acilit - ulisboalmc/research/pdfs/1997caise.pdflu s carr i co nu no guim ar ae s inst it u t o sup...

15

Upload: others

Post on 07-Jul-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

Facilitating Analysis and Diagnosis inOrganisations

Lu��s Carri�co� Nuno Guimar�aes

Instituto Superior T�ecnico�INESCR� Alves Redol �

���� Lisboa� Portugalflmc�nmgg�inesc�pt

Abstract� This paper addresses the de�nition of facilitation techniquesto the analysis and diagnosis of organisational problems� Its foundations�the cognitive mapping techniques and a speci�c organisational development approach� are introduced� Then the contributions of technology tothe use and augmentation of these methods are discussed and relevantwork is referred� Afterwards FADO� a facilitation tool set is presented�Extensions to the organisational methodology are proposed The introduction of manipulation and visualisation techniques� algorithms and navigation and query features is explored� The current status is reported andperspectives are drawn�

� Introduction

Diagnosing� de�ning plans� drawing strategies and making decisions are ubiquit�ous activities in organisational life� Their importance is emphasised further by thechange phobia of the last decade� buzzwords like Business Process ReengineeringBPR �� �� emerging with the automation of work processes or Total QualityManagement TQM ���� come to mind� Evidence exists that the indiscriminateuse of pet solutions� focusing merely on formal organisational aspects� encountersunexpected social and environmental drawbacks and permits recurring dysfunc�tional situations� Solutions must be found� with or without information systems�according to consistent and judicious processes� These rest on accurate identi�c�ation of problems� de�nition of alternative plans� formulation of sound strategies�and criterious choice� all supported by a former experience of recorded successesand failures�

Thus� understanding the way consultants� managers or speci�c multifacetedteams think and keep track of their reasoning processes is essential for theconscientious resolution of organisational problems� Here� the use of cognitivemapping techniques ���� ���� emerging from disciplines of Cognitive Science �����provides the necessary means for knowledge representation and comprehension�Orthogonally� the application of Organisational Development OD theories �����classi�es and explain some of the structures and behaviour within organisations�framing the approach and establishing its starting point on a platform of alreadyacquired knowledge� Finally� the usage of computer technologies o�ers mechan�

Page 2: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

isms to facilitate the perceptive activity of the persons doing the diagnose andanalysis tasks� that in some cases are only feasible through expedite methods�

This paper addresses the de�nition of facilitation techniques to the analysisand diagnosis of organisational problems� Its foundations� the cognitive mappingtechniques and a speci�c OD approach� are introduced� Then� the contributionsof computer technology to the application and augmentation of these methods arediscussed and relevant work is referred� Afterwards� FADO� a facilitation toolset is presented� Extensions to the organisational methodology are proposed�in view of a broader appreciation of cognitive mapping and as a result of theintroduction of computer technology� The current status of the system is reportedand perspectives are drawn�

� Foundations

The use of knowledge representation to cope with organisational modelling hasits roots on several disciplines clustered into two main streams� the Cognitive Sci�ences ���� and the OD theories ����� The �rst provides the means for representingand understanding the knowledge that managers� consultants and organisationpractitioners and theorists have for managing and making decisions� The lateraddresses the available knowledge experts already have on organisational dimen�sions and evolving characteristics� and allows the crystallisation of the former insound foundations�

Fig� �� Theoretic contributions for Organisation Analysis and Diagnosis�

Figure � shows the major contributions required to perform analysis and dia�gnosis of organisations� in a multidimensional perspective� that considers both itsformal and informal� material and human aspects� OD theory o�ers the frame�work validated through Social Cognitive notions� Cognitive Mapping providesthe techniques that can be used to capture the required knowledge�

��� Cognitive Mapping of Organisational Aspects

Cognitive maps provide a visual representation for understanding and repres�enting cognition� Gaines and Shaw ���� allude to cognitive or concept maps as

Page 3: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

visual languages� complementary to �natural language as a means for communic�ating knowledge�� Among others� they include semantic networks� bond graphs�PERT charts� Petri nets� and argument maps� ranging from very informal tostrictly formal representations�

