an approach to assess collaboration readiness

International Journal of Production ResearchVol. 47, No. 17, 1 September 2009, 4711–4735

An approach to assess collaboration readiness

Joao Rosas and Luis M. Camarinha-Matos*

Department of Electrical Engineering, New University of Lisbon, Quinta da Torre, 2829-516Monte Caparica, Portugal

(Revision received April 2009)

The level of readiness of an organisation to join a collaborative process dependson ‘hard’ factors such as competency fitness or technological compatibility, butalso on several other factors of a ‘soft’ nature such as an organisation’s character,willingness to collaborate, or affectivity/empathy relationships. Considering theseaspects, a modelling approach to assess how prepared an enterprise is to join acollaborative network is proposed. The approach is based on the notion of‘character’ of the organisation, addressing behavioural aspects regardingcollaboration. The distinction between collaboration readiness and preparednessis established. In order to deal with the incompleteness of information anduncertainties associated to the readiness assessment process, it is proposed to usebelief networks, which allow performing inference concerning behaviouralcharacteristics of organisations. The approach is then extended in order tohandle decision making under situations characterised by uncertainty. Anexample illustrates how the approach can be integrated in a partner’s suggestionmechanism.

Keywords: collaborative networks; collaboration readiness; collaboration pre-paredness; organisation’s character; virtual organisation breeding environment

1. Introduction

When involved in a collaborative network, entities work together towards the achievementof common or compatible goals. As collaboration goes on, they tend to adopt patterns ofbehaviour according to the situations they are involved in. While some of these patternsare both acceptable and desirable, some others might not be. Naturally, undesirablebehaviour should be avoided, as it affects collaboration and may lead to conflicts. The actof working in collaboration is by itself considered challenging and risky. Often, anorganisation performs well while working alone, but poorly in collaboration. This meansthat before joining networks, organisations should be adequately prepared forcollaboration, in order to be ready to react promptly and grasp business opportunities.

Most works in the past were focused on ‘hard’ factors, such as matching competence ortechnological fit, in order to assess the preparedness of an organisation to join a network orpartnership. However, these factors cannot always be an explanation for the causes ofpartnership failures. More recent research, gives evidence that other aspects of a more ‘soft’nature are recognised as greatly influencing collaboration and, therefore, significantlyinfluencing the partners’ preparedness to collaborate. Such aspects, just to name a few, arethe members’ behaviour, ethical issues, norms, values, and trust.

*Corresponding author. Email: [email protected]

ISSN 0020–7543 print/ISSN 1366–588X online

� 2009 Taylor & Francis

DOI: 10.1080/00207540902847298

http://www.informaworld.com

Therefore, in order to increase the chance of success in collaborative networks, weshould consider both ‘hard’ and ‘soft’ aspects in any approach for the assessment ofpartners’ readiness and preparedness to collaborate. Soft aspects are more challenging anddifficult to model. How important they are, how deeply they influence the success ofcollaborative networks, and which level of influence they have in collaboration readinessand preparedness, are aspects that need to be further investigated.

This research proposes an approach to perform assessment of the collaborationreadiness of members or candidates to collaborative networks. This assessment is mostlybased on the concept of an organisation’s character, as described in Section 2, in which aclear distinction between the concepts of collaboration readiness and collaborationpreparedness is also established. A Bayesian belief network modelling approach is used tomake predictions on collaboration preparedness based on an organisation’s character, asdescribed in Section 3. The approach is then extended into a decision network, in order toallow decision making. An example illustrates how the collaboration readiness concept canbe used in the implementation of a partners’ suggestion mechanism. Finally, relatedresearch concerning partners’ search and selection is presented in Section 4. This article is arevised and extended version of a preliminary work first presented at the BASYS’08conference (Rosas and Camarinha-Matos 2008a).

2. Collaboration readiness assessment

2.1 The organisation’s character concept

Organisations inside networks work and interact with each other towards the achievementof common or compatible goals. They typically manifest a variety of behaviour, accordingto the peers and situations they are involved in. In this sense, behaviour can be understoodas anything that an organisation does involving pro-active actions and responses toexternal events/requests.

The behaviour of organisations typically tends to show some repetition through time.This repetition usually leads to the formation of behavioural patterns. These patterns canbe associated to a set of identifiable traits. A trait represents a relatively stablepredisposition to act in a certain way or, in other words, the preponderance for theoccurrence of a certain behavioural pattern. These traits, together, form what is referred toas character. An organisation’s character can therefore be seen as a composition of a set oftraits that determine the behaviour or nature of the organisation. This underlying mappingbetween character traits and behaviour can be used to perform behaviour prediction. Thismeans, in turn, that collaboration readiness assessments can be performed using theconcept of an organisation’s character. Basically if the predictable behaviour is consideredpositive towards collaboration, then the readiness increases, otherwise it decreases. This isthe approach suggested in this paper for collaboration readiness assessment. It shall benoted that the intrinsic connection between character traits and behaviour hastraditionally been an extensive research topic in psychology, as expressed in Goldie(2004) and Webber (2006).

