16
What's the Problem? An Introduction to Problem Structuring Methods JONATHAN ROSENHEAD London school of Economics Houghton Street London WC2A 2AE England OR's traditional problem-solving techniques offer remarkably little assistance in deciding what the problem is. New problem structuring methods (PSMs) provide decision makers with sys- tematic help in identifying an agreed framework for their prob- lem. The result is either a well-defined project that can be ad- dressed using traditional OR methods, or a clarification of the situation that enables those responsible to agree on a course of action. In principle, PSMs can provide analysts with greater ac- cess to strategic problems—those engaging multiple relatively independent decision makers. PSMs' transparent methods of representation can capture differing perceptions of the situation, to help generate a consensus or to facilitate negotiations. P roblem structuring methods (PSMs) a range of problem situations for which are a broad group of problem-han- more classical OR techniques have limited dling approaches whose purpose is to as- applicability. sist in structuring problems rather than Operations researchers have been devel- directly with solving them. They are partie- oping and using individual problem struc- ipative and interactive in character and in turing methods since the mid-1960s. How- principle offer operations research access to ever, only recently have they recognized Copyright © 1996, Institute for Operations Research PROFESSIONAL—OR/MS IMPLEMENTATION and the Management Sciences PROBLEM DIAGNOSIS 0092-2102/96/2606/0117$01.25 This paper was refereed. INTERFACES 26: 6 November-December 1996 (pp. 117-131)

Problem Structuring Methods Psm

Embed Size (px)

Citation preview

Page 1: Problem Structuring Methods Psm

What's the Problem? An Introduction to ProblemStructuring Methods

J O N A T H A N R O S E N H E A D London school of EconomicsHoughton StreetLondon WC2A 2AEEngland

OR's traditional problem-solving techniques offer remarkablylittle assistance in deciding what the problem is. New problemstructuring methods (PSMs) provide decision makers with sys-tematic help in identifying an agreed framework for their prob-lem. The result is either a well-defined project that can be ad-dressed using traditional OR methods, or a clarification of thesituation that enables those responsible to agree on a course ofaction. In principle, PSMs can provide analysts with greater ac-cess to strategic problems—those engaging multiple relativelyindependent decision makers. PSMs' transparent methodsof representation can capture differing perceptions ofthe situation, to help generate a consensus or to facilitatenegotiations.

Problem structuring methods (PSMs) a range of problem situations for which

are a broad group of problem-han- more classical OR techniques have limiteddling approaches whose purpose is to as- applicability.sist in structuring problems rather than Operations researchers have been devel-directly with solving them. They are partie- oping and using individual problem struc-ipative and interactive in character and in turing methods since the mid-1960s. How-principle offer operations research access to ever, only recently have they recognizedCopyright © 1996, Institute for Operations Research PROFESSIONAL—OR/MS IMPLEMENTATIONand the Management Sciences PROBLEM DIAGNOSIS0092-2102/96/2606/0117$01.25This paper was refereed.

INTERFACES 26: 6 November-December 1996 (pp. 117-131)

Page 2: Problem Structuring Methods Psm

ROSENHEAD

them as constituting a coherent field that isimportant for the current practice and fu-ture prospects of OR.Limitations to the Orthodox Heritage

Since the late 1960s, analysts have ac-tively debated claims for the objectivity ofOR/MS models and the limitations im-posed on OR/MS practice by itsconcentration on well-defined problems.Critics (for example, Ackoff [1979a],Checkland [1983], and Churchman [1967])noted the assumption behind standard ORtechniques that relevant factors, constraints,and the objective function are both estab-lished in advance and consensual. Consis-tent with this, standard formulations of ORmethodology (for example, formulate,model, test, solve, and implement) take astheir foundation the possibility of a singleuncontested representation of the problemsituation under consideration. Within thisframework, there is no intellectual impedi-ment to adopting optimization as the cen-terpiece of OR's technical repertoire.

Critics have generally recognized thatOR practice has been considerably more di-verse than this and in particular is far fromdominated by considerations of optimality.They have argued, however, that the avail-able tools and the accepted wisdom onmethodology give scant guidance to ana-lysts confronting less well-behaved situa-tions.

An in-depth analysis by Greenberger,Crenson, and Crissey [1976] of the RANDCorporation's experience in its seven-yearengagement with the problems of urbangovernment in New York City providessupportive empirical evidence for this hy-pothesis. They found that the traditionalOR approach worked well where

—The client organization was structured ina tight hierarchy;—Few of its members were analyticallysophisticated;—The organization performed a well-defined repetitive task generating reliabledata; and—There was general consensus on priori-ties.In the New York fire service, it had consid-erable impact; in the New York publichealth service, it suffered ignominiousfailure.

The situations that satisfy the conditionsGreenberger, Crenson, and Crissey identi-fied are predominantly those in which uni-tary control is exercised over uncontentiousactivities. Within these confines, the stand-ard OR approach with its battery of power-ful techniques can be formidably effective.Many aspects of our complex and inter-locking social arrangements, in the absenceof such analytic inputs, could operate, if atall, only ineffectually or with unreasonablewaste of effort. This much is not in dispute.Many, however, have become concernedthat the exclusion of problem domains thatviolate some or all of the Greenberger con-ditions may be bad both for OR and for so-ciety at large. It is this concern that moti-vates and justifies the development ofproblem structuring methods.Dichotomies

This analysis of the RAND experiencehas already pointed to a dichotomy ofproblem situations. Other commentatorshave made broadly parallel observations:tame versus wicked problems [Rittel andWebber 1973], problems versus messes[Ackoff 1981]. Schon [1987] captures thisdistinction graphically in his extended met-

INTEREACES 26:6 118

Page 3: Problem Structuring Methods Psm

PROBLEM STRUCTURING METHODS

aphor contrasting the "high ground,"where problems are of great technical inter-est but of limited social importance, withthe "swamp," where messy, confusingproblems defy technical solution. In eachcase, the authors reach the same conclu-sion—that the methods for problem han-dling appropriate to pacified conditions donot transfer to more turbulent and prob-lematic environments.

