations.We identify three key attributes of scripted problem-solving facilitation, as well as threekey attributes of visual representations that function as boundary objects, to gain insightsinto pivotal experiences when group problem-solving efforts turned from collaboration toconflict and vice versa. We draw on three vignettes from facilitated group problem solvingto illustrate how these attributes can be deployed to move conflict-mired conversations intocollaborative discussions. This paper contributes to collaborative problem solving by usingthe formal sociological theory of boundary objects to offer a deeper, richer understanding ofsuccesses and shortcomings of visual representations as drivers of conflict resolution inmodel-building approaches. Copyright © 2012 John Wiley & Sons, Ltd.

Keywords boundary objects; collaboration; group model building; problem solving; systemdynamics

INTRODUCTION

Group model building is an effective approachto collaborative problem solving (Vennix, 1996;Vennix et al., 1996; Andersen and Richardson,

1997; Andersen et al., 1997; Vennix, 1999). It inter-actively involves stakeholders in problem identifi-cation and problem-solving processes using themethods of system dynamics modelling (Randers,1980; Wolstenholme, 1994), from behavior-over-time graphs through model conceptualizationand simulation analyses. The approach is particu-larly desirable when addressing ‘messy problems’(Ackoff, 1979; Vennix et al., 1996; Vennix, 1999)

*Correspondence to: Laura Black, Montana State University, Collegeof Business, PO Box 173040, Bozeman, MT 59715, USA.E-mail: [email protected]

Copyright © 2012 John Wiley & Sons, Ltd.

Systems Research and Behavioral ScienceSyst. Res. 29, 194–208 (2012)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/sres.2106

or ‘wicked problems’ (Rittel and Webber, 1973),when participants’ resources and objectives maydepend on or constrain other participants’resources, actions and objectives. Vennix (1996)has advocated that group model building canhelp people, quite literally, ‘get on the samepage’ in understanding the common groundand opportunities among them, and Andersenand Richardson (1997) have reported that theprocess has generated collective insights aboutwhy problems persist despite organizations’ bestefforts to address them decisively.Group model-building sessions rely on a

facilitator (Richardson and Andersen, 1995) tomove participants through a structured processof problem articulation and problem-solvingexplorations. The process includes both low-tech,high-touch aspects, such as capturing participants’contributions on flip charts or a white board orother large, visible writing space, and technologic-ally based representations that synthesize partici-pants’ input in sketches, maps and simulatingmodels.This paper focuses particularly on how visual

representations, both low-tech and high-tech,function in group model-building collaborativeefforts. We assert that when visual representa-tions function as ‘boundary objects’ they turnsituations of disagreement, tension and conflictinto collaborative problem-solving discussions.A construct from sociology, a boundary object(Star and Griesemer, 1989; Henderson, 1991) isa tangible representation of dependencies acrossdisciplinary, organizational, social or culturallines that all participants can modify. It caneffectively advance shared understanding whenparticipants can transform the representation toshow more clearly their understanding of thedependencies among them and the implicationsfor each participant’s resources, operations andgoals (Carlile, 2002).In the following, we describe three collab-

orative problem-solving efforts in which thegroups were struggling with conflict of one sortor another. Two of the vignettes have beenselected from our own repertoire of facilitationand modelling experiences and one is fromthe direct experience of a close professionalcolleague; each illustrates particular instances of

visual representations as part of facilitatedmodel-building efforts. In the section followingthe vignettes, we examine the theory and practicesof structured, planned activities of facilitation andunderscore how conflict and unintended resultscan emerge from scripted behaviors. Next, wesummarize the theory of boundary objects andexplain generally the conditions under whichvisual representations can serve as boundaryobjects or, alternatively, as conscripting or blud-geoning tools. We examine the role that visualrepresentations played in each of the threevignettes in light of these principles and interpretthe planned and unplanned dynamics emergingin each situation according to how visual represen-tations were used. Finally, we present an analyticdiscussion that presents in summary form a visionof why and how boundary objects can supportreaching agreement by conflictive groups andoffer concluding thoughts.

METHODS USED IN THIS PAPER

This study emerged from a combination ofaction research (Schein, 1969; Argyris, 1993;Eden and Huxham, 2006) and case study review(Eisenhardt, 1989). Collectively, the authors ofthis paper have extensive experience workingwith facilitated groups, with many of thoseyears involving computer-assisted face-to-facegroup modelling sessions. We each associatedwith several sets of long-term and experiencedmodeller–facilitator pairings, with these pair-ings having cumulative decades of experiencein this kind of work. Although most of ourefforts have been with collaborating groups, animportant minority of our work has of necessityinvolved groups experiencing conflict of oneform or another. Indeed, some of our most vividexperiences—opportunities to learn and grow inmodelling and facilitation practices—arise fromconflictive groups. Both authors also havesignificant experience conducting research ongroup model building and facilitated projects,motivated by a belief that systematic reflectionon and study of group processes can lead toimprovement in these practices for ourselvesand others.

Syst. Res. RESEARCH PAPER

Copyright © 2012 John Wiley & Sons, Ltd. Syst. Res. 29,194–208 (2012)DOI: 10.1002/sres

Using Visual Representations as Boundary Objects 195

THREE SCENARIOS OF CONFLICTS ANDCOLLABORATIONS

Vignette #1: Strategies for Public Security in aRegion of Ethnic Violence

Colleagues of ours were working with seniorpublic security offices in a region that had recentlyexperienced increased ethnic violence and terror-ism. Ethnic violence had been a long-standing partof the region’s history. Thismodelling and strategydevelopment conference was attended by seniorofficers of the police and security forces whoshared interests in keeping all the public safe,preventing bomb explosions in public places andeliminating assassinations. But the ethnic groupsthat these police and security forces representedhad a history of suspicion and mutual recrimina-tions, if not outright hate. Cooperating with oreven agreeing with people perceived as ‘the otherside’ risked being seen as small acts of ethnictreason.