These definitions of behaviour and traits pertain not only to individuals, as they in facttranscend the individual level. Sandelands and Stablein (1987) raised the possibility thatorganisations are mental entities capable of thought and that they could existindependently of particular individuals (Walsh and Ungson 1997). The idea of anorganisation having a character is not a completely new concept. For instance, in

4712 J. Rosas and L.M. Camarinha-Matos

Gothlich (2003) a model for collaborative business ecosystems is presented taking somemetaphors from biology, in which the behavioural patterns are described through a smallnumber of classification traits, namely resilience and responsiveness. In Wilkinson et al.(2005) an analogy is made between the idea of mating and sexual appeal and the idea ofbusiness mating. They describe matching factors for engaging in long term partnerships.These factors were grouped in financial issues, organisational (and strategic issues), andtechnological issues. In Chun (2005) a ‘virtuous ethical character’ scale, composed of sixdimensions (integrity, empathy, courage, warmth, zeal, conscientiousness) and 24 items, isdescribed to enable an assessment of the link between organisational level virtue and anorganisation’s financial or non-financial performance.

2.2 Elements for a collaboration readiness assessment model

The word readiness, according to the Oxford Dictionary of English (Oxford Dictionary ofEnglish 2003), refers: (1) to the state of having been made ready or prepared forsomething; (2) the willingness to do something; and (3) the quality of being immediate,quick and prompt. Following this definition, an organisation could be considered ready tocollaborate if it is prepared and willing to work in collaboration for the achievement ofcommon goals, performing tasks in an accurate and reliable way. This readiness conceptshould cover several aspects, ranging from technological and economical to behaviouraland social ones. In this research, however, the emphasis is put mainly on aspects related tothe organisation’s behaviour. Since traits represent predispositions to act in a certain way,an organisation can be considered prepared to collaborate if its character traits showvalues that favour the predisposition of occurrence of behaviour that is desirable in acollaboration context.

In this section a number of support concepts are defined in order to better understandthe context and suggested approach. In the notation used it is assumed that all singleattributes are named in small letters, while sets are named in capital letters. At the base, letus consider the sets listed in Table 1.

Table 1. Basic sets.

Set description Example

O¼ {o1, o2, . . .} – the set of organisations of a virtualorganisation breeding environment (VBE) (Afsarmaneshand Camarinha-Matos 2005).

{net1, org2, university3, . . .}

T¼ {t1, t2, . . .} – the set of trait identifiers that can be used tocharacterise an organisation’s character

{flexibility, creativity,reliability, . . .}

E¼ {e1, e2, . . .} – the set of empathy, affectivity or attitudesassumed by one organisation towards others

{trusts, distrusts, respects, relies,dislikes, . . .}

Vi¼ {vi,1, vi,2, . . .} – the set of values that trait ti can assume {low, fair, high};

OP¼ {op1, op2, . . .} – the set of comparison operators. Theoperator opi performs comparisons between the values ofthe set Vi (e.g., ‘near(v1,v2)’)

{‘5’, ’4’, ‘¼’, about, near, relia-bility_op, prestige_op, . . .}

C¼ {c1, c2, . . .} – the set of competences required for theachievement of a collaboration opportunity.

{DBA, logistics, ICT, CAD, . . .}

International Journal of Production Research 4713

Definition 1: Organisation’s behaviour – the way in which an organisation acts orconducts itself and toward others; the way it behaves in response to a particular event orsituation.

For instance, each of the negative or non-beneficial behaviour listed in Table 2 wasidentified as having occurred in inter-organisational relationships (Muskin 2000).

A monitoring approach based on a collaboration readiness concept model will beuseful if it triggers before the occurrence of any of these or other similar behaviour. In theevent that any of this behaviour actually manifested, what is required is, not assessments,but prompt remedy measures.

Nevertheless, it is not likely that the behaviour shown in Table 2 will become patternsdeveloped by members of a virtual organisation breeding environment (VBE), as theywould soon be kicked out of the strategic alliance. By contrast, it is beneficial in terms ofcollaboration the observation of patterns, such as:

. Strong effort put in undertaking assigned ‘business’ processes.

. Adhesion to established governance rules.

. Following agreed strategies and protocols.

. Sharing assets and exchanging knowledge.

. Promoting a team-spirit between participating organisations.

Definition 2: Organisation’s behavioural patterns – the regularities of behaviour that isobservable or discernible in the behaviour of an organisation.

Table 2. Non-beneficial inter-organisational behaviour.

Behaviour Short description

Conflict of interest Incompatible affiliation between organisationswhich has the potential of causing anunmerited flow of benefits.

Bribery Offering something, which causes unmeritedbenefits.

Purposefully misleading or false statements What is beneficial to one organisation, butresulting in a business behaviour that isharmful to another organisation.

Appropriation of intangibles Any unauthorised taking of ideas, informa-tion, design, processes, secrets or otherintangibles belonging to an originally pos-sessing party.

Non-performance of agreements Expectations and commitments not met by apartner, without an agreement of substituteprovisions.

Commitments beyond ability to perform Failing to perform on acceptable performancestandards; irresponsible announcement of‘technological’ capabilities.

Exploitation of relative power An organisation uses its position to inducebehaviour contrary to the reasonable inter-ests of a party exposed to this power.

Favouritism Activity carried by one organisation, in resultof some ‘non-business’ relationship, andwhich favours one organisation overanother of greater merit.


Definition 3: Organisation’s character – an organisation’s character can be seen as acomposition of a set of traits that determine the way it behaves. It can be modelled as atuple OC¼ (o, TV), in which:

. o – identifies the organisation being characterised;

. TV¼ {(ti, vi,k) j ti2T, vi,k2Vi} – is the trait set constituted of tuples, each onecomposed of a trait and a corresponding trait value.

