Appendix

К оглавлению
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 
136 137 138 139 140 141 142 143 144 145 146 

Supporting Software for the Elicitation, Construction,

Analysis and Comparision of Causal Maps

In all but the very simplest of applications, the use of computer software systems can

greatly assist the researcher in all stages of the causal mapping process, from knowledge

elicitation and map construction to individual and comparative analysis. This is equally

true not only in the case of applications involving the detailed analysis of single

(Cossette & Audet, 1992) or small numbers (Clarke & Mackaness, 2001) of maps, but also

in much larger-scale comparative studies (Markóczy & Goldberg, 1995) of causal maps.

Clearly there are times when small-scale, complex ideographic studies, exploratory and

inductive in nature, conducted in the context of under-explored knowledge domains, are

invaluable. In this type of application, which can result in maps containing as many as

several hundred concepts (Eden & Ackermann, 1998a), it is impracticable to analyze the

structure and content of the maps using basic manual procedures. However, much of the

utility of causal mapping techniques in organizational research lies in their application

to larger numbers of individuals and/or groups, comparing their similarities and differences

in a range of contexts and/or over multiple points in time. As noted in the main body

of the chapter, such comparisons are potentially unwieldy, but fortunately recent

developments in mobile computing and associated software advances are paving the

way for new support systems that will rapidly resolve these difficulties.

Generic computer software tools such as ATLAS/ti (Jasinski & Huff, 2002) are enabling

ideographic researchers to tackle more demanding problems and extend their analyses

considerably further than would have been possible using manual coding techniques.

Moreover, software systems devised for the structural analysis of social networks, such

as UCINET (Borgatti, Everett & Freeman, 1992), are potentially also suitable for the

analysis of causal maps, having common mathematical roots in graph theory. Indeed,

many of the structural indices commonly employed by network analysts and routinely

available in software packages to support the analysis of social networks bear a strikingly

close resemblance to those devised by Eden et al. (1992) specifically for the analysis of

causal maps. There is no doubt that these software tools are extending the range of

computer technology broadly capable of supporting causal mapping. One other system

Table A1. Selected software supporting causal mapping

Name of

Progam

Approach to Knowledge

Elicitation/ Map Construction

Output Statistics Strengths Limitations Illustrative applications Source of Software

ATLAS/ti Though not developed

specifically for causal mapping

purposes, this knowledge-based

computer system (KBS) has

been used to support the

construction of maps based on

ideographic interview

transcripts and could equally be

applied to facilitate the

construction of maps from other

forms of documentary data.

Not Applicable (N/A) The software assists researchers in

identifying patterns among key

concepts in documentary data,

enabling them to be depicted

graphically in the form of nonhierarchical

network structures.

Data are readily exportable to other

software packages (

e.g. SPSSTM)

suitable for performing quantitative

analyses of the text patterns

identified.

Data must take the form of plain ASCII

or ANSI text, i.e. there is no recognition

of word processor formats (though data

can be internally translated).

The reliability and validity problems

associated with the coding of maps

constructed from ideographic data

gathered through indirect means (as

detailed in the main body of this chapter)

apply when using this software package.

Jasinski &

Huff, (

2002)

The package is available commercially

from Scolari, Sage Publications

A demo version is downloadable from:

http://www.scolari.com

CMAP2 This software was developed to

support the construction of maps

indirectly elicited from

documentary sources, including

interview transcripts.

Documentary data is coded post

hoc. In addition to modules for

the inputting of raw data and the

creation of a standard language

vocabulary for coding natural

language expressions and

achieving comparability over

research participants, the

program also contains a variety

of tools for editing the input

data and the generation of

analyzable databases.

Key numerical outputs from

the program include

measures of the distances

between the maps of

individual participants or

clusters of participants.

CMAP2 is based on the observation

that causal maps can be decomposed

into causal units, i.e. conceptconcept

pairs assumed to be causally

linked to one another. The

computer can process these causal

units as semi-independent data

entries.

The software links the computer’s

databases with the input data

elements and thus with the source.

As such, it promotes good

organization of data, and enables an

audit trail from the source to the

final standardized concepts and

causal units.

The software is currently written in DOS

mode. As acknowledged by the author,

this does not permit easy export of the

various output data into better-developed

user interface environments for

supplementary statistical analysis, using

standard software packages such as

SPSS.