An early activity of this strategy mapping andmodelling session was the creation of an issuemap. The group task was simple—a script askingindividuals to write down one issue surroundingpublic safety in the region and then paste it on apublic space. They were to do this exercise overand over until all relevant issues were visiblefor everyone in the group to review and discuss.This activity is usually a fast way to surface andshare a robust mixture of problems, goals,assumptions and even hints for solutions. Later,these issue maps could be sorted into agreed-upon statements of common goals linked toaction steps and assumptions that would supporta map of common goals.

The special problem for this particular sessionstemmed from the long history of conflict amongmembers of the group. It was implicitly the casethat, ‘If they asserted or proposed something,then we will have to oppose it.’ On that day,therefore, facilitators arranged for the issue mapto be constructed using Group Explorer, a net-worked version of the Decision Explorer software.The room was equipped with a network of abouta dozen personal computers with pairs or smallgroups of security officers (from the same ethnicgroup) working at each machine. When they

entered an issue at their keyboard, it appeared ona common projected screen. Whereas the groupfacilitator and modeller could identify who hadsaid what, participants could not. Issues and ideaswere detached from their authors and could betaken (or rejected) only on their own merit.

After the initial brainstorming of issues, two ormore ideas could be clustered to craft a newthought. Members of the group could argue overa particular idea or over a linkage between twoor more issues without knowing to whom theoriginal ideas and links belonged. Security officerscould amend issue statements and craft newviews, all the while being ignorant of the ethnicorigins of the original ideas. This group eventuallycrafted a common statement of security goals forthe region and agreed upon a set of policy actions,tied to key strategic assumptions and resourcerequirements to support the proposed policyactions.

Vignette #2: Conflicting Views of PricingPolicies in a Commodity Market

A member of the authoring team was engaged ina system dynamics group model-building effortwith the executive team of a major multinationalfirm that operates capital-intensive plants in acommodity market. At the time of the modellingexercise, the market was experiencing a relativeover-expansion of capital infrastructure world-wide, and profit margins for the industry werefalling precipitously from pressures due to thetrough of a commodity cycle. This corporation,a technology leader in its field, had in the pastbeen able to differentiate its product from themainstream commodity market, but capitalplants around the world were increasingly ableto replicate innovations, and traditional pricepremiums enjoyed by the client were eroding.

The system dynamics group modelling projectwas sponsored by a corporate vice president withresponsibilities in areas of innovation and researchand development, but the CEO of the firm wasthe ‘meeting owner’, and in addition to other execu-tives, a number of senior operations managersresponsible for global operations and sales wereparticipants in the room. The groupmodel-building

RESEARCH PAPER Syst. Res.


196 Laura J. Black and David F. Andersen

session had begun with a concept model based onpublished simulations of commodity markets thatwas more elaborated than most initial conceptmodels (see Richardson, 2006 and Scriptapedia;Hovmand et al., 2011 for a description of conceptmodels). The next stage of the facilitated processinvolved the group’s elaborating the model’sinitial feedback structure to accommodateparticular details of the firm’s pricing formationas it interacted with the industry as a whole. Aparticularly thorny issue centred on mappinghow existing commodity market mechanismsinteracted with industry-wide price-reportinggroups and the sales policies of individualcompanies to create an informal, publicly visibleand commonly agreed upon price marker forvarious well-defined segments of the commod-ity’s international market.Sales, operations and finance all had slightly

different views of how the firm’s internal pricesetting interacted with the industry’s formalprice reporting. There was disagreement withinthe senior management team over how thesepricing interactions dynamically played out andwhat direction the firm should take in its internalpricing policies. The stakes were high because, inthese over-capitalized markets, small differencesin price margins could mean the differencebetween survival and having to close one ormore production facilities.The CEO, who had been following the vigorous

discussion and arguments with interest, finallyweighed in with his point of view. He had aforceful, hard-hitting, no-nonsense style of presen-tation, and he laid out what he believed the firmshould do next in clear and compelling terms.His intervention seemed to settle the emergentconflict in the room. The facilitator called a breakin the modelling process.During the break, a member of the modelling

and facilitation team took the CEO aside toreview the morning and to comment on hisdramatic intervention. The facilitator verbalizeda tricky observation, ‘Do you think that anyonein the room could disagree with you, given theway you just made your presentation?’ Thefacilitator then asked him to re-start the meetingafter the break, giving the group a sense of whatdirection he wanted to see.

After the break, the CEOopened themeeting andin less forceful terms re-stated his position aboutwhat action the firm should take. But he wantedto hear further debate and discussion. In theensuing silence, the facilitator directed the group’sattention to the pricing structure that they had beenmapping in the model just before the break. Thedisagreements that had been in the room slowlyre-emerged, but they were now focused on creatingand discussing details of the map on the white-board. Individual words, phrases and causalarrows were all called into question and discussed.

During these detailed discussions, the facilitatorkept the participants’ focus on the emergent mapon the white board rather than on the personalitiesvoicing comments, and gradually, a new view ofcollective corporate strategy emerged. In the end,the view that emerged on thewhite board ‘shouteddown’ in its own quiet way the CEO’s position sovividly staked out before the break. He waspleased to have his initial thoughts replaced by amore empirically grounded, commonly supportedand potentially more profitable idea. Futuremodelling efforts explored these collective ideasinmore detail with wide/growing support amongthe executives and operations team.

Vignette #3: Inter-agency Social Policy Reformat the County Level

Our third vignette has much in common withthe second, but with a less fortunate conclusion.The client group was an inter-agency team ofpublic sector managers from the United Statesassigned to work together on a county-wide setof cross-agency reforms in a specific area of so-cial policy. The participants in the group modelbuilding consisted of managers from severalcounty, state and municipal agencies as well asrepresentatives of key non-governmental orga-nizations who shared an interest in this particu-lar policy area. The commissioner from one ofthe lead agencies was the meeting owner andhad convened the entire group and givenparticipants their charge to articulate a strategyfor making cross-agency reforms.

As in the second vignette, facilitators beganby presenting a basic concept model, quickly




transitioning to causal mapping of system struc-ture. At almost every step along this process,disagreement erupted between members of thestrategy planning team. Individual members werearticulating and adhering to the points of view oftheir own organizations, and most parts of thecausal map emerged slowly from this vigorousdiscussion.