Examples of possible organisation’s character traits could be organisation’s agility,flexibility, openness, creativity, reliability, and honesty. The identification of a minimumset of traits that best represent a character regarding collaboration between organisations,is an open research issue.

As an example, the character of a hypothetical organisation org_1, using the abovedefinition, could be specified as the tuple (org_1, {(reliability, high), (creativity, fair),(honesty, high}).

Assessing how prepared an organisation is to collaborate, using the concept of anorganisation’s character, requires not a deterministic assessment of an organisation’s traitvalues, but instead a verification of whether these traits meet adequate preparednesscriteria.

Definition 4: Character-related preparedness conditions – the preparedness conditionsrelated to the organisation’s character are represented by a set CP of preparedness items.Each item is a tuple that specifies the condition or value that is required for a givencharacter trait of an organisation. The preparedness conditions’ set is formally defined as:

CP ¼ ðti, vi, k, opi, piÞjti 2 T, vi, k 2 Vi, pi 2 ½0, 1�, opi 2 OP� �

,

in which:

. ti – is the trait name;

. vi,k – is the trait value, such that vi,k2Vi;

. pi – expresses the desired probability/likelihood of the trait ti having the value vi,k;

. opi – is the operator used for comparing the values of probability pi.

As an example, a preparedness pattern would be represented by the following setCP¼ {(reliability, high, ‘4¼’, 0.7), (creativity, fair, ‘about’, 0.8)}.

Definition 5: Preparedness for collaboration – an organisation is considered prepared tocollaborate if its character (Definition 3) can satisfy a set of preparedness conditions(Definition 4).

Definition 6: Willingness to collaborate – an organisation is willing to collaboratewhenever it perceives that (a subset of) its interests can be better satisfied in collaborationwith other organisations than operating in isolation.

These interests can include, for instance, access to new markets, access to resources,complementing its competence and skills, sharing market risks, or increasing its ownbenefits. Sometimes this willingness can be negative; for instance, whenever anorganisation feels uneasy or when it perceives important concerns in the virtualorganisation (VO) or in the collaboration opportunity (CO) achievement (e.g., when itdoes not believe that the CO will provide the expected benefits). An academic or researchinstitution might be interested in a collaboration opportunity for the purposes of


knowledge creation, patent granting, or receiving royalties, but might not be willing toactively participate in the commercial results exploitation phase.

Definition 7: Competence fitness – an organisation fits, competency-wise, in somecollaboration scenario if it possesses adequate (or required) competence.

The competence adequacy depends on whether the context is either a VBE (bringingcompetence that fits the general scope of the VBE) or a virtual organisation (VO)(providing or complementing required competence for the achievement of the VO goals)(Camarinha-Matos and Afsarmanesh 2006).

Definition 8: Affectivity/empathy relationships – it is a set composed of elements thatspecify empathic relationships between organisations. It is formally specified as A¼ {(oi, oj,el, level) j oi, oj2O, el2E, level2 [�1,1],}. Each tuple represents a ‘feeling’ between oneorganisation oi, and a peer oj. The level parameter specifies the intensity of the feeling.Empathy relationships can be negative (e.g., when an organisation distrusts another).

Definition 9: Collaboration readiness – a concept that combines the organisation’spreparedness (Definition 5), willingness to collaborate given a concrete CO (Definition 6),competency fitness (Definition 7), and the affective/empathic relationships (Definition 8)between this organisation and the other entities participating in the CO.

Unlike preparedness, the concept of readiness is applied to a specific collaborationopportunity and typically defined for a short time window. Preparedness, on the otherhand, is more long-term oriented. Specific cases of readiness can be defined, e.g., readinessto join a VBE, readiness to join a VO.

Past research has put considerable effort in the area of ‘competence fitness’ (e.g.,matching algorithms for partner selection). However, the other elements of Figure 1 havereceived little attention so far, and yet they are quite important for the success of acollaboration process. In the remainder of this paper we will focus on the issue ofcharacter’s related preparedness. The other aspects are the subject of future research. Forthis purpose, let us make the following assumptions:

Assumptions: – organisation’s behaviour predictability

As mentioned earlier, an organisation performs actions or shows behaviour that tends torepeat through time, leading to the formation of behavioural patterns. These patterns can

Figure 1. Collaboration readiness concept (Rosas and Camarinha-Matos 2008a).


be associated to a set of identifiable traits. Given the underlying correspondence between

traits and behaviour, then the organisation’s character can be used in behaviour

prediction. In other words, the character of an organisation can be used to perform

collaboration preparedness assessment, in the following way:

. If the predicted patterns are seen as favourable to collaboration, then the

collaboration preparedness increases.. If these patterns are mostly positive, then in terms of its character the organisation

is considered prepared to collaborate.. On the opposite side, if these patterns are considered undesirable or unfavourable

to collaboration, then the collaboration preparedness decreases.. If these patterns are mostly negative, then in terms of its character the

organisation is considered not prepared to collaborate.

With the above definitions together with these assumptions, it is possible to formulate

the following two axioms of collaboration preparedness. However, before stating the first

axiom, it is necessary to introduce the concept of belief in a notation that is useful for the

approach being described.