Why this should be so emerges readilyfrom a consideration of the nature ofswamp conditions/messes/wicked prob-lems. Typically there is a range of actorswho are not in subordinate-superordinaterelationships with each other; they have aconsiderable degree of autonomy. The dif-ferent actors have their own interests andperspectives that lead them to pursue dif-ferent objectives and to identify differentfactors as relevant. The potential is here forconflict, often exacerbated by the high lev-els of uncertainty that actors commonlyhave to endure about their own options,their likely consequences, the objectivesand possible tactics of others, and so forth.

Where no one can give orders, knowingan optimal solution is of Httle use, espe-cially if it is the optimal solution to onlyone party's version of the problem. Whereparticipants need to interact and negotiateto reach agreement, mathematics and com-puter algorithms that are hard to under-stand will tend to get in their way. Whereuncertainty reigns, their confidence tomake decisions may not be best advancedby the unambiguous specificities of a singleOR formulation.

Various ways around this difficulty havebeen proposed. Thus optimizing techniquesmay be valued less for the solutions they

find than for their well-developed sensitiv-ity-analysis routines, which a skilled crafts-person can use as sophisticated diagnosticinstruments [Beale 1980]. Alternatively,Cyert [1981] suggests that more and yetmore powerful analytic work will enableOR to "tame" these as yet ill-structuredproblems one by one. Decision analysts aimto extend the range of traditional analysisby incorporating subjective estimates ofprobabilities and of the values of outcomeswhen they cannot obtain objective informa-tion. Simon [1987] proposes that OR andartificial intelligence are complementary:where formulating problems precisely isimpossible, we can develop expert systemscapable of replicating human judgment. Intheir various ways, these prescriptions forOR's relevance problem outside its strictdomain all amount to—more of the sametreatment that generated the debilitatingsymptoms in the first place.The Characteristics of ProblemStructuring Methods

Problem structuring methods provide amore radical response to the poor fit of thetraditional OR approach for wicked prob-lems—a response based on the characteris-tics of swamp conditions rather than on apreexisting investment in high-tech solu-tion methods. These conditions suggest thatdecision makers are more likely to use amethod and find it helpful if it accommo-dates multiple alternative perspectives, canfacilitate negotiating a joint agenda, func-tions through interaction and iteration, andgenerates ownership of the problem formu-lation and its action implications throughtransparency of representation. These socialrequirements in turn have various technicalimplications. Representing problem com-

November-December 1996 119

Page 4: Problem Structuring Methods Psm

ROSENHEAD

plexity graphically (rather than algebrai-cally or in tables of numerical results) willaid participation. The existence of multipleperspectives invalidates the search for anoptimum; the need is rather for systematicexploration of the solution space. Lay peo-ple can generally express their judgmentsmore meaningfully by choosing betweendiscrete alternatives rather than across con-tinuous variables. Estimating numericalprobabilities will need to give way to iden-tifying relevant possibilities. And alterna-tive scenarios will substitute for futureforecasts.

The specification I have outlined for adecision-aiding technology more appropri-ate to messy, strategic problems eliminatesmuch of the scope for advanced mathemat-ics, probability theory, and complex algo-rithms (as practiced, for example, in deci-sion analysis and the analytic hierarchyprocess). It identifies, rather, an alternativeapproach employing representation of rela-tionships, symbolic manipulation, and lim-ited quantification within a systematicframework.

These technical attributes are among theproperties that unite the family of problem-structuring methods. Other common fea-tures concern the process of the engage-ment through which analysis assistsdecision making [Eden and Radford 1990].This process is participative and interactive.Little or nothing happens in back rooms orblack boxes; those who must take or recom-mend decisions are participants in or exec-utants of the analysis. The purpose of theanalysis is to elicit relevant knowledge andto reflect it back in structured form in an it-erative process of problem construction.Typically PSMs operate nonlinearly.

switching freely between different modesor phases of the method in response to thedynamics of group discussion. Outputsmay be visible (recommendations, plans,policies) or invisible (changed apprecia-tions, shared values, better working rela-tions). Of these, the visible outputs, giventhe differing agendas of participants, arelikely to consist of partial rather than com-prehensive commitments.

Virtually all PSMs are designed for useby groups (although they have been widelyappropriated as individual aids to problemclarification). The rationale for this is that ifthe problem situation involves multiple in-terest groups and plural rationalities underconditions in which no group can impose

The assumption is thatrelevant factors, constraints,and the objective function areestablished in advance.

its will, then negotiation of a way forwardmust involve representatives of these par-ties. Such arrangements are common bothwithin organizations—the corporate board,interdepartmental task forces—and be-tween them.