By noon of the second day, the scheduledadjournment for the meeting was within sight.Although progress had been made on manyareas, the group had not reached consensus onseveral key points. It seemed clear to the facilita-tors that the session would reach its scheduledending time with several of the meeting’s keypoints still unsettled. During the noon break,facilitators consulted with the meeting owner.Up to this point, he had been actively engagedin both trying to build consensus while tryingto advocate for his own agency’s perspective.

After the break, the commissioner/meetingowner moved into a decisive leadership role.Because most of the disagreement had centredon details of the causal map being constructedby the group, he decided to sidestep these contro-versial images. Rather, he identified approxi-mately eight key tasks that had emerged ascandidate ‘next steps’ during the discussion thatmorning. In an effort to create momentum for theoverall project, for each task, he identified a leadstaff person, almost all from his own agency, andthen assigned one or more of the other meetingmembers to the task. Sub-groups were expectedto report on their progress at the next meeting.

The next meeting never happened. The cross-agency teams assigned by the meeting owner topush forward on an assigned task simply did nottake up the work. Attempts to re-schedule a nextmeeting failed. The social policy strategy forreform under discussion was never implemented.

SCRIPTED FACILITATED PROBLEM-SOLVINGPROCESSES AND THEIRVISUAL ‘PRODUCTS’AS POTENTIAL BOUNDARY OBJECTS

In this section, we examine previous work on facili-tated problem solving using ‘scripts’ and researchon visual representations as boundary objects.

Looking at the Vignettes’ Scripted Processes forTechnology-Supported Problem Solving

We identify three key attributes of facilitatedtechnology-mediated problem solving:

• The process of the conversation and the contentof the conversation are separated by designand by different individuals playing differentroles (facilitators manage the process; partici-pants manage the content).

• An implicit contract of trust between facilitatorsand participants exists.

• Explicit sub-processes—sometimes calledscripts—focus clearly on creating products ordeliverables that are visually shared anddiscussed.

Facilitated, technology-mediated problemsolving requires the use of some technology(not necessarily sophisticated technologies) anda modicum of participants’ trust in facilitatorsand in the processes and activities they propose.In each of the three vignettes described, thefacilitation and modelling team was followinga plan, a scripted plan of action for the session.These plans for sub-processes exist independentof the content the group wants to address.Almost all the scripted plans we know ofpresume that the client group is a collaboratingteam, or at least working on a shared problemor seeking to attain a shared goal. Scriptedsub-processes also assume that the purpose ofthe facilitated modelling session is to create aconsensus or agreement—an ‘alignment ofmental models’ (Kim, 2009) as it were—throughprocesses of exploring different perspectives onparts of a complex problem.

As evident in the aforementioned vignettes,however, scripted activities designed to helpgroups collaborate and create consensus can alsoprove useful in mitigating conflicts. We believethat the capacity to synthesize conflicting viewsand mitigate personal tensions relies largely onthe boundary objects that are evoked and putinto play in the facilitated modelling sessions.Before we describe what boundary objects areand why they can be useful both for creatingconsensus and for resolving conflict, we pause toexamine literature on and approaches to scripted




design of facilitated face-to-face group modellingsessions. This look at the published literature onscripted approaches to group model buildingoffers some additional insights into what wasgoing on in the three vignettes as well as intowhy two of the vignettes seem to have had rela-tively good outcomes whereas the third did not.In the first vignette, the modelling team began

by building an issue map. With the use ofGroupExplorer (Ackermann and Eden, 2001),an issue map for nearly 100 concepts can be con-structed by a group in less than 1 h (Ackermannet al., 2011). The detailed procedures for how tobuild these maps are described in Ackermannet al. (2005), which also provides a detailed processdescription, or script, for creating a goal map andan agreed-upon set of policy actions while work-ing with a group. More recently, Ackermannet al. (2011) addressed a number of questionsleft open when one adopts a scripted approachto developing group model-building sessions.Should some scripts be performed first, whereasothers wait until later? Are some scripts properlyseen as prerequisites for others? In general, whatguidance, if any, exists for practitioners who wishto assemble a series of scripts into a wholeintervention plan that makes sense? ‘ScriptsMap’(Ackermann et al., 2011) is a tool for addressingthese questions, offering a network framework ofsequences of scripted activities, products anddeliverables to enable facilitators to constructappropriate combinations for workshops. Theinitial work laid out a map that combines scriptsfrom traditional group model-building practiceswith Eden and Ackermann’s (1998) approachto strategy development working directly withclient groups. Eden et al. (2009) further elab-orated on a number of practical and more theo-retical dilemmas associated with attempts tointegrate group modelling projects using diverseanalytic methods.In the second and third vignettes, the facilita-

tion and modelling team was using scripts inmuch the same sequence as prescribed by theScriptsMap (Ackermann et al., 2011). Both ofthese vignettes began with a ‘graphs-over-time’script, designed to help a group identify a set ofvariables to include in the simulation model.The next script used was a ‘concept model’ script,

designed to teach the basic vocabulary andiconography of system dynamics modelling in25min or less through a demonstration of themodelling using a simplified version of thepresenting problem or objective. The groupmodel-building sessions giving rise to the secondand third vignettes were using a series of scriptsdesigned to elicit group consensus on causesand interactions among causes of the problemwhen strong disagreements emerged. Thesescripts include a ‘ratio script’ and a scriptdesigned for ‘elaborating model structure’. Fi-nally, both meetings used a script for resolvingconflict by facilitators talking with the meetingowner during a break. Of course, these twomeetings ended with different outcomes, andthe nature and effectiveness of participants’subsequent activities differed dramatically.

Each of the scripts has been more carefullydescribed by Hovmand et al. (2011) in a catalogueof scripted behaviors called the ‘Scriptapedia’.Each script begins with some ‘products’, or mate-rials produced by earlier discussions, that areavailable at the beginning of the script anddescribes the group process used to transformthese products into artefacts that represent thecontent of the group’s discussion.