Moreover, the absence of methods for

direct raw data acquisition and input by

the participants themselves, in

conjunction in with primary-level

standardization facilities and

opportunities for concurrent feedback,

renders the whole process extremely

labor intensive on the part of the

researcher.

Laukennan (1994, 1998) This software is available on a noncommercial

basis from the developer.

For details, see Laukennan (1998: 189)

Decision

Explorer

This software was developed to

permit the direct elicitation and

construction of cause maps.

Typically, maps are constructed

iteratively, in face-face meetings

in situ

This software enables the

graphical representation of

maps as well as the

calculation of a

variety of

quantitative indices of a

structural nature.

This software has proved to be of

immense benefit in areas where

interactively generated maps have

helped to build up a comprehensive

qualitative map or model, which is

then explored and analysed to help

develop strategy, decision making

and business problems

Decision Explorerallows

researchers to manipulate the data

and hence view it from a

variety of

perspectives. This is not only

beneficial from an analytical

standpoint, but enables the

researcher to actively gain and

maintain the interest of participants

in the research process.

Data can be readily imported from

and exported to a

number of

standard software packages for

supplementary qualitative and/or

quantitative analysis (e.g. QSR

NVivo and SPSSTM).

Overly simplistic or complex and

‘messy’ maps may be constructed at the

original stage of data input.

As with ideographic methods more

generally, the highly idiosyncratic nature

of the data renders problematic the

systematic comparison of maps,

particularly where large numbers of maps

are involved.

There are no features to support the

analysis of inter-coder and code-recode

reliability. (

For an explanation of these

terms see the section entitled

‘Psychometric Issues’ in the main body

of this chapter).

The software does not permit the links

between variables to be formally

weighted numerically, ,

although basic

polarities (positive and negative) can be

depicted and important variations can be

identified through the use of contrasting

color codes and/or variations in the

relative thickness of the lines

interconnecting the concept nodes.

Cropper, Eden & Ackermann

(1990)

Eden & Ackermann (1998a,

1998b).

This software package is available

commercially from Banxia Software

Ltd.

A demonstration copy can be

downloaded from:

http://www.banxia.co

Table A1. Selected software supporting causal mapping (continued)

Name of

Progam

Approach to Knowledge

Elicitation/ Map Construction

Output Statistics Strengths Limitations Illustrative applications Source of Software

Distrat/

askmap

suite of

programs

Not applicable.

This suite of programs was

devised facilitate the comparative

analysis of cause maps elicited

using the hybrid procedures

devised by ________    

               

Goldberg

(1995)

The programs perform several

of the analytical tasks

associated with the ________-

Goldberg approach.

As stated in Goldberg’s (

1996) analysis

guide, the software is inflexible and of little

benefit to anyone not using the particular

methods described. The input and output

data are configured such that they are not

readily transferable to and from other

systems, especially of the larger, fully

integrated, interactive variety.

__

_

____

_

               

_

_

__           _

__          

_____

__

_

____

_

                              

Goldberg (1995)

The complete collection of programs is

available to researchers on a noncommercial

basis. For details see

Goldberg (

1996), down loadable from:

http://www.goldmark.org/jeff/programs/d

istrat/software/drdoclet.ps.gz.

KNOT

(Knowledge

Network

Organizing

Tool)

Not applicable.

This software package was not

developed specifically for the

purpose of causal mapping. Its

purpose is to implement the

Pathfinder network algorithm.

The Pathfinder algorithm seeks to

represent the information

contained in the input proximity

matrix in as few links as possible,

consistent with the parametric

values set by the investigator. The

aim is data reduction to facilitate

comprehension of the resulting

network.

The software performs a

similar range of functions to

programs for the analysis of

social network structures, i.e.

the input distance measures

are represented as nodes, with

links representing the relations

between objects. Weights

associated with the links,

derived from the original

proximity data, reflect the

strength of the relations.

The KNOT system includes several

programs and utilities to facilitate

Pathfinder network analyses of

proximity data. The system is oriented

around producing graphical

representations of the solutions, but

representations of networks and other

information are also available in the

form of text files, which can be used in

conjunction with other software.

Although it was designed for in DOS

mode with minor modifications it runs

in a Windows environment.

This software does not support the initial

elicitation of data for the construction of

cause maps.