Each member of the PSM family incorpo-rates as a core element the explicit model-ing of cause-effect relationships. This givesPSMs their unambiguous operational re-search identity. It distinguishes them, forexample, from non-OR modes of groupworking, such as organizational develop-ment. PSMs can also be demarcated fromother OR approaches that purport to tacklemessy, ambitious problems (for example,the analytic hierarchy process). PSMs are

INTERFACES 26:6 120

Page 5: Problem Structuring Methods Psm

PROBLEM STRUCTURING METHODS

distinctive in their transparency of method,their restricted mathematization, and theirfocus on supporting judgment rather thanrepresenting it. These limits are impreciseand arguable: approaches developed forother or broader purposes (for example,spreadsheet models) can be used in a simi-lar spirit. Methods that have some degreeof similarity to PSMs, but which are bestregarded as falling outside the category, in-clude decision analysis, decision conferenc-ing, PROMETHEE, scenario planning, sys-tem dynamics, and viable system diagnosis.Other parts of the PSM perimeter are bor-dered by the focus group approach, and byRapid Rural Appraisal and other participa-tive methods for third-world developmentplanning.Some PSMs Summarized

I will attempt here the perhaps unwisetask of summing up in a few sentenceseach the. intention, range of application,and mode of operation of methods that arethe product of years of experience and ofsuccessive refinement.

Eor more detailed coverage of particularmethods, a first stop might be my introduc-tory text [Rosenhead 1989], which containsaccounts of six leading PSMs, each writtenby its principal developer. Another sourceis Flood and Jackson [1991], which offerssummaries and critical discussion of arange of systems-based methods, not all ofwhich could be described as PSMs.

I will describe the methods in alphabeti-cal order:—Hypergame analysis is an interactive ap-proach to taking action in conflict situa-tions. It emphasizes (1) exploring the pat-tern and nature of interactions between theactors, and (2) the effect of differences of

perception among the actors about whatactions are possible, about preferences be-tween outcomes, and so forth [Bennett andCropper 1986].

—Interactive planning (also called idealizedplanning) is a method with the ambitiousaim of designing a desirable organizationalfuture and ways of bringing it about. Ana-lysts generate a reference scenario to dem-onstrate the dire consequences of not tak-ing action. This motivates a participativeprocess in which participants create anideal design for the future of their organi-zation. Other stages of the method dealwith how to bring this future into existence[Ackoff 1979b].

—Metagame analysis is an interactivemethod of analyzing cooperation and con-flict among multiple actors. Analysts usingsupporting software work with one of theparties. They elicit from them decision op-tions for the various actors, from whichthey construct possible future scenarios.Analysts and actors use these as a frame-work to explore their ability to stabilize theoutcome at a more preferred scenario, bythe use of threats and promises [Howard1993].

—Robustness analysis is an approach that fo-cuses on maintaining useful flexibility un-der uncertainty. In an interactive process,participants and analysts assess the com-patibility of alternative initial commitmentswith possible futiire configurations of thesystem being planned for, and the perform-ance of each configuration in feasible futureenvironments. This enables them to com-pare the flexibility maintained by alterna-tive initial commitments [Rosenhead 1980].—Soft systems methodology (SSM) is ageneral method for system redesign. Par-

November-December 1996 121

Page 6: Problem Structuring Methods Psm

ROSENHEAD

ticipants build ideal-type conceptualmodels, one for each relevant worldview. They compare them with percep-tions of the existing system in order togenerate debate about what changes areculturally feasible and systemically de-sirable [Checkland 1981; Checkland andScholes 1990].

—Strategic assumption surfacing and testing isa method for tackling ill-structured prob-lems where differences of opinion aboutwhat strategy to pursue are preventing de-cision. Participants are divided into groups,each of which produces a preferred strat-egy and identifies the key assumptions onwhich it is based. The reunited groups de-bate these strategies and assumptions, mu-tually adjusting their assumptions on theway to an agreed solution [Mason andMitroff 1981].

—Strategic choice approach (SCA) is a plan-ning approach centered on managing un-certainty in strategic situations. Facilitatorsassist participants to model the intercon-nectedness of decision areas. Interactivecomparison of alternative decision schemeshelps them to bring key uncertainties to thesurface. On this basis, the group identifiespriority areas for partial commitment anddesigns explorations and contingency plans[Friend and Hickling 1987].—Strategic options development and analysis

(SODA) is a general problem identificationmethod that uses cognitive mapping as amodeling device for eliciting and recordingindividuals' views of a problem situation.The merged cognitive maps provide theframework for workshop discussions, anda facilitator guides the group towards com-mitment to a portfolio of actions [Eden,Jones, and Sims 1983].

PSM ApplicationsThe extent of practical experience with

these methods varies. Undoubtedly themethods that have the most extensive trackrecord, in terms both of numbers of appli-cations and of their migration away fromtheir original developers, are SSM, SCA,and SODA. However, there has been verylittle systematic research into the extent oftheir penetration. An exception is the sur-vey by Mingers and Taylor [1992], dealingexclusively with SSM. Though the samplewas nonrandom, the survey did indicatelarge numbers of users, a wide dispersionof application types, and a considerable de-gree of user satisfaction.

There are no "typical" applications ofthese methods. Nevertheless I will indicatehere some accessible accounts of practicalapplications, which can provide guidancefor those attempting to assess the costs andbenefits of getting to know one or other ofthe PSMs in more detail.