Looking at Scripts’ Visual ‘Products’ as PotentialBoundary Objects

Many of the products of group model-buildingprocesses are shared visual representationsthat portray participants’ perspectives on thecomplex problem that brought them together.Visual images provide a significant input to thethinking process, and they can play a critical rolein helping people re-conceptualize abstractproblems (McKenzie and Winkelen, 2011). Butthe visual representations used in collaborativeproblem solving and group model buildingplay an especially significant role in facilitatingand shaping consensus when they function asboundary objects (LJ Black, 2011, working paper).Specifically, visual products of facilitated pro-cesses can serve as boundary objects when theyhave the following characteristics:




• They are tangible two-dimensional or three-dimensional shared representations (considera diagram, sketch, prototype model, map,etc.—these representations can include text but,in our experience, the text is always sparse).

• They portray salient dependencies and relation-ships among participants’ objectives, expertise,decisions and actions.

• Critically, they can be modified by input fromevery participant.

Here, we briefly describe the theoreticalbackground related to boundary objects, whichwe then use to interpret how the visual represen-tations functioned in each of the vignettesdescribed in the preceding sections.

Lave’s (1988) theory of distributed cognitionasserted that cognition is not ‘all in our minds’but distributed among our minds and bodies andthe locations in which we exercise our compe-tences. Although using visual representations incollaborating groups is not new, distributedcognition provides a strong theoretical basis forgiving special focus to them in facilitated technol-ogy-supported problem-solving sessions. Thevisible ‘products’ of scripts provide tangiblerepresentations of how participants conceptualizethe problem and therefore how solutions mightbe identified. In the group model-building setting,the representations customary to the systemdynamics method offer opportunities to representa familiar problem in ways that are different from‘just talking’ about the problem. These visualrepresentations especially focus on showingrelationships among the problematic aspects thatparticipants care about. If we take seriously theconcept of distributed cognition, the emergingmaps and models of a group model-buildingsession offer content-rich, socially shared experi-ences that help participants recognize and changetheir individual and collective cognition, or howthey think about the problem.

Shared visual representations do not facilitate col-laboration, however, unless they function as bound-ary objects. Of course, boundaries can arise fromdifferences in locations, organizational structuresor entities, knowledge disciplines, expertise levelor time frames for decision making because any ofthese gaps can create challenges in coordinating

and collaborating on interdependent resourcesand activities. The term ‘boundary object’was firstcoined by Star and Griesemer (1989), who definedit as an object ‘adaptable’ enough to be interpreteddifferently by people whose expertise differedwithout losing a coherent identity across the socialworlds it is spanning. Star and Griesemer (1989)suggested that boundary objects are usefulbecause they aid in negotiations when ‘each socialworld has partial jurisdiction over the resourcesrepresented by th[e] object’ (p. 412), but those juris-dictions overlap or are mismatched or in conflict.

Henderson (1991, 1998) elaborated on thenotion of ‘adaptable’ objects, emphasizing theimportant role that many kinds of artefacts, toolsand technologies play in the coordination ofknowledge and competence, when one kind ofexpertise is insufficient to accomplish the task athand. She studied the ways that drawings andsketches of products under development elicitedvoluntary innovative ideas or ‘conscripted’ partici-pation in aspects of the products under develop-ment. Carlile (2002) also studied artefacts used torepresent dependencies across functional bound-aries in the context of product development andarticulated three aspects characterizing a robustboundary object: representative of dependenciesamong the actors, relatively ‘concrete’, given theparticipants’ differing expertise and experienceand transformable by all actors. If an objectis transformable, then anyone involved has theability to manipulate and alter the representationto show more clearly the consequences of thedependencies she/he perceives.

INTERPRETING THE VIGNETTES IN LIGHTOF SCRIPTS AND BOUNDARY OBJECTS

Although neither boundary objects nor scriptshave redistributing power as a primary objective,both boundary objects and scripts have significantimplications for the use of power among par-ticipants during a collaborative problem-solvingeffort (Black, 2011, working paper). When visualrepresentations function as boundary objects, theyde-personalize issues in a way that spoken wordsdo not and focus the conversation on dependenciesand their consequences for participants’ objectives




and operational choices. The design of technology-supported facilitated problem-solving activitiesdescribed in the previous section is premised oncollaborative use of visual representations thatsuccessively build on (and sometimes challenge)previous agreements by participants. As men-tioned in the previous section, facilitation andtechnology use enable a temporary rebalancingof power among participants as they relinquishto facilitators responsibility for managing theprocess and for making productive use of thecontent generated by participants. How the groupproblem-solving session unfolds, however, affectswhether trust in technology-supported processand in facilitators managing the process grows ordiminishes. If we use the lens of boundary objectsto examine how the sessions unfolded in each ofthe aforementioned vignettes, we gain usefulinsights about the role that visual representationscan play in the dynamics of collaboration.

Vignette #1 Re-interpreted: Strategies for PublicSecurity in a Region of Ethnic Violence

Conflict was part of the context at the outset of thegroup process in the first vignette. Conflict lay inthe individual and community histories of partici-pants, and violent conflict gave rise to a sharedgoal for public safety, the purpose in meeting.When participants were first asked to create anissue map by generating ideas and portrayingthem on a shared space, what in other settingscan be a companionable brainstorming sessionhere raised threats to participants’ identities andloyalties. Their conflicts prevented them frombuilding on one another’s ideas because each ideawas observed to be associated with the personwho shared it. Notably, collaborative model build-ing seldom employs anonymity because its goal isto increase participants’ capacity to talk to oneanother productively about how to use theirmutu-ally dependent resources and actions to achievetheir individual and shared objectives; maskingthe identity of participants’ comments customarilyundermines that goal. But in the public safetyconversation, by altering the group process to usethe GroupExplorer method (Ackermann and Eden,2001) in which participants’ comments were

portrayed anonymously, the facilitators createda de-personalized, shared representation thatpermitted participants to suspend suspicions andbegin to think together. The tool allowed partici-pants to transform the shared representation ofindividual comments and build on one another ’sissues and ideas. It therefore functioned as aboundary object because it facilitated partici-pants’ transforming a shared representation,which helped them navigate in less threateningand risky ways to a shared understanding ofwhat they could individually and collectivelydo to improve public safety.