For details of the general

Pathfinder method and its

application see Schvaneveldt

(1990)

KNOT is a commercially distributed

software package, available from

Interlink. For details see

http://www.interlinkinc.net/Pathfinder.ht

ml

UCINET

N/A.

This package was devised for the

analysis of social network data

(typically questionnaires or

documentary sources).

A number of basic output

statistics for the identification

of social network structures

including, for example,

centrality measures, predicated

upon elementary graph theory

are equally suited to causal

mapping applications. In

addition, the package has a

number of matrix analysis

routines useful in the context

of cause map analysis.

The software program NetDraw is

integrated with UCINET for drawing

diagrams representing social networks,

but is equally suitable for drawing

networks of concepts, as in causal

mapping applications.

This software does not support the initial

elicitation of data for the construction of

cause maps.

Borgatti, Everett, & Freeman

(1992)

A demonstration copy of this

commercially distributed software

package is available at

http://www.analytictech.com

Cognizer

This software

package is

currently being

developed by

Mandrake

Technology

This software enables direct

elicitation of cause maps.

Building on __

_

____

_

                              

Goldberg (1995), data elicitation

and map construction are achieved

by participants first selecting from

a predefined pool of variables

shown on screen (randomised on

an individualized basis, so as to

minimize potential order effects in

variable selection and pairwise

evaluation tasks). Participants

then perform on-screen pairwise

evaluations of their chosen

variables (including explicitly

detailing the strength of

relationship between each pair of

variables). The resulting cause

map can next be immediately

available for viewing and can be

edited by direct manipulation of

the on screen graphical outputs.

Many basic analytical

functions are incorporated,

including, a number of map

content measures (e.g.

indegree, outdegree and

reachability values) and

structural measures (e.g. linkto-

node ratio and map density)

(see Table 1 in the main body

of the chapter for a description

of these measures). Distance

ratios, reflecting the degree of

overall dissimilarity between

pairs of cause maps

(Langfield-Smith & Wirth,

___         __

_

___

_

_

                               _______

_

_            

1995) can be readily computed

and employed in order to

investigate patterns of

similarity and difference

among subgroups of

participants.

Cognizer is essentially being

developed to permit the elicitation of

cause maps in a manner that is

meaningful to participants and

amenable to mass-comparison.

However, the software also allows

cause maps to be elicited in other

ways, each of which may be more or

less acceptable in particular contexts

of application. For example, maps can

be constructed directly, by drawing a

weighted digraph (a

sophisticated

variant of the basic influence diagram,

as discussed in the main body of this

chapter under ‘basic metrics for the

analysis of individual cause maps’).

The results are easily exportable to

other appropriate analytical packages,

such as SPSSTM.

The limitations of the software will depend

upon the chosen method of elicitation and

analysis (see the main body of the chapter

for a

detailed discussion regarding the pros

and cons of each method).

Clarkson, Hodgkinson &

Fearfull (2001)

.

For additional details and/or a

demonstration copy of Cognizer, e-mail:

mandraketech@

fsbusiness.co.uk

worthy of brief mention in this connection, before turning to consider more specialist

software tools specifically devised for the analysis of cause maps, is the general

approach known as Pathfinder (Schvaneveldt, 1990; Schvaneveldt & Durso, 1981;

Schvaneveldt, Dearholt & Durso, 1988, 1989). In a similar vein to UCINET, the Pathfinder

algorithms, as implemented in software systems such as KNOT (The Knowledge Network

Organizing Tool) (http://www.interlinkinc.net/Pathfinder.html), are used to explore network

structures derived from proximity data (i.e., distance matrices reflecting the degree

of overall (dis)similarity, or some other proximity measure, between concepts). Within

the specific domain of information technology, Pathfinder has been successfully applied

to a variety of problems concerning the design of user interfaces (e.g., Gillan, Breedin &

Cooke, 1992; Roske-Hofstrand & Paap, 1986). As observed by Gillan & Schvaneveldt

(1999), in general, applications in this context (typically involving the analysis of

relatedness ratings) have demonstrated that users are more effective in using interfaces

derived from their revealed models of the system, as identified by the Pathfinder

algorithm, in comparison with existing interfaces.