Definition 10: Belief – a function that receives a query and known evidence, and returns

the probability of the query being true. It can be abstractly specified as:

belief : query � evidence! ½0, 1�,

in which:

. query2 fðti, vÞjti 2 T, v 2 Vig;

. evidence2 2fðti, vÞjti2T, v2Vig.

As an example, these two parameters could be instantiated as: query¼ (reliability, high)

and evidence¼ {(prestige, neutral), (creativity, high)} (we recall here the concept of

power set1, from set theory).

Axiom 1: An organisation org is prepared according to a given set of character-related

preparedness conditions PC if for each preparedness item p2PC, there is a corresponding

belief b, such that org’s character complies with this preparedness item p:

8org8PCðis preparedðorg,PCÞ 8p9bððbelongsðp,PCÞ

^ beliefðorg, p, bÞÞ ! compliesðp, bÞÞÞ

In this axiom, ‘belief ’ takes the notation of a predicate, being its functionality

supported by a belief network, which is described in Section 3. It is used to obtain the

probability or likelihood of a given trait having a value b. The ‘complies’ predicate verifies

whether this likelihood meets or does not meet the condition p specified in PC.

Axiom 2: A VO satisfies a given set of preparedness conditions PC if all its members are

prepared according to PC:

8VO8PCðpreparednessðVO,PCÞ 8orgðbelongsðorg,VOÞ ! is preparedðorg,PCÞÞÞ

It shall be noted that often there is not enough information to perceive and characterise

an organisation’s character. This fact results in traits that might be unknown, specified

with imprecision or subjectively. This lack of knowledge increases the uncertainty


regarding behaviour’s prediction and, consequently, limits the collaboration preparedness

assessment. Therefore, the output of the assessment process should be of a probabilistic

nature. These aspects are illustrated by an application example in Section 3.4.

3. An illustrative modelling experiment

3.1 Belief networks basics

A Bayesian belief network is a kind of probabilistic model that represents causal

relationships on a set of variables (Figure 2). It is composed of two parts: the structural

part, which consists of a direct acyclic graph, in which nodes stand for random variables

and edges for direct conditional dependence between them; and the probabilistic part that

quantifies the conditional dependence between these variables. Each variable can have

state values (such as, ‘no’, ‘yes’ or ‘low’, ‘high’). If the value of a variable in a node is

known, then that node is said to be an evidence node. More on belief networks can be

found in Jensen (1996). In Figure 2, the arc pointing from node C to node E, for instance,

can be perceived as C causing or influencing E. Each of the child nodes has a conditional

probability table that quantifies the effects that the parents have on them. For the nodes

without parents, the corresponding table only contains prior probabilities. Due to these

conditional dependencies, if a node becomes an evidence node, then the probabilities

(or likelihood) of the other nodes change.For any node of the network, the computation of conditional probabilities is done

using the Bayes’ rule, exemplified in the next section. For the above example, the

probability of variable E being in state yes or no is conditioned by its parent C being in

state low or high and its parent D in state left or right.Belief networks can be used to perform queries in distinct ways:

. To perform predictions. This is useful whenever some causes are known and it is

necessary to determine the probability of possible effects/consequences. For

instance, when B¼ low and C¼ high, the probability of E¼ yes is given by the

query P(E¼ yes jB¼ low, C¼ high).. To perform diagnostics. For instance, when the fact F¼bad is known, it is

necessary to determine the likelihood of eventual causes: P(A¼ yes jF¼ bad).. It is also possible to make queries on the joint distributions, without providing

evidences. For instance, the probability of F¼ fair, without further evidence, is

given by P(F¼ fair).

Figure 2. An example of a Bayesian belief network.


3.2 Performing decision making

As just described, a belief network is used to reason under uncertainty and uniquely

represents a probability distribution for computing any probability of interest. It models

the behaviour of a part of the world worthy of interest and can be used to perform

behaviour prediction. What we do with these predictions, however, is another matter that

goes into the realm of decision theory (Charniak 1991).A belief network does not provide, by itself, a mechanism for decision making.

A typical decision making process characterised by uncertainty involves, as illustrated in

Table 3, the alternative decisions at a decision maker’s disposal, the characterisation of the

situations in which these decisions must be taken (usually characterised as states of the

‘world’), and the decisions’ uncertain consequences or outcomes (e.g., benefits, costs,

rewards, penalties, joy, regret, . . .).Belief networks can, nevertheless, be extended to handle decision making and, in this

case they take the form of decision networks or influence diagrams. Decision networks

offer formalism for capturing the various types of knowledge involved in a decision

problem and offer methodologies for computing preferred decisions (Renooij and Gaag

1998). Though, a decision network allows for encoding not only a probability distribution

on a set of variables, as belief networks do, but also the decisions that a decision maker can

make and the desirability of their (uncertain) consequences.As illustrated in Figure 3, a decision network is built from a belief network with the

addition of two additional types of nodes, the decision and the utility nodes. The nodes of

the decision’s belief network side are now named as nature nodes. A decision node is

usually drawn as a rectangle and models a decision variable that represents the various

decision alternatives at a decision maker’s disposal. A value node is usually drawn as a

Figure 3. Example of a decision network.

Table 3. Example of information involved in a decision-making problem.