Vignette #2 Re-interpreted: Conflicting View ofPricing Policies in a Commodity Market

In the case of the multinational commodity-producing firm, participants shared a goal toagree on a pricing strategy that would help thefirm survive the current price pressures in theindustry. Interpersonal conflict among the topmanagers may have been latent but was not ex-plicit at the outset; during the session, however,as facilitators captured in a diagram participants’experiences and views on the problem, conflictemerged along clear lines of power asymmetryas the CEO offered his own clearly articulated,no-nonsense view of how the group should ad-dress the pricing problem. Following the breakin the meeting, when participants reconvened,the facilitators directed attention from the CEOand his words back to the shared representationof the aspects of firm and industry interrelation-ships that reflected their emerging collectiveunderstanding of pricing, sales and operations.

This representation, which depicted dependen-cies among what participants knew from theirown experiences, was gradually transformed withthe facilitators’ help during the subsequent con-versation. By directing attention to the shared vis-ual representation, the facilitators de-personalizedthe conflict because participants could discuss thediagram and suggest changes to it without directlychallenging the ranking executive. By solicitingsuggestions from all the participants and modi-fying the diagram to reflect what was said, thefacilitators also made the power asymmetry in




the room, although still present, less relevant.Attention to the visual representation also permit-ted participants to build on and transform therepresentation of their integrated understanding.Through this tool, possible policies suggested byparticipants were represented in modifications tothe visual representation, enabling the group tothink together about the implications of thedependencies. Because the facilitators wieldedthe representation as a boundary object, gradualconsensus on a policy emerged from the discus-sion through reasons and hypotheses shared bythe group, not because of compliance with theopinion of the highest-ranking participant.

Vignette #3 Re-interpreted: Inter-agency SocialPolicy Reform at the County Level

Conflict—in this case manifested passively as theabsence of joint action (Blumer, 1969)—was partof the context in individuals’ and agencies’experiences of the persevering problem necessi-tating cross-agency reforms in a specific area ofsocial policy. During the session, with the facilita-tors’ help, some substantive interrelationshipswere identified in a shared representation, butthe slow process also served to re-entrench partici-pants along agency lines as they expressed primar-ily their sponsoring organizations’ points of viewon the complex issue. When the commissioner,the highest-ranking participant, rushed to closureby disregarding the map of interrelated causesand effects, he unilaterally pushed aside the sharedvisual representation. Why didn’t the commis-sioner’s list of eight action items serve as a bound-ary object?He put forth a newvisual representationnot generated by participants’ conversation andconsensus. The commissioner’s action item list didnot show dependencies among agencies’ actionsand was not transformable by anyone in the room(including—or maybe especially—by the facilita-tors). Removing the opportunity for participantsto work with a visual representation that couldfunction as a boundary object served to end theconversation in that session so definitively that,in the months following, the conversation wasnot re-opened by participants.

ANALYSIS: HOW BOUNDARY OBJECTSPROMOTE AGREEMENT IN CONFLICTIVEGROUP MODEL-BUILDING SESSIONS

We view the emergent process of helping partici-pants move toward joint action as multiple stagesof building incremental agreements, many ofwhich are represented visually during thecollaborative problem-solving process. Repre-senting participants’ input visually, identifyingdependencies among participants’ perspectivesand depicting those visually and transformingthe accumulating understanding by altering thevisual representation place a continually revised“picture” of incremental agreements before parti-cipants. Even if participants do not agree withone another ’s perspectives, they agree that theirperspectives are different and yet share somedependencies—that itself comprises a series ofagreements. This shared focus, sustained throughfacilitated discussions creating and modifyingvisual reminders of a common ground, simultan-eously builds participants’ trust in one anotherand in the process generating the emergentagreements. Figure 1 indicates that each stage ofa four-stage, facilitated problem-solving processplays a role in reinforcing participants’ mutualtrust, which in turn enables their greater partici-pation. Critically, the visual representations asboundary objects serve as the group’s point offocus and stimulate the activities that create thegradual accumulation of agreements.

We identify four stages of activities, eachgenerating some accumulation of visually repre-sented ideas: (1) Participants generate ideas thataccumulate into a stock of salient ideas ‘on thetable’ for the group to consider; (2) Participantsidentify dependencies among some of the salientideas and perspectives, creating an accumulationof ideas that also portray consequences of inter-dependencies among them; (3) Participants trans-form a portion of the salient ideas by someprocess of group discussion or idea manipulationto create another stock of tangible and salientideas modified by their shared input; and (4)Working with these transformed ideas and someprocess of prioritization, participants createagreements about how the group can moveforward. Members’ trust in one another and in the




process that advances participants through thesestages plays a critical mediating role amongactivities and accumulations of facilitated scriptproducts. The visible script products, wielded asboundary objects, provide early and growingevidence that participants are being heard byfacilitators and by one another. This evidencebuilds trust and at least a limited sense ofpsychological safety that enables continuinglearning behaviors (Carmeli et al., 2009) throughthe facilitated activities, which in turn createsmore evidence that reinforces participants’ mu-tual trust and leads them to participate furtherin the collaborative problem-solving processes.In our experiences, we find that structured

processes for interaction and facilitation enablethe activities that generate and transform ideas,portray them visually and then selectively shapesome transformed ideas into paths for jointaction. It is straightforward to see that the keyattributes of boundary objects—tangibility, repre-sentation of dependencies among participants’objectives and resources and transformability—ease the emergent processes. Idea generation issupported by boundary objects that are relativelyconcrete, given participants’ expertise and ex-perience, and able to depict salient dependenciesamong members of the group. They can ‘seewhat they say’ and the consequences of their

interdependencies. Because representations aretangible and show salient dependencies, visualrepresentations that function as boundary objectssupport the ability of the group to transforminitial ideas placed ‘on the table’ into alternativesthat retain their salience and tangibility evenas they are modified to represent additionaldependencies and consequences to participants.