My edited collection [Rosenhead 1989] in-cludes chapter-length accounts of case stud-ies for each of six PSMs to complement themore theoretical expositions of scope, model-ing formalism, and procedures. Other in-structive case studies include the following:—Best, Parston, and Rosenhead [1986] de-scribe an application of robustness analysisto planning health services for Ottawa andthe surrounding region of Ontario. Theyused DELPHI and cluster analyses in con-junction to generate alternative possible fu-tures. Extensive consultation with healthcare deliverers and with the public wereother features of the study.—Checkland and Scholes [1990] providethe fullest description so far of SSM in ac-tion, with detailed case studies of work in

INTERFACES 26:6 122

Page 7: Problem Structuring Methods Psm

PROBLE]VI STRUCTURING METHODS

government agencies, major corporations,and other bodies. The most extended ac-count concerns reshaping the internalorganizational structure and processes of adivision of a major transnational corpora-tion.

—Eden [1985] describes an application ofSODA in the publishing industry to deter-mine strategic directions for a group ofmagazines. The facilitator adapted theSODA procedures to involve members ofthe large management team in waysreflecting their distinctive potentialcontributions.

—Howard [1992] discusses and comparesthree applications of metagame analysis: (1)to improve bidding success for an engi-neering firm, (2) to help a supplier of rawmaterials to increase margins by squeezingdistributors, and (3) to develop strategy forgovernment negotiations with industry rep-resentatives over reductions in air pollu-tion.

—Moulin [1991] helped women dissatisfiedwith health service provision to articulatetheir demands for better birthing facilities.He used the strategic choice approach toenable them to crystallize what theywanted. Their greater effectivenessenabled them to influence the servicesprovided.

Brief accounts like these cannot providereaders with a sense of how problem struc-turing methods operate in practice. To re-duce this difficulty, I will give more ex-tended accounts of two contrasting casestudies. The first illustrates the potential ofPSMs to work interactively to produce awell-defined project that can be addressedusing a more traditional OR method. Thesecond indicates their ability to clarify the

problem situation, so that participants canagree on action without "hard" analyticassistance.Developing Models to Support a Claimfor Damages

Analysts can employ SODA in a range ofways as a "front end" for analyses whichare then conducted by more conventionalmethods. They can use SODA for the prob-lem formulation stage, after which, for ex-ample, they may find the problem amena-ble to evaluation by multicriteria decisionmethods; or the process may focus atten-tion on a key uncertainty that requires tar-geted market research.

In a study reported by Williams et al.[1995], SODA was used in a more complexway in conjunction with system dynamics.A team from the University of Strathclydecarried out the work for the Canadian-based company Bombardier. Through asubsidiary. Bombardier had contracted todevelop and supply the shuttle wagons totake cars and buses through the ChannelTunnel linking Britain and France by rail.The client was pursuing a legal claimagainst the tunnel's builders, Trans MancheLink, based on the delay and disruption italleged had been caused to its activities byTML's introduction of revised specifica-tions and by TML's delays in approvingdesign documents. Revised specificationsfor a particular component may not onlyresult directly in extra design time; theycan also have extensive ramifications forother parts of the system. Delayed approv-als can disrupt the design schedule, resultin the need to work on designs in parallelrather than in series, and cause designwork to be carried out in ignoranceof the specification of related parts, leading

November-December 1996 123

Page 8: Problem Structuring Methods Psm

ROSENHEAD

to the need to redesign. The analytic prob-lem was to develop models capable ofquantifying the resulting delay and disrup-tion, and capable of carrying conviction incourt.

The importance of feedback cycles in theproblem situation made system dynamics(SD) a natural modeling language. How-ever, the problem was very large (some 300variables were eventually required), knowl-edge was distributed over the managementteam, and individual managers had differ-ent ideas about how the various factors in-teracted. The first use of SODA on the pro-ject was as a knowledge acquisitionprocess; the academic consultants inter-viewed each member of the client team,and recorded his or her perceptions of rele-vant factors and cause-effect relationships

The consultant is a facilitator.

as a cognitive map. The Strathclyde teammerged these maps into a group map,showing the network of ideas elicited. Theclient team, together with lawyers workingon the claim, debated, corrected, and elabo-rated the structure of this group map at aseries of workshop sessions, assisted bymembers of the Strathclyde team.

This validated SODA model would forma continuing reference framework for theremainder of the study—as a reminder ofthe broader relevance of particular detailedanalyses, as a means of identifying whichaspects of the resulting SD model wouldneed revision when new assumptions ordata were introduced, and as a medium forexplaining progress to those outside theproject team. More immediately, it served

as the raw material for constructing an in-fluence diagram as an intermediate stagetoward constructing the SD model. Theconsultants generated the influence dia-gram by identifying and extracting thefeedback loops (98 in all) embedded in theSODA model. Clustering those loops withshared components helped them to priori-tize areas for analytic work—primarily thedesign stage and then secondarily the pro-cess engineering and manufacturing stagesof the project.

As the team moved from influence dia-gram to SD model, the SODA model re-mained in play as a knowledge repositoryfor the study and as a medium of commu-nication between analysts and clients. Itwas of particular value that, whereas an SDmodel represents only one view of theproblem at a time, the SODA model couldstore and relate inconsistent perspectives.The views of corporate managers often dif-fered from those of plant managers, whileverbal reports conflicted with much of thehard data. As new inputs became available,they could be used in conjunction with theSODA model to reevaluate the validity ofthe viewpoints that had been incorporatedin the SD model.