States

Decisions s1 s2 . . . smd1 u1,1 u1,2 . . . u1,md2 u2,1 u2,2 . . . u2,m. . . . . . . . . . . . . . .dn un,1 un,2 . . . un,mProbability P(s1) P(s2) . . . P(sm)


hexagon and represents the desirability of the consequences that may arise from thevarious decisions made under the current situation. Arrows into decision nodesrepresent the information that is available at the time a decision is made. Arrows intovalue nodes collectively represent the influence of the parent nodes in the desirability value(Szolovits 1995).

The process of obtaining a decision solution for a decision network corresponds toselecting, for each node, the optimal choice given any possible set of informational inputsto the decision node. This is performed with the principle of maximal expected utility, inwhich a rational choice is made whenever the decision that yields the maximal expectedutility is selected (Horvitz et al. 1988, Boutilier 2003).

The decision network in Figure 3 could be inside a guiding system of a vehicle, whichfrom the given sensory information (encoded in the belief network’s nature nodes) and thedesirability values for hypothetical alternative routes (encoded in the value nodes) suggestswhich way to turn, left or right (the choices in the decision node). The feature ‘F’represents the information that is available whenever a decision is taken by the system. Thenumber at the right hand side of each decision alternative in the decision node representsthe expected utility of making that choice. In this example, if no evidence is given, therational decision would be to turn left, which is the action with highest desirability.

3.3 Modelling the predictor

In simple situations a Bayesian network can be specified by an expert and then used toperform inferences, as illustrated in Figure 4 (phases 1 and 2). In many cases this task istoo complex to be done by hand. Alternatively, both the structure (nodes and arcs) andparameters of the local distributions can be learned from historic data, using machinelearning techniques (Pearl 1996, Cheng et al. 1997, Friedman 1997, Cheng and Greiner2001).

In order to guide the belief network design process for this experiment, we made a fewassumptions related to members’ behaviour, among potentially many others, which shouldbe taken as merely illustrative. Therefore, for building the modelling example, thefollowing conjectures were considered:

c1. An organisation of fragile economical condition, in order to benefit from others’competence (that usually it cannot afford to own), is more willing to accept the risks

Figure 4. Belief network modelling and utilisation (Rosas and Camarinha-Matos 2008a).


of collaboration. On the other hand, due to its fragile condition, it tends to be less

reliable.c2. An organisation in good economical condition might be more reliable, but does not

feel the same pressure, as the previous case, to collaborate and therefore tends to be

more risk conservative considering collaboration/partnerships.c3. A small size organisation (e.g., an SME) might possess less competence and, with

the goal of complementing it, accepts to be more exposed to the risks of

collaborating with other organisations.c4. The prestige of an organisation, which is an attribute that is perceived by its peers,

is fundamental in collaboration and adds directly to the preparedness level.c5. An organisation characterised by poor reliability has a downgrade of its prestige.c6. The creativity of an organisation, which can be roughly estimated by evaluating its

rate of generated innovations, might also be important for collaboration, and adds

directly to the preparedness level.c7. Higher reliability adds to preparedness; higher tolerance to the risk (of being in

collaboration) also adds to preparedness.

Certainly, these conjectures are arguable, but they are considered here only for the

elaboration of an illustration. An example belief network, modelled using the above

guidelines, for the inference of the organisation’s preparedness level is shown in Figure 5,

using NeticaTM (Norsys 1997). The causal links are labelled with previously specified

conjecture(s), justifying the causality between the random variables, which in this case are

taken as the organisations’ traits.For this belief network, the joint probability distribution, from which the predictions

and diagnostics can be made, is the following (showing only the initials for the nodes

names):

P(PD,ES,RP,R,C,P,PL) = P(PD) × P(ES|PD) × P(RP|PD,ES) × P(R|PD,ES,RP) × P(C|PD,

ES,RP,R) × P(P| PD,ES,RP,R,C) × P(PL|PD,ES,RP,R,C,P)

This function can be simplified by considering the conditional independence statements

implied in the belief network. For instance, the ‘partner dimension’ variable does not

directly influence the ‘preparedness level’, as ‘reliability’ does. This is because

Figure 5. A Bayesian network example to assess the preparedness level.


P(PLjPD,R)¼P(PLjR), so PD can be removed from the above expression. In other words,

PL and PD are conditionally independent given R. The same approach can be applied to

the other conditional probabilities, which helps removing more variables (the shaded ones)

from the above expression. This results in the expression:

PðPD,ES,RP,R,C, P, PLÞ ¼ PðPDÞ � PðESÞ � PðRPjPD,ESÞ � PðRjESÞ

�PðCÞ �PðPjRÞ �PðPLjRP,R,C, PÞ

As an illustration for the given problem, and assuming most of the nodes as evidence

(to reduce calculations), the probability of preparedness level PL¼high, given that

PD¼ high, ES¼ fair, C¼ high, and P¼high is given by:

PðPLhighjPDhigh, ESfair, Chigh, PhighÞ ¼PðPLhigh, PDhigh, ESfair, Chigh, PhighÞ

PðPDhigh, ESfair, Chigh, PhighÞ

¼ 0:928:

After the belief network description, it is now possible to give more explanations about

the behaviour of the ‘belief ’ predicate, used in Axiom 1. This predicate, through the belief

network, provides the likelihood that, for a given character, the trait ti specified in

preparedness item p (Definition 4) has the value vi,k also specified in that item. As an

illustration, let us consider the preparedness item t¼ (reliability, high, ’4’, 70) and observe

the VBE example in Table 4. This predicate would provide the values for belief b (see