Facilitating collaborative problem solving doesnot solely depend on visual representations.Clearly, we believe that facilitators’ using struc-tured scripts also accelerates the process, and fur-ther, we believe that participants must begin withat least a modicum of trust in both the facilitatorsand the processes they bring. Especially withtechnology-mediated processes, facilitators drawon a ready source of ways to represent ideas.Facilitators make sure that ideas are representedvisually, often unselectively and democratically,because they bear responsibility for the process,not the content. But by ensuring that these repre-sentations show dependencies and are modifi-able and modified by ongoing participant input(again, often supported by technologies andusing structured scripts), they can ensure thatthe representations serve as boundary objects.

Critically, not every visual representationfunctions as a boundary object. Let us considereach of the vignettes in light of Figure 1.

members'mutual trust

tangiblyrepresented

ideas

tangiblyrepresentedtransformed

ideas

tangiblyrepresented

ideas agreed tofor moving

forwardgeneratingideas

selectingideas

visualrepresentation

tangibility

visual representationdependency-portrayal

visualrepresentationtransformability

we areall

heard

we buildnew sharedunderstand-

ingswe are

all in thistogether

butmaybewe arenot allstuck

tangiblyrepresented

ideas showingdependencies

identifyingdependencies in

ideas

transformingideas

we agreewe canmove

forward

ourprogress

fuelsworkingtogether

we canbe clearabout

how weare

affected

Figure 1 Properties of boundary objects enabling activities that build trust and agreements incrementally




• In the first vignette, participants initially pro-gressed no farther than generating ideas, thefirst activity in Figure 1, with written wordsrepresenting their tangible ideas. Becauseoriginally their written ideas for improvingpublic safety were observed to be connectedwith the participants’ themselves, all werereluctant to identify dependencies among theideas because that could imply there weredependencies—perhaps in contradiction toconventional lines of community conflicts—among themselves as well. When the facilita-tors introduced anonymity into the process ofgenerating ideas, then the visually representedideas could function as boundary objectsbecause participants then trusted the processsufficiently to identify dependencies amongthe ideas, building the second stock in Figure 1.

• In the second vignette, the executives andoperations managers were progressing throughstages of problem solving. They built accumu-lations of tangibly represented ideas withportrayed interdependencies related to pricingpressures and transformed them with facili-tated help. When the CEO expressed his no-nonsense opinion about how to move forward,he effectively voiced his desire to ‘jump stages’directly to the last accumulation of agreementsin Figure 1, rupturing the process of graduallybuilding agreements through transformedvisual representations. After the session break,the facilitators deliberately returned partici-pants’ focus to the visual representation thatserved as the most recent picture of their agree-ments thus far. By allowing the discussion tofocus on transforming that representation oftheir ideas, participants—including the CEO—returned to the process of gradually buildingagreements and eventually arrived together atagreements on how to move forward.

• In the third vignette, agency representativeswere slowly building accumulations of tangiblyrepresented ideas, portraying some dependen-cies and encountering some difficulties in trans-forming them, as they returned to voicing theirsponsoring agencies’ views rather than thinkingabout the image of interdependencies in thecontext of the problem before them. When thecommissioner took the pen from the facilitator’s

hand and began writing an action-item listbefore the participants, he not only rupturedthe process of building gradual agreements, healso expressed disregard for the process itselfand proposed his own view of how to moveforward. His action list was a visual representa-tion, but it did not function as a boundary objectin any respect. Rather, his un-transformablerepresentation was an expression of power,with which he hoped to persuade or coerceparticipants to his view of appropriate action.

When faced with a room full of people atthe beginning of a structured problem-solvingsession, facilitators must move quickly to buildvisual representations of small, incrementalagreements because the modicum of trust initiallyextended to facilitators and their process will beextended for a limited time. They thereforerecognize the need to build participants’ trustconcomitant with the accumulations of tangiblyrepresented ideas and interdependencies—andespecially so in situations of conflict. With repre-sentations that function as boundary objects, thisis possible. Richardson and Andersen (2010)reported using a form of computer-supportedboundary object to create an issue map with morethan 100 salient and linked concepts in less than40min. Howick et al. (2006) reported that afacilitated group of collaborating managers wasable to start recombining and transforming ideasin less than 2 h.

The ability to rapidly fill a stock of participant-legitimated, transformed ideas using boundaryobjects and scripted group model-building tech-niques has two additional downstream impactson future agreements. First, in the absence ofinitial trust among group members, the rapid ac-cumulation of tangible, transformed ideas, repre-senting legitimate compromises or new ideas,can help build trust among participants. Theysee that the process has the potential to surfacealternatives or compromises that could benefittheir position, and they may work with oppo-nents to reach a possible beneficial solution fromtheir own point of view. Second, if the formalleadership of the group, that is, the ranking man-ager or officer participating, is open to letting thisprocess work, then participants’ experiencing




early agreements can rapidly build confidencenot only in the facilitators and in the processbut also in the leader-participant and in her orhis ability and willingness to support comprom-ise agreements. McCartt and Rohrbaugh (1989)have identified managerial openness as a key pre-dictor of success in a large-scale retrospective studyof facilitated face-to-face computer-supportedmeet-ings generally. We propose that for groups inconflict this managerial openness is especiallyimportant. We saw in vignettes 2 and 3 that theleadership’s ability to support the facilitated,structured process became a key factor affectingthe success or failure in conflict-mired groups.Our high-level view of how boundary objects

promote agreement in conflictive groups is this:Boundary objects’ tangibility and dependencyrepresentation, coupled with free expressionand sometimes an intervening variable of ano-nymity, promote a rapid ability to get salientideas ‘on the table’. Just doing this builds sometrust early, as members of a low-trust group seetheir own and others’ ideas rapidly laid out in alegitimate and transparent fashion. Facilitators’transforming representations from participant in-put, especially when skilfully used with scriptedprocesses, almost always assures that a few com-promises or new ideas can emerge. If the rankingleader-participant demonstrates openness, thenboth trust among members and trust in thefacilitated process can grow rapidly. With theseconditions in place, the stock of transformedideas can grow even further, creating opportun-ities for identifying agreements on how to moveforward. If this process goes well—and for rea-sons as demonstrated in the vignettes, it mightnot—the reinforcing processes described inFigure 1 dominate in favourable ways. First, aflow of agreeing behaviors leads to a stock ofagreements, which increases members’ mutualtrust, thereby enabling a larger flow of agreeingbehaviors. Similarly, the same initial flow ofagreeing behaviors leading to accumulatingagreements can further increase the dynamicability of an open leader-participant to brokerfurther agreeing behaviors.These reinforcing loops do not always operate