The resulting SD model was required toreproduce historically observed behavior(for example, the total number of freelancedesigners employed and their distributionover the lifetime of the project) with credi-ble accuracy. It would then be used to em-ulate what would have happened if actionsthat were the subject of legal action had nottaken place. The difference between pairsof runs would then form the basis of acosted claim for delay and disruption.

In the event, the legal action was settled

INTERFACES 26:6 124

Page 9: Problem Structuring Methods Psm

PROBLEM STRUCTURING METHODS

out of court. Bombardier was clear that theevident power of the modeling approachused to support the claim for costs was in-fluential in securing a very satisfactory set-tlement.

Contesting Health Strategy in a DeprivedInner City Area

Tower Hamlets is probably the most de-prived borough in Greater London. It hasabsorbed successive waves of immigration,most recently from Bangladesh; much ofthe population, of whatever ethnic origin,is subject to vicious circles of disadvantage.As might sadly be expected of this sort ofarea, health services inadequate to the needare part of the problem.

In September 1987,1 received an invita-tion (through a former student) to workwith three local community health organi-zations to develop an external critique ofthe methodology used by the Tower Ham-lets Health Authority to determine health-service provision. I was quick to involveJohn Friend, principal developer of thestrategic choice approach, which it seemedmight be appropriate. Friend [1994] de-scribes the resulting engagement.

One of the three groups was the TowerHamlets Community Health Council, astatutory agency responsible for represent-ing the views of the local population on thehealth care it receives. The second organi-zation was the Tower Hamlets HealthCampaign, a group with a strong trade un-ion involvement, which was active in de-fending the local health service from cuts.The third organization was the TowerHamlets Health Strategy Group. Chairedby the local vicar, this free-standing organi-zation aimed to channel local pressure forimprovement in health services.

On an exploratory visit to meet represen-tatives of the groups, we found that thingshad moved on. They saw an imminentthreat that the district health authority(DHA) would close the accident and emer-gency unit at Mile Fnd Hospital. Mile Endwas effectively an outstation for the RoyalLondon, a prestigious teaching hospital notfar away. However, Mile End was viewedas a key health care resource for the localcommunity. The general view was thatclosing Accident and Emergency (A&E)would prepare the way for the end of acuteservices and the eventual closure of thehospital. The groups' representativesagreed that this impending crisis must takepriority as the focus of the planned strate-gic choice workshop, which should be con-vened with urgency.

The workshop was held on October 15th,with 13 representatives of the three localorganizations attending. Matters had al-ready moved on again. The DHA was nowdue to confirm a management proposal toclose the A&E Unit at a meeting scheduledfor the following week.

The strategic choice approach most com-monly operates entirely in workshop for-mat, as it did at the October 15th meeting.The participants, assisted by one or two fa-cilitators, use the SCA framework to repre-sent their understanding of their situation.Along the way they engage in, perhapsmore than once, a shaping activity, in whichthey establish key areas for decision; a de-signing activity, in which the method helpsthem to identify feasible combinations ofoptions for action in these areas; a compar-ing activity, in which they evaluate these al-ternatives against a range of criteria theysee as important; and a choosing activity in

November-December 1996 125

Page 10: Problem Structuring Methods Psm

ROSENHEAD

which the method leads them towardsagreement on commitments in some areasand exploratory investigations in others.There are decision aiding tools for each ofthese four modes.

During the discussion in the shapingmode at the October 15th workshop, theparticipants identified seven key decisionareas, and the pair-wise connections be-tween them. One such area was possible al-ternative arrangements for A&E facilities inthe borough, which in turn linked to thefunctions of other local hospitals. The fu-ture role of Mile End Hospital (the decisionarea of most urgent concern) was linkednot only to the provision of A&E cover, butalso to decision areas concerning the futureof "community and priority services," andto care arrangements for mental illness andhandicap. (These were two alternative usesfor Mile End if it should cease to be anacute hospital.) Another decision area thatemerged was a proposal, still at an earlystage, to open a medical college in associa-tion with both Mile End Hospital (if it wasstill functioning as an acute hospital) and anearby college of the University of London.

Normally at a strategic choice workshop,participants agree to work with a problemfocus of about three important, urgent, andinterlinked decision areas. In view of theshortness of time—we had only four hoursfor the workshop—and the dominatingthreat of A&E closure, we agreed to takeonly one decision area as the problem fo-cus: the future of Mile End Hospital. Thiswas indeed "planning under pressure"[Friend and Hickling 1987].

The participants shortened the designingphase by agreeing to consider just two con-trasting options for Mile End: maintaining

the full range of acute services, and reduc-ing it to "cold acute" status by the closureof A&E. The implications of these two op-tions were spelled out in more detail byadding in compatible options from otherdecision areas.

The group started the process of compar-ing by listing a range of criteria on whichthese options should be evaluated. Thesewere not only the group's own criteria butthose they thought other parties wouldconsider important. The group tried toidentify where the balance of advantagebetween the two options lay on each of thekey criteria: cost, management control.

Where no one can give orders,knowing an optimal solution isof little use.

teaching, staff morale, and prestige. Thisprocess usefully threw up areas of disagree-ment or uncertainty—in particular the costof continuing to duplicate services at twosites (Royal London and Mile End), the via-bility of cold acute operation, the costs thatclosure would generate elsewhere in thehealth system, and the state of the fabric ofMile End Hospital (and hence the costs ofbringing it up to an acceptable standard).