Although software systems such as ATLAS/ti, UCINET and the Pathfinder algorithm are

proving generally useful as basic support mechanisms in the conduct of causal mapping

research, fully integrated software systems, dedicated to the elicitation, construction,

analysis and comparison of causal maps are ultimately required, if causal mapping is to

fulfill its true methodological and substantive potential. To this end, there have been a

number of advances over the past decade or so and in the remainder of this appendix we

highlight what we consider to be the most significant of these. Due to space limitations

we shall confine our attention to a brief consideration of just three of the more popular

software packages presently available for the dedicated analysis of causal maps, namely,

CMAP2 (Laukkanen, 1994), Decision Explorer(Eden et al., 1992) and the suite of

programs developed by Goldberg (1996), known as distrat/askmap, in addition to

reporting some ongoing developments of our own. Clearly, all of these systems are

constrained (albeit to varying degrees) by virtue of the underlying assumptions and

concomitant choices that their developers have made in relation to the various issues

discussed in the main sections of this chapter.

CMAP2 (Laukkanen, 1994, 1998) was developed for the comparative analysis of causal

maps derived through interview transcripts and/or documentary sources. A data-basedorientated

PC program, it is intended specifically for use in settings where the input data

take the form of natural communication and key parameters such as the number of

concepts explored, the number of mapped relationships and indeed the number of

participants must be flexible (Laukkenan, 1998). Unfortunately, as observed by Jenkins

(1998), CMAP2 is limited in several important respects. First, no research has been

undertaken to assess the reliability of the processes by which the input data are

transformed into comparable units of analysis. As noted in our discussion concerning

the relative merits of direct vs. indirect elicitation procedures, this is clearly not a problem

unique to CMAP2 but is common to a number of applications of causal mapping

procedures more generally, where the maps have been inferred from interview transcripts

and/or other indirect documentary sources. Clearly, however, if the practice of causal

mapping and the associated application of particular procedures such as CMAP2 are to

gain credence in terms of their scientific legitimacy, there is an urgent need to increase

the volume and quality of research addressing these and other equally pressing issues

concerning their psychometric efficacy.

Decision Explorer(Eden et al., 1992), a re-launch of Graphics COPE, the system

developed several years earlier by Eden and his colleagues for use in the context of group

decision support (e.g., Ackermann, Eden & Cropper, 1990; Eden & Cropper, 1990), has

proven to be of immense benefit in the context of building comprehensive cognitive maps

of complex organizational problems. Decision Explorerallows the researcher to manipulate

data in ways that enable it to be viewed from a variety of perspectives (Eden &

Ackermann, 1998b). This is helpful not only from an analytical standpoint, but also in

enabling the researcher to actively gain and maintain the interest of participants in the

research process. However, Decision Explorer, as with Laukkanen’s software package,

was designed primarily for use in the context of local settings, where the focus of

attention is on the intensive analysis of ideographic data, gathered from small numbers

of individuals. It is less suitable for use in the context of larger-scale studies.

In contrast, Goldberg’s (1996) computer programs were designed to perform several of

the tasks associated with the Markóczy-Goldberg approach to causal mapping (Markóczy

& Goldberg, 1995). As discussed in the main body of the chapter, this approach is

potentially very useful in situations that demand the comparative analysis of large

numbers of maps. Unfortunately, however, a number of the statistical procedures as

devised and implemented by Markóczy and Goldberg (1995), including those building on

the earlier work of Langfield-Smith and Wirth (1992), have no accompanying software

provision within the distrat/askmap system, thus rendering their implementation difficult,

if not impossible, using these programs. Nevertheless, the fact remains that the

Markóczy-Goldberg approach to causal map elicitation, analysis and comparison – and

the earlier work of Langfield-Smith and Wirth, which laid the foundations for these

innovations – represents a major methodological breakthrough. However, if the ultimate

potential of this approach is to be realized, there is an urgent need for further developments

in the provision of user-friendly software, capable of readily implementing the full

range of associated procedures, from elicitation through analysis to comparison, in realtime

environments. At the time of writing, the present authors are in the advanced stages

of actively evaluating such a system. To date, this Windows-based system, known as

Cognizer, has been successfully implemented in the elicitation, analysis and comparison

of well over 200 maps, all gathered in the context of face-to-face interviews, in situ, with

busy employees. (Further details of all of the individual software systems discussed in

this Appendix, including a summary of their main strengths and limitations, together with

information concerning their availability, are presented in Table A1.)