in a benign fashion, however. They can also serveto trap a conflictive group in failure. Imagine that

for some reason an external event in the roomcauses members’ trust to decrease—some off-hand remark or reaction to an external event suchas a news broadcast. Then this decrease in trustmakes it harder to evoke agreeing behaviors,leading to a lack of accumulating agreements,making it even harder to build future trust, orworse yet, leading to an actual erosion of trust.A similar vicious cycle working in the ‘wrong’direction can occur with an initial drop in trustthat the ranking leader-participant is open toideas, perhaps a sour opening remark or off-handquip, and lead to a similar trap due to partici-pants’ perceiving that they cannot effectivelytransform others’ ideas. Furthermore, whethergrowing momentum of agreeing behaviors leadsto substantive, potentially innovative solutionsor deteriorates into group-think rests largely onboth the scripted processes for systematicallyprobing a problem space and on the facilitators’abilities to ensure that salient representationsdepict dependencies, the consequences of thosedependencies and all participants’ ongoingcandid input.

CONCLUSIONS AND IMPLICATIONS

In this paper, we have argued that the presence ofwell-functioning boundary objects can and doesdrive conflict resolution in collaborative groupmodel-building projects. Recognizing and know-ing how to use boundary objects wisely is a keyskill for resolving conflict in groups of partici-pants who are working on a shared problem orissue. Although the notion of boundary objectshas been evoked as an organizing frame forgroup model building (Zagonel, 2002), this paperfocuses specifically on the theory of boundaryobjects as induced and formalized in sociologicalstudies, which refers specifically to transformablevisual representations. This paper contributes toresearch on collaborative problem solving bydemonstrating how the formal theory of bound-ary objects offers a deeper, richer understandingof successes and shortcomings of visual represen-tations as conflict-resolution tools in model-building approaches. The benefits of developinga theoretical understanding of our practices lie




in helping us recognize the salient elements inmany varied complex situations that can serve aspoints of leverage to accomplish desired out-comes—in this case, making sure visual representa-tions are used as transformable boundary objects tocreate and sustain collaborative discussion.

We first presented three vignettes illustratingboth how conflict can arise and be resolved ingroup model-building sessions. After presentingthree key characteristics of structured groupmodel building as well as three additional char-acteristics of boundary objects, we used theseideal-type characterizations to re-interpret thedynamics of conflict generation and resolutionin the three vignettes. We concluded by present-ing a four-stage conceptual model of how tech-nology-supported, facilitated problem-solvingsessions create tangibly represented agreementsfor moving forward. Group members must firstbe able to tangibly represent ideas from theirown perspective. Represented ideas must bedepicted in ways that identify dependencies,and these dependencies can often depict conflictamong members of the group. Critically, allmembers of the group must be able to transformthe ideas that are under consideration. Only thencan concretely represented and transformedideas be converted into agreements for movingforward.

We conclude with three broad ‘take-away’mes-sages to inform collaborative problem-solvingpractices:

1. Make sure that visual representations areactually boundary objects. Not all goodvisualizations are good boundary objects. Aswe saw in the third vignette, the creation of alisting of ‘tasks to do’, often an artefact usedto galvanize alignment around collective nextsteps, can become a polarizing bludgeonwhen not used properly. Boundary objectsare artefacts whose key properties areactivated by the social processes defininghow they are used. Good artefacts manipu-lated with bad social processes do not makeboundary objects.

2. Use boundary objects to support ideageneration and transformation, especiallyto map dependencies. The attributes of

well-functioning boundary objects makethem ideally suited to help groups of indivi-duals in conflict to represent visually their(often conflicting) ideas and to transformunaligned or conflicting initial ideas intoideas that depict points of overlap, inter-dependencies and often hidden or emergingwin–win action points. As we have seen inall three vignettes, when used in scriptedsettings by neutral facilitators who bothhave the trust of the group and the abilityto separate process from substance, bound-ary objects can almost always assure thatconflicting ideas are placed on the tableand points of conflicting interdependenciesexplicitly mapped. That is, boundary objectscan almost always constructively depict, ifnot resolve, conflicts when used properly.

3. Use boundary objects to create reinforcingtrust cycles. Getting to ‘yes’ or getting groupsin conflict to agree almost always involves therapid creation of some sort of intangible stockof trust or goodwill, if only a calculated formof trust that lasts for the duration of the meet-ing. We believe these stocks of trust accrue asmembers in conflict work together to createand transform visual representations of theircollective ideas. The very act of co-creating avisual representation of conflicting ideas,when accomplished as a joint activity by parti-cipants, can build mutual respect and trust.Vignette 1 illustrated how an anonymous,face-to-face and computer-supported brain-storming process can touch off three reinfor-cing processes that we labelled ‘We are allheard’, ‘We can be clear about how we areaffected’ and ‘We are all in this together ’.Rapidly built stocks of trust and mutualrespect are essential to reaching commonagreement by the end of the group’s facilitatedprocess. As discussed in all three vignettes,leadership openness is a key interveningvariable that has been demonstrated to drivea group’s ability to agree.

In sum, visual representations must be createdand transformed with processes that produce thethree powerful effects of boundary objects: (1)representations tangible to participants, given




their different experiences, (2) an ability to por-tray dependencies among participants’ resourcesand goals and (3) transformability. Boundaryobjects can and do drive conflict resolution incollaborative model-building meetings whencombined with the following: (A) process andcontent separation, (B) a contract of trust betweenfacilitator and participants and (C) explicit,scripted sub-processes that focus on visualproducts and deliverables.