From another point of view� in Hu� ����� cognitive maps are introduced asrepresentations for understanding the conceptual structures used in decision mak�ing and strategic thought� Several applications and suggestions of contributionsto strategic management issues are referred� In particular she proposes ���� thefollowing classi�cation�

Concept Relevancy Maps usually result from statistic techniques of text ana�lysis drawing out degrees of importance and association of concepts� Essen�tially they are considered starting points for basic identi�cation of centralconcepts� and themes related sets of concepts�

Concept De�nition Maps follow from the natural classi�cation process thatleads to understanding� Its use aims at the identi�cation of the dimensionsthrough which people group their concepts in categories and the de�nition ofthe abstraction hierarchies those concepts can be classi�ed on� Kelly�s per�sonal construct theory ���� � � ��� �� ��� has been widely used for identi�ca�tion of relevant dimension in� management strategy� collaborative evaluation�decision making understanding and knowledge acquisition in general� Withthese purpose� concept de�nition maps usually combine category and hier�archy identi�cation and are extended with maps pertaining causal reasoningor more generic relations between concepts� In the limit they constitute whatis known as semantic networks �����

Causal Maps rest on the notion of cause or in�uence as the construct aroundwhich understanding is organised in ways that support explanation and choiceof alternative actions ����� Hu� provides an extensive list of application ofcausal map to social domains� Causal maps may vary in the way they areobtained� visualised and analysed� in order to emphasise di�erent aspectsof the cognition process or of the knowledge itself� Visual representations�supported by directed graphs� may be complemented with matrix counter�parts ��� ��� Variations may also occur on the represented causal relations�On decisional issues� Axelrod ��� identi�es six main causal associations per�vading variants of causality positive� negative� undetermined� not positive�not negative and none� Hu� ���� adds four more� though not causal� that sheclaims help the coding process� equivalence or its absence and example A isa subset of B or its denial� Both authors suggest extensions that augmentthe expressiveness of these maps that include the association� of� certaintylevels relations that are speculative or supported by experimental facts�time delays� conditions� irreversibility� etc�

Argumentation Maps convey the structured nature of argumentation� Its ap�plication in negotiation� decision making and diagnosis seems quite adequate�specially when classi�cations or relations are unclear or controversial� Com�mon argumentation maps are those de�ned by Toulmin�s argumentationscheme ���� and Ritel�s IBIS method or those derived from them ��� ��� �� ����

Page 4: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

Basically� Toulmin�s scheme compulsively attaches to a claim some evidencefacts datum� optionally supported by a warrant statement and backed upby a relevant theory� Exceptions to the claim can also be expressed rebuttal�In IBIS an issue proposes an idea that can be answered by one or severalpositions supported or objected by arguments� that can raise new issues�

Frame Maps structure cognition into schemes of frames� constituting complexhierarchical frameworks of expectations� rooted on previous experiences� Lan�guage itself is taken as a manifestation of this underlying structure �����

Several drawbacks that can be found in cognitive mapping techniques ���� ���Of particular relevance in concept de�nition and causal maps are the misun�derstandings about a concept when various subjects are involved� or the multiplemeanings of concepts under di�erent contexts� Here� the elicitation of ontolo�gical terms ���� or the application of already existing taxonomic theories� helpto reduce the advent of those misconceptions� Therefore� in an organisationaldomain it appears normal to make use of OD theories that o�er� to some extend�systematic classi�cation of the dimensions found in organisational settings�

��� Cognitive Maps under an Organisation Development Theory

Stream Analysis ���� is an approach to diagnose� plan and monitor organisationalchange� As expressed by Porras� the whole process focuses on the activities theexpert consultant� manager� or Organisational Change Teams OCT �must un�dertake in e�ecting planned change�� emphasising �the steps and proceduresneeded to change the system� �����

Stream Analysis proposes a methodology for elicitation and a graph�basedrepresentation� articulated within an OD model� It de�nes a sequence of threecharts addressing the problem diagnosis� planning and monitoring phases�

The OrganisationalDevelopmentModel Stream Analysis proposes a classi��cation of the organisational phenomena� by means of the Stream OrganisationalModel SOM� Four streams are identi�ed� organisational arrangements embraceall formal aspects� including goals and strategies� formal roles and communication�ows� co�ordination and rewards systems� social factors comprise the individualand social characteristics� and interaction patterns of people in an organisation�technology encompasses all the factors related to the execution of work� includingtools and equipment as well as job and work �ow design� physical settings en�circles all the structures and objects that constitute the environment where peoplework e�g� space con�guration� physical ambience and interior and architecturaldesign are components of this stream�

To each of these groups� the SOM provides a �ne grain classi�cation of or�ganisational phenomena that helps the OCT in the categorisation of problemsand actions in the organisational change process� Its complete description is outof the scope of this paper and can be consulted in �����

Page 5: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

Problem diagnosis The OCT identi�es� classi�es and describes problemsemerging in the organisation as graph nodes� Then� directed links are drawn toconvey that the problem on the links� origin contributes to the manifestation ofthe problem described in the destination node see �gure �� Apart from the cat�egorisation model this graph constitutes a simple form of causal mapping ���� ���as it represents causal beliefs of the OCT about a delimited domain ����� Simpli�city arises from the fact that only one type of causality is used problem A is thecause of B � positive e�ect� although Porras admits the attachment of weightingschemes to represent relevancy of causality� The elicitation methodology advisesthe drawing of the most important causal relations only and to avoid loops andmutual causality links� Contributing as well is the fact that the elicitation processbegins with a thorough synthesis of the perceived problems� thus reducing thenumber of relevant nodes in the chart�

O6

S1NAs blame others

for problems

No informal

rewards for NAs

S2

...

NA positions are entry-level jobs: no mobility and low wages

O1 No physical

space specially

assigned to NAs

P1

Organizational Arrangements

SocialFactors

TechnologyPhisicalSettings

No means of

communication

with NAs

P2

T2Inefective supply

handling process

T3Breakdown of

instr. washers

No formal

definition of NA

responsibilities

O3

Sporadic work:

NAs in either

over/under load

T4

T1No standard

thechnical policies

and procedures for

patient transport,

supply ...

O4No forlmal incen-

tives for NAs.

NAs

socializing rather

than working.

S5

Responsibility

for supply handling

not clear.

O5

O2No formally

assigned super-

visor for NAs

Inadequate supply

and instrument

room facilities

P3NAs in conflit

with nursing staff

S3

Mechanization ...

creates confliting

demands in NAs...

T5

S4Nas hiding

or leaving job

location

Core problems

a) Stream Problem Diagnossis Chart b) Example Problem Story

S1 S5

O3T5 P3

T1

T4

S3

S4

Fig� �� Stream Problem Diagnosis adapted from Stream Analysis pages ��� ����� Thesample problem relates to the di culties of control of an employee group� the nursingassistants NAs� as a consequence of growth on a community hospital�

The problem diagnostic map will facilitate the identi�cation of �����

core problems� the fundamental problems that originate the malfunctions forwhich the diagnose was undertaken� First the indegree and outdegree� of

� In its simplest form the indegree of a problem de�nes the number of arrows or causes�that point to in�uence� a problem whereas the outdegree speci�es the number of linksemerging from a problem� More sophisticated techniques exist� that consider� secondorder in�uences and the weight of importance each causality was assign with�

Page 6: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

all identi�ed problems is determined� The bigger the outdegree or outde�gree�indegree ratio the bigger the bene�ts that the resolution of that par�ticular problem has on the organisational malfunctions� Then� core problemsare elected from those using a criterion supported by the OCT�s knowledge�

stories� clusters of interconnected problems that� as a group� describe a morecomplex phenomenon than its components� Usually� an articulated solutionmust be provided for the group� These clusters can then be analysed in isola�tion from the whole problem diagnosis chart� thus focusing the OCT e�orts�Problem stories are identi�ed by tracking the causal chain for a speci�edsymptom problems with lower outdegree�indegree ratio� It is not consideredthe whole path of causes� causes of causes� ��� but a criterion determines theselection of problems that enter a story� An example of a problem storyselected from the diagnosis map is shown in �gure � b�

themes� sets of core problems that relate to a common issue and usually requirean integrated resolution� Among its components there usually is no explicitrelation and the theme elicitation relies again on the OCT knowledge�

The premises of simplicity of the diagnosis maps are supported by an intens�ive work and discussion among OCT members� The process of reaching a �nalmap requires consensus in several steps� identi�cation of problems� its classi�c�ation in streams and dimensions� its grouping into more abstract descriptionsor division into more concrete manifestations� de�nition of which are the criticalcausal relations� and the identi�cation of the most relevant causality sense� Otherimportant direct human interventions are the de�nitions of the knowledge basedcriteria used in the analysis of the result diagnosis map� Its relevancy is evenmore stressed if we note that core problems� critical stories and themes will formthe triggering factor for the following steps of the methodology� These decisionsstrongly determine the success or failure of the change process�

Planning and Monitoring Change Planning and Monitoring Charts are es�sentially PERT charts where the OCT describes the actions to be taken plan�ning and what actually happened monitoring along a vertical axis representingtime� A major distinction from simple PERT diagrams is that the categorisation�applied in the problem diagnosis chart� is present in planning and monitoringmaps as well� Porras observes that in these maps links convey triggering or per�mitting instead of causality�

An important aspect emphasised in ���� is that the monitoring chart directlyrelates to the planning one in that a projected action should have a correspond�ing description� positive� negative� or extended in time� in the monitoring map�Also� the planning map should refer the particular problem or set of problemsidenti�ed in the diagnosis map� The classi�cation of the action plans into eachstream enforces that association as actions should be located in the stream of theproblem they preferably solve rather then on the kind of action they are�

Page 7: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

� The Role of Technology

The contribution of computer supported tools for facilitation of the processesinvolved in cognitive mapping is patent in the multiplicity of tools provided�Gaines and Shaw ���� refer several works� emphasising its relation to humancomputer interaction issues� to hypertext and hypermedia systems� and arti�cialintelligence� Other areas also contribute to the adequacy of computer facilitationfor cognitive mapping�

The use of algorithms for expedite statistical analysis of cognitive maps en�counters applications through the whole set of cognition charts categories andtools� in particular in what concerns content analysis concept relevance andcausal mapping� Birnbaum�More and Weiss ��� refer the application of programsfor content analysis� using regression and statistical identi�cation of words andphrases� For causal mapping� the identi�cation of loops� causality paths andrelevance can also be accelerated� CMAP� ���� o�ers several statistics on theuse and relevancy of terms and causal relations for the comparison of causalmaps� COPE � ��� is a strategic decision support tool� based on causal maps�that provides several analysis facilities� ranging from clustering� loop detection�centrality� etc�

The relation between arti�cial intelligence concepts and techniques and cog�nitive maps is obvious� In fact� some of the referred cognitive maps semanticnetworks and frame systems were developed in arti�cial intelligence studies�with the purpose of designing systems able to demonstrate something close toreasoning� The full range of cognitive maps is covered� Logic�Line � is an expertsystem used by Birnbaum�More and Weiss ��� to elicit words which are relatedto a set of speci�ed keywords� Aquinas ���� and KSSn ���� use Kelly�s personalconstruct theory for knowledge acquisition that is then used in evaluation and in�ferences supported by the underlying cognitive maps� The inference mechanismsare drawn from the inclusion of rules or causal relations� thus extending thebasic concept de�nition maps elicited fromKelly�s repertory grids with causal orin�uence relations� KL�ONE � � is a knowledge representation language support�ing frames� Fikes ��� refers the use of frame�based representations in reasoningwithin the KEE system�

Hypertext systems provide a twofold contribution to ease the process of cog�nitive mapping� One derives from the graph support underlying hypertext sys�tems� that �nds direct application in argumentation maps� Several argumenta�tion systems based of Toulmin�s scheme or IBIS have been built ��� ��� �� ����These systems explore the clustering facilities associated with hypertext systemsand visualisation systems in general in the search for solving the problem ofpresenting huge graphs� The other contribution emerges from the need to relatethe cognitive maps with the large amounts of information obtained in the dataacquisition procedures that precede the cognitive mapping practice�

Cognitive maps can de�nitely bene�t from the computers ability to improvevisualisation� Together with the factors associated with hypertext� we can con�

� Renammed to Decision Explorer

Page 8: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

sider shape� font and colour schemes to improve perception� centring the re�searchers attention on particular aspect of a cognitive map� Multiple and di�er�entiated perspectives� emphasising speci�c facets� or allowing the visualisation ofdi�erent maps may provide a broader understanding of the charts� COPE� abovementioned� also provide multiple views of maps and analysis results� and attrib�ute schemes for concepts and links� The manipulation of constructs for editingpurposes can also be enriched� o�ering robust techniques for creation and modi��cation of the maps� This is particularly true in the case of maps supported by aformal syntactic structure� like hierarchy or rigid category maps� PERT or Petrinet charts� Gaines de�nes a visual language ����� covering concept de�nition andextended causality maps� Apples� MacDraw is used with additional features thatenforce the correctness of the edited graphs� according to the syntax de�ned bythe language�

Other interaction enhancements can be envisaged that are hardly coveredin existing cognitive mapping tools� For example� perception can be facilitatedwith the advent of multimedia and multi�modal interfaces� by using alternative orcomplementary channels for conveying the information expressed in the cognitivemaps� Manipulation can be enriched with drag and drop behaviour� conveyingnot only the syntactic formal aspects of the cognitive map� but also the contextand semantic of its particular application�

Now that some systems that explore computation support to facilitate cog�nitive mapping were revised� it is time to address the in�uence of the socialaspects� that of OD theory� At this point� however� the number of computational

Fig� �� Computer contribution for Organisation Analysis and Diagnosis�

tools becomes rather reduced� The solutions usually provided are �at in termsof organisational dimensions� even when they pertain to management or decisionmaking in organisations� The tool set presented next tries to cope with this ques�tion articulating the contributions from computer facilitation� cognitive mappingtechniques and OD theories see �gure ��

Page 9: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

� FADO

FADO is a tool set aiming at the Facilitation of Analysis and Diagnosis of Or�ganisational aspects� based on cognitive mapping techniques and framed withinan OD classi�cation� It departs from the Stream Analysis approach and extendsits methodology�

��� The cognitive mapping enhancements

The enhancements proposed to the Stream Analysis are designed to cope withparticular knowledge acquisition situations or to express knowledge that allowsthe active participation of the tool in the diagnosis processes� From the descrip�tion of cognitive mapping techniques and the observation of the critical points inthe method� some major extensions were designed�

� addition of attributes to causal relations e�g� certainty and relevancy�� addition of levels of certainty to problem classi�cation and restructuring�� use of argumentation maps� sustaining the decision of controversial steps�� addition of implicit and explicit typed relations�� use of concept de�nition maps for generalisation of problem descriptions�

Certainty levels and relevancy weights� in relations and problem classi�ca�tion� allow a more accurate analysis of the diagnosis chart� Particularly� in thealgorithmic identi�cation of core problems� relevancy is used� together with inde�grees and outdegrees� In the de�nition of stories� they are optionally applied tothe determination of relevant causal branches� decreasing the number of problemsthe OCT has to take into account when its decision is required�

Certainty levels also contribute to the identi�cation of contradiction levelsbetween the members of the OCT� in the comparison of diagnosis charts� Forexample� once problems are identi�ed� each OCT member is required to clas�sify and later de�ne the causality relations� expressing certainty and relevancyweighs� The resulting maps are then merged weights are combined accordingto prede�ned formulas� Contradictory classi�cations or causal links assignedwith high certainty levels in more that one chart or low in all of them areselected for further discussion��

The visual representation of certainty and relevancy is done varying graphicalattributes like the continuity and thickness of lines� More relevant links and thebox containing the potential core problems are drawn with thicker lines� Doubtfulrelations or classi�cations are shown with dotted lines�

The use of argumentation maps is enforced each time� a contradiction is per�ceived in problem classi�cation and restructuring or causal relations de�nition� apotential core problem is chosen or rejected� a branch is rejected in a story de�n�ition� a theme is de�ned� and a mismatch is found between the planned actions

� Since certainty can take values other than high or low� the selection criterion is usuallymore complex�

Page 10: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

and the monitored results in their respective charts� Toulmin�s argumentationschemes are used where top most claims refer to one of the above situations�

Causal variants and typed relations can be added to the diagnosis chart� Itspurpose� besides retaining knowledge� is to facilitate the expedite determinationof stories� themes or alternative patterns relevant in other forms of analysis� Re�lations like �must be solved with�� for instance� are used in pre�de�ned criteria�that can be chosen to �nd stories and themes� Other kinds of relation� associatingproblems through common speci�c keywords in their description text� are alsopossible� For example� the use of role names e�g� NA in the example of �gure ��of the formal organisational structure� are included in criteria available to de�termine stories and themes� The relation types are represented attaching a textstring or symbol to the corresponding link� Keywords� relevant in the de�nitionof relations between problems� are emphasised in bold�

Concept de�nition maps aim at the de�nition of the problems at a moreabstract level� These permit the introduction of inference mechanisms� that cansuggest causes and e�ects for equivalent problems in later diagnosis� in the sameor other targeted organisations� The technique can also assist the OCT in theclassi�cation of problems� identi�cation of stories and themes� For this purpose asimple visual language is de�ned� maintaining as much as possible a presentationcoherent with the causal map itself� The language re�ects concepts representedby an oval� hierarchies and the causality rules that can be applied between them�Uncertainty and relevancy are also included�

��� The computer role

FADO is implemented in C��� It uses an improved version of Edgar ����� agraph editing and visualisation library� enhanced with advanced manipulationtechniques ��� that support multiple modalities and constrained compliant beha�viour� An abstract level for access to the underlying expert system was de�nedand experiments have been done with Mike �����

Creating diagnosis maps FADO provides direct manipulation techniques tocreate� describe and classify problems into the columns representing streams�Two alternatives are o�ered� In the �rst� the user creates a node in a streamand then enters its text� classi�cation detail and level of certainty� In the second�problem descriptions are created in an unconstrained space no streams andthen moved to the diagnosis space� This allows di�erent OCT members to departfrom a common� unclassi�ed� set of problems and then create their own diagnosischarts� These charts are later merged using the technique previously referred�

Once problems are de�ned� they can be restructured� For that purpose theuser selects a set of nodes collapsing them into a single one� A new descriptionand certainty level are requested� The system suggests a proper classi�cation ofthe node� placing the node in one of the streams� based on a weighted averageof its components and considering their classi�cation certainty� Problems canalso be split� In collapsing and splitting� an hyper�link to the original problemdescriptions is maintained in the resulting nodes accessible on a pop�up menu�

Page 11: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

For classi�cation� restructuring and linking� the user can associate argument�ation maps� The editing of these maps is done on a speci�c space that enforcesToulmin�s argumentation schema� The claim and datum nodes are automaticallyplaced and the �rst is �lled with the action on debate� For datum� an emptyrectangular node is created and a text editor is presented� Warrant� backing andrebuttal can be selected out of an available toolbar� Hyper�links are maintainedfrom nodes and relations to the associated argumentation�

FADO also provides means for the creation of formal organisational chartsof roles and �ows in speci�c spaces� When these are created� the names andmnemonics of the roles and �ows are searched in problems� textual description�When found� hyper�links are automatically assigned to the words� o�ering ac�cess from the diagnosis map to the organisation�s description� Search criteria ofproblems pertaining to a speci�c role or �ow are also available� They can beused� in the identi�cation of clusters of problems in the restructuring phase or�as referred� as aids for the analysis of stories and themes�

Common to all maps provided by FADO are visualisation techniques forproblem focusing and selection� like zoomed views or simply scroll facilities�Criteria based �lters can also be applied to the nodes and links of the diagnosischart� hiding or emphasising speci�c paths and patterns�

Analysis of diagnosis maps The analysis of the diagnosis map can take ad�vantage of computer facilities� even though the semantic�knowledge based se�lection process di�cults the implementation of a fully automatic process� FADOprovides an automatic identi�cation and gradation of core problems� and a guidedselection of themes and stories�

� Gradation of core problems is done automatically by selecting an available cri�teria outdegree� outdegree�indegree ratio� with or without weights or quad�ratic factors� considering or not speci�c relations� The result is presented ina scaled view� where one of the axis conveys the variables used in the scalingcriteria� It is also possible to specify �lters that de�ne thresholds below whichproblems are not visible� The other axis can also be used to express a factorlike indegree or problem evidence� On this chart� the user can decide whichare the actual core problems�

� The selection of stories starts after the core problem gradation phase or analternative algorithm� since it normally departs from the most symptomaticphenomena bigger indegree or smaller outdegree�indegree ratio problems�Once a problem is selected� the systems computes a sub�graph recursivelycontaining all the causes of that problem� that obey a particular chosen cri�teria� The user is then able to suppress irrelevant branches if the criteria isnot restrictive enough� Alternatively� the process is stepwise� and for eachproblem the user is requested to select the pertinent causes for the workingstory� A directed acyclic graph layout is used for stories presentation�

� Finally� problems are gathered into themes� The choice is done from the elec�ted core problems� Here� pre�de�ned criteria can be also applied for problemselection�

Page 12: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

Fig� �� Framework for Analysis and Diagnosis of Organisations�

The de�nition�visualisation editors that complement the problem diagnosismap allow a focused study of each pattern core problems� stories and themes�These patterns can be also visualised together allowing the OCT to examine asituated perspective� comparing stories and schemes and completing each docu�mentation with relevant global aspects and relations hyper�links� The visualisa�tion of the clusters all core problems and each story or theme are emphasisedby using speci�c marks or convenient colour�schemes� Those can also be used toview the results of the scaling algorithms for out or indegree�

Large graph visualisation techniques� like �sh�eye views ����� can be applied toany of these graphs� facilitating the focusing of attention� An alternative� speciallyuseful for core problems and symptoms emphasis� performs a distortion contraryto the �sh�eye view over a line vertical or horizontal thus providing a largerperspective of the graph extremes core and symptoms�

Each time a semantic�based decision criteria is involved in a pattern selection� that is� when core problem is selected� a causal branch is dismissed in a story�or a problem is included in a theme � a backing argumentation can be stored�Hyperlinks are automatically placed between the problem description and therecorded decision� Future consultations can easily follow those links� selecting onthe menus attached to the problem diagnosis map or on the speci�c pattern mapsand view the corresponding argumentation�

Page 13: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

��� Planning and Monitoring

The computer support provided to the construction of planning and monitor�ing maps are in most aspects similar to those described above for the problemdiagnosis map� Some aspects can� nevertheless be emphasised specially in whatrelates to the domain of application itself�

In fact� the provision of a computational system allows the enforcement of therelations between the three involved maps� FADO immediately begins planningand monitoring map skeletons� drawn from the identi�ed core problems� storiesand themes in the �rst� and the whole planning map in the second� Furthermore�hyper�links are added between those entities� allowing the OCT to easily �nd theprescription for the resolution of problems and the results of its application�

� Current Status and Future Perspectives

FADO is currently in �nalisation phase in what concerns the edition� visualisa�tion� speci�c analysis and hypermedia issues� Work must still be done in theextension of the basic causal maps with typi�cation of links and application oflayered visualisation techniques�

The integration of computer supported group decision techniques� like thoseproposed by Antunes ���� is envisaged� to facilitate the resolution of controversialsteps� However� a previous validation process of both tools is required� For FADO�this validation phase� among a team of expert consultants for organisationaldiagnosis� is quite relevant for the evaluation of the gains obtained with thecomputer support �� but also of the impact and acceptability of the extensionsproposed in the methodology�

Finally� we intend to give a deeper importance to the tool active participationspecially in the problem diagnosis map constructions� The di�culties that emergeare� the identi�cation of recurrent abstract patterns� for example in stories andthemes� that can be used to infer and suggest to the user the existence of similaroccurrences in new diagnosis� and the inherent automatic classi�cation of prob�lems for matching with those patterns� In both� the automatic or semiautomaticanalysis of the problems textual description assisted by the methodology classi��cation and the association with formal structures� like roles and work �ows�seems to provide some feasible ground�

References

�� P� Antunes and N� Guimar�aes� Structuring elements for group interaction� InSecond Conference on Concurrent Engineering� Research and Applications �CE����Washington� DC� August ����� Concurrent Technologies Corporation�

�� R� Axelrod� Structure of Decision� The Cognitive Maps of Political Elites� Princeton University Press� Princeton� New Jersey� �����

�� P� BirnbaumMore and A� Weiss� Discovering the basis of competition in �� industries� Computerized content analysis of interview data from the us and europe� InA� Hu�� editor� Mapping Strategic Thought� John Wiley � Sons� �����

� Stream Analysis is already used in real cases

Page 14: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

�� R� Brachman and J� Schmolze� An overview of the klone knowledge representationsystem� Cognitive Science� ������������ April �����

�� L� Carri�co� Interaction techniques for modelbased tools� Working document available upon request� �����

�� J� Conklin and M� Begeman� gIBIS� A hypertext tool for team design deliberation�In Hypertext ��� �����

�� C� Eden� Strategy development and implementation� Cognitive mapping for groupsupport� In J� Hendry and G� Johnson with J� Newton� editors� Strategic Thinking�Leadership and the Management of Change� AddisonWesley� �����

�� R� Fikes and T Kehler� The role of framebased representation in reasoning� Com�munications of the ACM� ������������� September �����

�� G� Fisher� R� McCall� and A� Morch� JANUS� Integrating hypertext with aknowledgebased design environment� In Hypertext ��� November �����

��� G� Furnas� Generalized Fisheye Views� In Proceedings of the SIGCHI � Confer�ence� pages ������ �����

��� B� Gaines� An interactive visual language for term subsumption languages� Know�ledge Science Institute Home Page� ����� http���ksi�cpsc�ucalgary�ca�

��� B� Gaines and M� Shaw� Knowledge acquisition tools based on personal construct psychology� Knowledge Science Institute Home Page� �����http���ksi�cpsc�ucalgary�ca�

��� B� Gaines and M� Shaw� Concept maps as hypermedia components� KnowledgeScience Institute Home Page� ����� http���ksi�cpsc�ucalgary�ca�

��� M� Hammer and J� Champy� Reengineering the Corporation� Nicholas Brealy������

��� A� Hu�� Mapping strategic thought� In A� Hu�� editor�Mapping Strategic Thought�John Wiley � Sons� �����

��� A� Hu�� editor� Mapping Strategic Thought� John Wiley � Sons� �������� A� Hu�� V� Narapareddy� and K� Fletcher� Coding the causal association of con

cepts� In A� Hu�� editor� Mapping Strategic Thought� John Wiley � Sons� �������� G� Kelly� The Psychology of Personal Constructs� New York� Norton� �������� M� Laukkanen� Comparative Cause Mapping of Management Cognitions� Helsinki

School of Economics and Business Administration� �������� Open University� Mike� ����� Available by anonymous ftp from

ftp���hcrl�open�ac�uk�pub�software�src���� V� Paulo� Modelos� arquitecturas e ferramentas para a edi�c�ao de grafos� Master�s

thesis� Instituto Superior T�ecnico� �������� J� Porras� Stream Analysis� Organization Development� AddisonWesley� �������� M� Posner� editor� Foundations of Cognitive Science� The MIT Press� �������� R� Reger� Managerial thought structures and competitive positioning� In A� Hu��

editor� Mapping Strategic Thought� John Wiley � Sons� �������� G� Rein and C� Ellis� rIBIS� A realtime group hypertext system� Int� J� Man�

Machine Studies� ������������� �������� J� Ross� Total Quality Management� Kogan Page� �������� D� Schuler� P� Russo� J� Boose� and J� Bradshaw� Using personal construct tech

niques for collaborative evaluation� Int� J� Man�Machine Studies� ��� �������� J� Sowa� editor� Principles of Semantic Networks� Explorations in the Represent�

ation of Knowledge� MorganKaufman� San Mateo� California� �������� N� Streitz� J� Hannemenn� and M� Thuring� From ideas and arguments to hyper

documents� traveling through activity spaces� In Hypertext ��� �������� S� Toulmin� The Uses of Argument� Cambridge University Press� �����

Page 15: acilit - ULisboalmc/research/pdfs/1997caise.pdfLu s Carr i co Nu no Guim ar ae s Inst it u t o Sup er ior T ecnicoINESC R Alv e s Re dol Li sb oa P ort ugal f lmcnm g in e scpt A bstract

This article was processed using the LATEX macro package with LLNCS style