All of this information was available indiagrammatic form, on the flip charts onwhich we had been recording the problemstructure as it emerged, for the final choos-ing phase. In view of the prevailing uncer-tainties, many of the possible actions thatemerged from discussion were exploratoryin nature, including the possibility of earlydiscussions with the DHA. However, themost creative discussion focused on what

INTERFACES 26:6 126

Page 11: Problem Structuring Methods Psm

PROBLEM STRUCTURING METHODS

form of public reaction to the closure pro-posal the organizations should issue.

Before the workshop, the organizationshad been expecting defeat and thinkingonly in terms of standard responses to thethreat of A&E closure—a letter to the localpaper, a demonstration with placards out-side the hospital. As a result of the morn-ing's meeting, the group could see at leasttwo added dimensions that gave its cam-paign a good chance of fighting off the clo-sure threat. First, the DHA was seen asheavily influenced by the views of power-ful consultants at the Royal London. How-ever, the discussion had established that animportant subset of these consultants wasinvolved in the planned medical college,which depended on Mile End remaining infull operation. There was scope, therefore,to at least undercut the DHA's enthusiasmfor closure.

The second innovative element in the or-ganizations' response was sparked by theinclusion of management control as a di-mension of evaluation. This had broughtthe North East Thames Regional HealthAuthority (NETRHA) into the picture. (Thestructure of the National Health Service in1987 gave regional health authorities veryconsiderable powers over the DHAs withintheir regions.) Group members recalled thatthe DHA was in NETRHA's bad books forfailures of management control in the re-cent past. The group could therefore mountan argument that the proposed closure wasboth a consequence of poor managerial ef-fectiveness by the DHA and a recipe forcontinued inadequate managerial practicesin the future.

Several members went away directly tocompose appropriately targeted letters to a

range of recipients. The workshop ended inexcellent spirits, with a commitment to fur-ther workshops to address longer-term is-sues. Two weeks later, at the first of these,the DHA had already conceded the caseagainst early closure of the accident andemergency unit at Mile End Hospital. TheDHA sent representatives to each of thesubsequent workshops.

Many factors in this account are uniqueto the particular case. However, it is typicalof strategic choice engagements in certainrespects: the ability to work with multipleorganizations in contested areas, to social-ize and purposefully organize knowledgepreviously fragmented among participants,and to generate decision outputs in theform of immediate commitments combinedwith explorations to reduce the more cru-cial uncertainties.Some Current Developments

I have presented the different PSMs asseparate and distinctive entities, which nev-ertheless constitute a recognizable family ofmethods. The family resemblance is rein-forced by the fact that these individualPSMs can be (and have been) disassembledand their component phases used eitherseparately or hitched up to another of themethods.

Many PSMs consist of a loosely articu-lated set of processes (part social, part tech-nical), which permits users considerablefreedom to switch mode or recycle. Theytherefore lend themselves to creative reas-sembly. Users may carry through onlysome stages of a method, or they may com-bine methods: for example, SODA andSSM, strategic choice and robustness, hy-pergames and SODA [Bennett and Cropper1990; Gains and Rosenhead 1993]. There is

November-December 1996 127

Page 12: Problem Structuring Methods Psm

ROSENHEAD

also some limited evidence of work acrossthe divide between hard and soft methods[Bryant 1988]. Experience reported fromShell International [Gibb 1993; Maplestonquoted in Holt 1993] suggests a potentialfor using particular aspects of a PSM with-out embracing the method in its entirety;using a PSM for only a part of the journeyfrom problem appreciation to firm commit-ment; and using PSMs not for stand-aloneprojects but as elements embedded inlarger projects.

One shared feature of PSMs, which dis-tinguishes them from OR's more conven-tional approaches, is the nature of the inter-action between consultant and client group.In orthodox OR, the consultant is an ana-lyst committed to extracting from perhapsrecalcitrant data usable knowledge aboutthe content of the problem confronting herclients. When operating with PSMs, theconsultant is a facilitator, attempting tomanage the complexities and uncertaintiesof problem content while simultaneouslymanaging the interpersonal processes anddynamics of the client group. Facilitation isa specialized and demanding activity; ifPSMs are to establish themselves in the ORrepertoire, changes will be needed in thecontent of OR training [Eden and Radford1990; Phillips and PhUlips 1993].

Almost all the major PSMs now incorpo-rate purpose-built software, either as an op-tional or an indispensable element. Friend[1993] has expressed concern that introduc-ing a computer into a workshop situationwill adversely affect group dynamics. Con-versely, Eden [1989] has suggested that thecomputer should be introduced preciselyfor its potential to depersonalize messagesthat might otherwise generate unproduc-

tive responses by nature of the position orpersonality of their originators. Anotherpositive argument for computer involve-ment is its potential, given the wide avail-ability of PCs and the difficulty in conven-ing full group meetings, to enable work toproceed wherever subsets of the projectgroup can find time and space to gather.Evidently these issues place PSMs along-side group decision support systems[Finlay and Marples 1991] and computersupported cooperative work.Future Prospects for Problem StructuringMethods

It is reasonable to be sanguine about thefuture prospects for PSMs. They offer aroute around the frustrating roadblock thathas held operational research at a distancefrom many of society's more demandingproblems. I see this problem of OR/MS'snonacceptance as the result of a failed at-tempt at inappropriate technology transfer.PSMs appear to provide a more appropri-ate model-based assistance than do conven-tional OR methods to problem specificationand resolution in nonconsensual, multi-party predicaments. If so, they have the po-tential to open up the road to a construc-tive OR engagement with strategy. Theseare early days, and this statement dealswith promise rather than fulfilment. How-ever, we can reasonably expect furthermethodological innovations as this fast de-veloping field multiplies its experience.