ACKNOWLEDGEMENTS

The authors would like to thank the editors andthree anonymous reviewers for helpful sugges-tions and comments.

REFERENCES

Ackermann F, Andersen DF, Eden C, Richardson GP.2011. ScriptsMap: a tool for designing multi-methodpolicy-making workshops. Omega 39(4): 427–434.

Ackermann F, Eden C. 2001. Contrasting single user andnetworked group decision support systems for strategymaking. Group Decision and Negotiation 10(1): 47–66.

Ackermann F, Eden C, Brown I. 2005. The Practice ofMaking Strategy. Sage: London.

Ackoff RL. 1979. The future of operational research ispast. The Journal of the Operational Research Society30(2): 93–104.

Andersen DF, Richardson GP. 1997. Scripts for groupmodel building. System Dynamics Review 13(2):107–129.

Andersen DF, Richardson GP, Vennix JAM. 1997.Group model building: adding more science to thecraft. System Dynamics Review 13(2): 187–201.

Argyris C. 1993. Knowledge for Action. A Guide toOvercoming Barriers to Organizational Change. Jossey-Bass Wiley: San Francisco.

Blumer H. 1969. Symbolic Interactionism: Perspectiveand Method. University of California Press: Berkeley,CA.

Carlile PR. 2002. A Pragmatic View of Knowledge andBoundaries: Boundary Objects in New ProductDevelopment. Organization Science 13(4): 442–455.

Carmeli A, Brueller D, Dutton JE. 2009. Learning behav-iors in the workplace: the role of high-quality interper-sonal relationships and psychological safety. SystemsResearch and Behavioral Science 26(1): 81–98.

Eden C, Ackermann F. 1998.Making Strategy: The Journeyof Strategic Management. Sage: London.

Eden C, Ackermann F, Bryson J, Richardson GP,Andersen DF, Finn C. 2009. Integrating modes ofpolicy analysis and strategic management practice:requisite elements and dilemmas. Journal of theOperational Research Society 60(1): 2–13.

Eden C, Huxham C. 2006. Researching organizationsusing action research. In Handbook of OrganizationStudies, Nord W (ed.). Sage: Beverly Hills; 388–408.

Eisenhardt KM. 1989. Building theories from case studyresearch.Academy ofManagement Review 14(4): 532–550.

Henderson K. 1991. Flexible sketches and inflexible databases: visual communication, conscription devices,and boundary objects in design engineering. Science,Technology, & Human Values 16(4): 448–473.

Henderson K. 1998. The role of material objects in thedesign process: a comparison of two design culturesand how they contend with automation. Science,Technology, & Human Values 23(2): 139–174.

Hovmand PS, Rouwette EAJA, Andersen DF, et al. 2011.Scriptapedia 3.04. http://www.systemdynamics.org/conferences/2011/proceed/papers/P1404.pdf [21July 2011].

Howick S, Ackermann F, Andersen DF. 2006. Linkingevent thinking with structural thinking: methods toimprove client value in projects. System DynamicsReview 22(2): 113–140.

Kim H. 2009. In search of a mental model-like conceptfor group-level modeling. System Dynamics Review25(3): 207–223.

Lave J. 1988. Cognition in practice: mind, mathematics,and culture in everyday life. Cambridge UniversityPress: Cambridge, New York.

McCartt AT, Rohrbaugh JW. 1989. Evaluating groupdecision support system effectiveness: a perform-ance study of decision conferencing. Decision SupportSystems 5(2): 11.

McKenzie J, Winkelen CV. 2011. Beyond words: visualmetaphors that can demonstrate comprehension ofKM as a paradoxical activity system. SystemsResearch and Behavioral Science 28(2): 138–149.

Randers J. 1980. Guidelines formodel conceptualization.In Elements of the System Dynamics Method, Randers J(ed.). Productivity Press: Cambridge, MA; 117–139.

Richardson G, Andersen DF. 1995. Teamwork in groupmodel building. SystemDynamics Review 11(2): 113–137.

Richardson GP. 2006. Concept models. Paper pre-sented at the International Conference of the SystemDynamics Society, Nijmegen, The Netherlands.

Richardson GP, Andersen DF. 2010. Systems thinking,mapping, and modeling in group decision and negoti-ation. In Handbook of Group Decision and Negotiation,vol. 4, Kilgour DM, Eden C (eds.). Springer: Dordrecht;313–324.

Rittel HWJ, Webber MM. 1973. Dilemmas in a generaltheory of planning. Policy Sciences 4(2): 155–169.

Schein EH. 1969. Process Consultation: Its Role inOrganizationDevelopment. Addison-Wesley PublishingCompany: Reading, MA.




http://www.systemdynamics.org/conferences/2011/proceed/papers/P1404.pdf

http://www.systemdynamics.org/conferences/2011/proceed/papers/P1404.pdf

Star SL, Griesemer JR. 1989. Institutional ecology, ‘transla-tions’ and boundary objects: amateurs and profes-sionals in Berkeley’s Museum of Vertebrate Zoology,1907–39. Social Studies of Science 19(3): 387–420.

Vennix JAM. 1996. Group Model Building: FacilitatingTeam Learning Using System Dynamics. John Wiley &Sons: New York.

Vennix JAM. 1999. Groupmodel-building: tacklingmessyproblems. System Dynamics Review 15(4): 379–401.

Vennix JAM, Akkermans HA, Rouwette E. 1996.Group model-building to facilitate organizational

change: an exploratory study. System DynamicsReview 12(1): 39–58.

Wolstenholme E. 1994. A systematic approach tomodel creation. In Modeling for Learning Organiza-tions, Morecroft JDW, Sterman JD (ed.). ProductivityPress: Portland, OR.

Zagonel AA. 2002. Model conceptualization in groupmodel building: a review of the literature exploringthe tension between representing reality and negotiat-ing a social order. Palermo, Italy (July 28–August 1).Winner of the 2002 Dana Meadows Award.




using visual representations as boundary objects to resolve conflict in collaborative model-building...

Documents