Axiom 1 for ‘b’), using the belief network, as illustrated by the following cases:

. For organisation o_1, the belief that reliability¼high is b¼ 100%, because o_1

has the trait ‘reliability’ defined with value high in its character profile. It would

be represented by an evidence node in the belief network of Figure 5.. For organisation o_3, the belief that reliability¼ high is b¼ 0%, because o_3 has

low reliability in its character profile. It would be represented by an evidence node

in the belief network of Figure 5, but with different evidence (low reliability).. For organisation o_2, the belief is b¼ 53.6%. This is because, the reliability of this

organisation is unknown and, therefore, this value is obtained using the query

b¼P(‘reliability¼ high’ j known_traits(o_2)) on the belief network of Figure 5.

Table 4. Competence and traits of the VBE’s members.

Organisation traits

Organisation Competence PD ES RP R C D

o_1 c1, c2 high good ? high high higho_2 c4, c6 med ? high ? low higho_3 c2, c5 med fair high low high higho_4 c1, c2 ? good high low ? ?o_5 c1, c3, c4 high bad high high high lowo_6 c2, c3 low good high ? high high

Note: PD: partners dimension; ES: economical situation; RP: risk profile; R: reliability; C: creativity;P: prestige; ?: not known.


The predicate ‘known_traits(org)’, provides the known values of an organisation’s

traits.

3.4 An example

The example described below illustrates the estimation of collaboration preparedness

based on the organisations’ character, which, as mentioned before, is one of the aspects

considered in the readiness assessment approach being researched.Let us consider the existence of a virtual organisation breeding environment composed

of a group of organisations. These organisations, together with their corresponding

competence and character traits, are defined as shown in Table 4. For illustrative purposes,

the traits used in this example are the ones defined in the belief network of Figure 5,

Section 3.3. Aspects related to the orthogonally of these traits are yet to be considered in

future research. As illustrated in Table 4, one important aspect to emphasise here is that,

for the given organisations, the traits values are qualitative, in the sense that these values

might not be precise, and can even incorporate some subjectivity. As observed in the table,

some traits are even unknown. These aspects add uncertainty to the problem.Figure 6 illustrates two distinct cases of joining a collaborative network. In the first

case, organisation o_12 is a candidate to join the VBE. In the second case, organisation

o_6, already a member of the VBE, is being considered to join a virtual organisation (VO),

namely vo_1.As a newcomer, little information is known about o_12’s character. The only known

evidence about this candidate is that it is in good economical situation and is of low

dimension. We can query the belief network about the probability of this organisation

showing a high preparedness level, using the conditional probability:

Pð‘preparedness level’ ¼ high j‘partner dimension’ ¼ low , ‘economical situation’

¼ goodÞ ¼ 60:1%:

Figure 6. A virtual organisation breeding environment with an existing VO (? states the lack ofinformation concerning 0_12’s character traits).


If we want to assess whether the virtual organisation vo_1 is composed of members that

are prepared to collaborate, we can define some preparedness conditions, supposedly

adequate for a given situation or context, and run the predicate ‘preparedness’ specified in

Axiom 2. We would invoke the following query:

‘preparednessðvo 1, fðreliability, high, ‘ >¼ ’, 70Þ, ðcreativity, fair, ‘about’ , 80)g)’:

In this case vo_1 is not prepared according to the specified preparedness conditions,

because organisation o_2 does not comply with the preparedness conditions. This

organisation has reliability P(reliability¼high, know_traits(o_2))¼ 0.63, which is less than

0.7, as specified in the conditions. It also fails in terms of creativity, because

P(creativity¼ fair, know_traits(o_2))¼ 0. In other words, its creativity level is low andthe conditions of the query require it to be fair.

3.5 Deciding who to ‘invite’

Previous sections showed how a methodology for collaboration preparedness could be

modelled by means of a belief network, for which its predictive capacity could be used topartially validate the preparedness assumptions formulated in Section 2. Now it is time to

illustrate how this methodology can be used for decision making in a context of

collaborative networks. This is done by formulating a simple decision problem over the

belief network.In a given situation, a broker is responsible for selecting a set of organisations as

candidates for a new collaboration opportunity. These candidates are relatively unknown

to the broker, so he has little information characterising these organisations. Nevertheless,

the broker could obtain an idea of the prestige of each organisation after enquiring of

some partners he already knows. Moreover, he can also characterise each organisation interms of their size. Beyond the prestige, he is also very concerned with the eventual

reliability of selected candidates, for which he could not obtain concrete information.