The development of PSMs generatedsome imperialist proclamations from eitherside of the hard-soft divide. Both areequally misguided. Problem structuringmethods can no more perform the tasks ofconventional OR than the reverse. WhatPSMs provide is an extension of the OR

INTERFACES 26:6 128

Page 13: Problem Structuring Methods Psm

PROBLEM STRUCTURING METHODS

repertoire, rather than a driving out of for-mer evil spirits.

That said, I am persuaded, with Schon,that the wicked problems, the problems ofthe swamp, are of greater social signifi-cance than those relatively tame problemsthat OR to date has been technically andmethodologically equipped to tackle. Theproblems of the high ground are tractablebecause they have many givens. It isthrough the processes in the swamp thatthese givens are established. The scope toinfluence social developments is thus farwider. Using model-based methods we canachieve a purchase on these situations—butonly if we abandon any notion that theirrole is to replace subjectivity, that is, judg-ment, by analysis. Instead we need meth-ods that interleave judgment and analysis,which aim to support judgment, not substi-tute for it. From many perspectives this lib-eratory aspect of PSMs (by contrast withthe control paradigm of more orthodox ap-proaches) is one of its principal attractions.

The elimination of the assumption of asingular rationality renders PSMs particu-larly appropriate in the public domain, inwhich much of the terrain is contestable.How successfully PSMs can be employedactively in public debate is, however, opento doubt. The small group dynamicsthrough which PSMs operate can generatea high level of commitment among thosewho have shared the experience. However,the resulting conclusions do not necessarilycany conviction to a larger constituency.Contributions to debate may, therefore, bemore indirect.

One area of unconventional OR practicethat has already made extensive use ofPSMs is community operational research.

PSMs have proved themselves welladapted to a nonmanagerialist, nonhier-archical environment, where resource allo-cation is not an issue (because there are noresources) [Thunhurst et al. 1992; Thun-hurst and Ritchie 1992]. There is also someevidence of their applicability to problemsof development in third world countries[Bornstein and Rosenhead 1990; Rosenhead1993].

Public policy making itself is perhaps themost significant area of potential applica-bility for problem structuring methods. Inthe 1970s, overenthusiastic advocates madesuch inflated claims for a high-tech versionof policy analysis that it was discredited,and that left the field open to the danger-ous form of hyper-irrationalism that nowprevails worldwide in governmental circles[Rosenhead 1992]. Multiple advocacy, anecessary basis for sound policy analysis, isa process that could be well served by par-ticipative, multiple-perspective approachesof the kind that problem structuring meth-ods can deploy. Achieving a more struc-tured debate over policy formation wouldbe a prize not just for operational research,but for the society we serve.

ReferencesAckoff, R. L. 1979a, "The future of operational

research is past," Joumai of the Operational Re-search Society, Vol. 30, No. 2, pp. 93-104.

Ackoff, R. L. 1979b, "Resurrecting the future ofoperational research," Journal of the OperationalResearch Society, Vol. 30, No. 3, pp. 189-199.

Ackoff, R. L. 1981, "The art and science of messmanagement," Interfaces, Vol. 11, No. 1, pp.20-26.

Beale, E. M. L. 1980, "Operational research andcomputers: A personal view," Journal of theOperational Research Society, Vol. 31, No. 9, pp.761-767.

Bennett, P. G. and Cropper, S. A. 1986, "Helping

November-December 1996 129

Page 14: Problem Structuring Methods Psm

ROSENHEAD

people choose: Conflict and other perspec-tives," in Recent Developments in O.R., eds. V.Belton and R. O'Keefe, Pergamon Press, Ox-ford, England, pp. 13-25.

Bennett, P. G. and Gropper, S. A. 1990, "Uncer-tainty and conflict: Combining conflict analy-sis and strategic choice," Journal of BehaviouralDecision Making, Vol. 3, No. 1, pp. 29-45.

Best, G.; Parston, G.; and Rosenhead, J. 1986,"Robustness in practice: The regional plan-ning of health services," Journal of the Opera-tional Research Society, Vol. 37, No. 5, pp. 463-478.

Bornstein, C. T. and Rosenhead, J. 1990, "Therole of operational research in less developedcountries: A critical approach," European Jour-nal of Operational Research, Vol. 49, No. 2, pp.156-157.

Bryant, J. 1988, "Frameworks of enquiry: ORacross the hard-soft divide," Journal of theOperational Research Society, Vol. 39, No. 5, pp.423-435.

Gheckland, P. B. 1981, Systems Thinking, SystemsPractice, John Wiley and Sons, Chichester,England.

Gheckland, P. B. 1983, "OR and the systemsmovement: Mappings and conflicts," Journal ofthe Operational Research Society, Vol. 34, No. 7,pp. 661-675.

Gheckland, P. B. and Scholes, J. 1990, Soft Sys-tems Methodology in Action, John Wiley andSons, Ghichester, England.

Ghurchman, G. W. 1967, "Wicked problems,"Management Science, Vol. 14, No. 4, pp. 141-142.

Gyert, R. M. 1981, "The future of operationsresearch," in Operational Research '81, ed.J. P. Brans, North-Holland, Amsterdam, Neth-erlands, pp. 3-10.

Eden, G. 1985, "Perish the thought!," Joumai ofthe Operational Research Society, Vol. 36, No. 9,pp. 809-819.

Eden, G. 1989, "Using cognitive mapping forstrategic options development and analysis(SODA)," in Rational Analysis for a ProblematicWorld: Problem Structuring Methods for Com-plexity, Uncertainty and Conflict, ed. J.Rosenhead, John Wiley and Sons, Ghichester,England, pp. 21-42.

Eden, C; Jones, S.; and Sims, D. 1983, MessingAbout in Problems, Pergamon Press, Oxford,England.

Eden, G. and Radford, K. J., eds. 1990, TacklingStrategic Problems: The Role of Group DecisionSupport, Sage, London, England.

Einlay, P. and Marples, G. 1991, "A review ofgroup decision support systems," OR Insight,Vol. 4, No. 4, pp. 3-7.

Flood, R. L. and Jackson, M. G. 1991, CreativeProblem Solving: Total Systems Intervention,John Wiley and Sons, Ghichester, England.

Friend, J. K. 1993, "Planning together: The evo-lution of a decision-centred approach," paperpresented at IFORS XIII, Lisbon, July 1993.

Friend, J. K. 1994, "Gommunity involvement inhealth strategy in Tower Hamlets," in Commu-nity Works: 26 Case Studies Showing CommunityOperational Research in Action, eds. G. Ritchie,A. Taket, and J. Bryant, PAVIG Publications,Sheffield, England, pp. 137-145.

Friend, J. K. and Hickling, A. 1987, Planning Un-der Pressure, John Wiley and Sons, Ghichester,England.

Gains, A. and Rosenhead, J. 1993, "Problemstructuring for medical quality assurance,"working paper 93.8, Department of Opera-tional Research, London School of Economics,London, England.

Gibb, J. 1993, "Soft Methods in Shell," talk toLondon and South East Operational ResearchSociety, 22 September 1993.

Greenberger, M.; Grenson, M. A.; and Grissey,B. L. 1976, Models in the Policy Process, RussellSage, New York.

Holt, J. 1993, "No waffle in Durham," OR News-letter, January, pp. 8-9.

Howard, N. 1992, "Dealing with other people,"OR Insight, Vol. 5, No. 1, pp. 18-24.

Howard, N. 1993, "The role of emotions inmulti-organisational decision-making," Joumaiof the Operational Research Society, Vol. 44, No.6, pp. 613-623.

Mason, R. O. and Mitroff, 1.1.1981, ChallengingStrategic Planning Assumptions, John Wiley andSons, New York.

Mingers, J. and Taylor, S. 1992, "The use of softsystems methodology in practice," Journal ofthe Operational Research Society, Vol. 43, No. 4,pp. 321-332.

Moulin, M. 1991, "Getting planners to take no-Hce," OR Insight, Vol. 4, No. 1, pp. 25-29.

Phillips, L. D. and Phillips, M. 1993, "Facilitatedwork groups," Joumai of the OperationalResearch Society, Vol. 44, No. 6, pp. 533-549.

INTERFACES 26:6 130

Page 15: Problem Structuring Methods Psm

PROBLEM STRUCTURING METHODS

Rittel, H. W. J. and Webber, M. M. 1973, "Dilem-mas in a general theory of planning," PolicyScience, Vol. 4, No. 2, pp. 155-169.

Rosenhead, J. 1980, "Planning under uncertainty2: A methodology for robustness analysis,"Journal of the Operational Research Society, Vol.31, No. 4, pp. 331-341.

Rosenhead, J., ed. 1989, Rational Analysis for aProblematic World: Problem Structuring Methodsfor Complexity, Uncertainty and Conflict, JohnWiley and Sons, Chichester, England.

Rosenhead, J. 1992, "Into the swamp; The analy-sis of sodal issues," Journal of the OperationalResearch Society, Vol. 43, No. 4, pp. 293-305.

Rosenhead, J. 1993, "Enabling analysis: Acrossthe developmental divide," Systems Practice,Vol. 6, No. 2, pp. 117-138.

Schon, D. A. 1987, Educating the Reflective Practi-tioner: Toward a New Design for Teaching andLearning in the Professions, Jossey-Bass, SanFrancisco, California.

Simon, H. 1987, "Two heads are better than one:The collaboration between AI and OR," Inter-faces, Vol. 17, No. 4, pp. 8-15.

Thunhurst, C. and Ritchie, C. 1992, "Housing inthe Deame Valley: Doing community OR withthe Thurnscoe Tenants Cooperative Part 2,"Journal of the Operational Research Society, Vol.43, No. 7, pp. 677-690.

Thunhurst, C; Ritchie, C; Friend, J.; and Booker,P. 1992, "Housing in the Deame Valley: Doingcommunity OR with the Thurnscoe TenantsCooperative Part 1," Journal of the OperationalResearch Society, Vol. 43, No. 2, pp. 81-94.

Williams, T.; Eden, C; Ackermann, F.; and Tait,A. 1995, "The effects of design changes anddelays on project costs," Joumal of the Opera-tional Research Society, Vol. 47, No. 7,pp. 809-818.

November-December 1996 131

Page 16: Problem Structuring Methods Psm