To summarise, this broker has a decision problem with two alternative choices – invite or

not invite – and the decision has to be based on two uncertain organisations’

characteristics, prestige and reliability.Due to the fact that we already have a belief network that models the behaviour of

organisations from their character traits, the approach is now to extend this belief network

into a decision network, so that it can also handle decision making, as described in a

previous section. For this decision problem, it is sufficient to have a decision network withone additional decision node and one value node. In the decision node, the alternative

choices of the decision, invite or not invite, are specified. The value node specifies which

values of prestige and reliability are more desirable for each alternative decision, according

to the broker’s personal scale of judgment, as illustrated in Table 5.Although, contrary to the belief network component, this part can be quite subjective,

because it depends on the broker’s personal preferences. The required decision network for

the ‘invitation’ decision problem is presented in Figure 7, separated in behavioural and

decision parts.Looking at the decision network, we can see that without any evidence of candidate’s

character traits, the expected value of inviting a candidate is 45.6 and that for not inviting

is 41.2; thus, in this situation, the choice would be to invite. In order to understand how

this desirability value of 45.6 is obtained, let us organise the information in a typical


decision matrix as shown in Table 6. This table represents a snapshot for the current

decision problem presented in Figure 7, in which no evidence is provided to the belief

network. The values of desirability, given the traits reliability and prestige, are taken from

the value node.The values of probability in each column are determined using the concept of joint

probability distribution of our belief network, as previously explained, and assigning the

values of the traits present in each of these columns. For the column with the state (H, H),

that probability is:

PðRhigh, PhighÞ ¼ PðPD,ES,RP,Rhigh, C, Phigh, PLÞ ¼ � � � ¼ 0:385:

The remaining probabilities are obtained in a similar way. In order to proceed, let us

consider the sets D¼ {yes, no} and S¼ {(H,H), (H,N), (H,L), (L,H), (L,N), (L,L)}.

The determination of the expected utility of each alternative choice, for the specified

Table 5. Values of desirability for character traits and decision’s choice.

Utility node

Reliability Prestige Invite? Desirability

High High Yes 100High High No 0High Neutral Yes 50High Neutral No 10High Low Yes 10High Low No 50Low High Yes 10Low High No 50Low Neutral Yes 5Low Neutral No 60Low Low Yes 0Low Low No 100

Figure 7. A decision network for VO partners’ suggestion.


decision variable, requires the concept of expected utility function (Renooij and Gaag

1998), which is formulated as:

EUðdiÞ ¼P

sj2S

Uðdi, sjÞ � PðsjjdiÞ, di 2 D,

sj 2 S:

In this equation, U(d,s) represents a relation U :D � S desirability, which for a given

state and decision choice, provides a value of desirability. For our decision problem, these

values are specified inside the value node of our decision network in Figure 7, also

enumerated in Table 5. P(sj j di) yields the same probability as P(sj), because no arrows go

from the decision to the causal side of the decision network. Proceeding this way, the

expected values of utility for our decision choices are:

EUðyesÞ ¼ 100� 0:385þ 50� 0:110þ 10� 0:055þ 10� 0:045þ 5� 0:135þ 0� 0:270

¼ 45:67

EUðnoÞ ¼ 0� 0:385þ 10� 0:110þ 50� 0:055þ 50� 0:045þ 60� 0:135þ 100� 0:270

¼ 41:20,

which are the same as the values of desirability expressed in the decision node of Figure 7.

Under the considered circumstances, the option that is chosen is the one with maximal

expected utility.The values in Table 6 are a snapshot of a decision situation and do not allow for

interactive querying, as decision networks do. To illustrate this idea, let us see how our

model works when modelling a ‘gossip’. Let us imagine that a broker’s friend tells him

that he is aware that one of the candidates in the shortlist is not in an adequate

economical situation, which could cause difficulties in his regular ‘business’ activities.

The broker, concerned with the impacts that his friend’s tip could have on his decision,

adds this new evidence in the corresponding node of the belief network, as illustrated in

Figure 8.Moreover, the broker is aware that now his previous outlook on the candidate’s

prestige should not weight the same as before, so he stays neutral on prestige. This is

because, as established in the causal links of the belief network, a problematic economical

situation can prevent an organisation from developing a reliable behaviour. Less

reliability, in turn, is not healthy for an organisations’ prestige.With this new evidence, we have EU(‘yes’)5EU(‘no’), as shown in Figure 8, stating

that now the rational choice is to take this candidate out of the broker’s shortlist.

Table 6. A snapshot for the decision instance in Figure 7.

Traits: (reliability, prestige)

Invitation? (H,H) (H,N) (H,L) (L,H) (L,N) (L,L)

Yes 100 50 10 10 5 0No 0 10 50 50 60 100Probability 38.5% 11.0% 5.5% 4.5% 13.5% 27.0%

Note: H: high, N: neutral, L: low.


3.6 Use in a partners’ suggestion mechanism

In order to illustrate a practical utilisation of this character based collaboration

preparedness model, a partners’ suggestion mechanism was modelled. This mechanism

implements the concept of competence fitness (Definition 7) in order to select a set ofcandidates for a given collaboration opportunity. In this sense, and according to Figure 1,

the readiness concept is partially illustrated by its components of preparedness and

competence fitness. The mechanism also integrates simulation, and project managementconcepts, for which given the collaboration opportunity’s business process and suggested

VOs, performs simulation in order to provide performance measures.This framework was implemented using a rule-based knowledge base, developed in

Prolog, which was used to develop the preparedness axioms described in Section 2.The belief network was integrated in the suggestion mechanism, by the use of the

NETICATM API.The way the partners’ suggestion mechanism works is illustrated in Figure 9.

The business process needed to satisfy the CO and preparedness conditions are provided as

inputs. Then the partners’ suggestion functionality selects sets of candidates according

to the competence fitness concept, as described in Rosas and Camarinha-Matos (2008b).Using only the competence fitness concept for generating suggestions yields a significant

Figure 8. The decision network with additional evidence.

Figure 9. Suggested partners’ suggestion mechanism (Rosas and Camarinha-Matos 2008b).


number of solutions. However, considering the preparedness concept (Definition 5)

integrated in the suggestion process, the mechanism refines the suggestions to only select

candidates that appear to be more prepared to work in collaboration, according to the

specified preparedness conditions (Definition 4). Finally, the VO and CO’s business

process are given to the simulation module. Additional details of this partners’ suggestion

mechanism can be found in Rosas and Camarinha-Matos (2008b).The corresponding axioms for the partners’ suggestion mechanism, as specified in

Section 2 and in Rosas and Camarinha-Matos (2008b), were translated into Prolog

predicates, as illustrated in Figure 10.These axioms can be invoked using the query below. The shaded argument represents

the preparedness conditions required for the suggested organisations. The characters and

competence of organisations, represented in Table 4, are modelled as facts in the memory

of Prolog’s inference engine.

?- suggest_vo(co_1,{(creativity,high,’>’,60), (preparedness_level,high,’>’,70)}.

The label ‘co_1’ represents the business process plan for the collaboration opportunity

that was given as input for the partners’ suggestion mechanism (Figure 11). The ‘co_1’

plan’s activities are specified in a PERT-like approach (Martinich 1997).The initial VO suggestions, as shown in Table 7, are based on a simple competences’

matching approach, according to Axiom 2. Each line in this table represents a

Figure 10. Prolog predicates for the partners’ suggestion axioms of Section 2.

Figure 11. Example of a business process plan for a given collaboration opportunity.


VO suggestion. For instance, solution 1 represents a VO composed of the members in the

set {o1, o2, o3, o5}. For each suggestion, the simulation module provides the duration of

the simulated business process plan, helping spot the best suggestions. In order to restrict

the number of provided suggestions, it is imposed that each member can be assigned to

only a single competence otherwise the number of suggestions would be unnecessarily

large for our illustration purposes.In this solution, we did not consider any preparedness conditions. Therefore, some

suggestions may in fact be composed of members with low reliability and the VO might fail

in achieving its goals. On the other hand, as shown in Figure 12, if we now provide

desirable preparedness conditions to the suggestion mechanism (see Definition 4 and

Axiom 1), the suggestions would be those in Table 8. As the preparedness conditions

restrict the number of suggestions, each partner can now be assigned with more than one

competence.For this case, the mechanism selected only organisations with both high reliability and

prestige. Organisations with these traits undefined are also selected, provided that the

likelihood of having a high value is at least 30% and 50%, respectively. As mentioned in a

previous section, this likelihood is determined by the predicate belief of Axiom 1, using the

belief network of Figure 5.

Table 7. Example of VO suggestions.

Solution o1 o2 o3 o4 o5 o6 o7 Duration

1 c1 c4 c2 c3 382 c1 c4 c2 c3 393 c1 c4 c3 c2 394 c2 c4 c1 c3 405 c4 c2 c1 c3 406 c4 c1 c3 c2 417 c1 c4 c2 c3 388 c1 c4 c2 c3 399 c1 c2 c4 c3 3810 c1 c2 c4 c3 3911 c2 c4 c1 c3 4012 c4 c2 c1 c3 4013 c2 c1 c4 c3 4014 c2 c1 c4 c3 4015 c2 c4 c1 c3 4016 c4 c2 c1 c3 4017 c4 c2 c1 c3 41

Figure 12. Suggestions influenced by preparedness conditions.


4. Related research

There have been plenty of research methods on partners’ selection for collaborative

networks. Initial approaches were mainly focused on partners’ skills, capacities, and

resources to be shared. For instance, Gupta and Nagi (1995), developed a decision support

system for the selection of manufacturing partners in which the selection criteria consider

a variety of partners’ attributes, including quality, price, delivery, production, manage-

ment, and other enterprise services (e.g., packaging ability, R & D). The selection is

performed by a genetic-algorithmic search approach, providing nearly optimal groups of

manufacturing partners.Another example is found in Mowery et al. (1998), which relies on a resource-based

view of enterprises emphasising the role of partners’ technological capabilities. Patent

citation data and statistics were used as measures of enterprise-specific technological

capabilities, which were then used to determine technological overlap between these firms,

the extent to which their technological resources draw from the same pools of

technological knowledge. Partners’ selection is derived from the measurement of these

overlaps. Chu et al. (2000) established the premise that partners’ selection is strongly

coupled to the product design process for a new product or business opportunity. As such,

they propose a group technology-based approach for partners’ selection, which is mostly

driven by ‘hard’ factors such as cost, time, quality, and enterprises’ financial stability.There are many agent-based approaches for partners’ selection in collaborative

networks. In Petersen and Divitini (2002), an agent-based approach is described in which

partners are represented by agents that compete to become partners of a VE. The

partners’ selection problem is addressed through the definition of a number agent’s

attributes that allow establishing better fitness of partners inside teams. These attributes

include goals alignment, partners’ skills, performance rating (of tasks), levels of

commitment, and the risk (e.g., the risk of a breakdown). The multi-agent system

described in Zheng and Zhang (2005) simulates an artificial market-place, in which

partners organise themselves in VOs. They adopted the motivational quantities frame-

work for the agent’s reasoning processes and to quantitatively represent the progress

towards the organisational goals.In Viswanadham and Gaonkar (2003) a mixed-integer programming model for partner

selection in a supply-chain context is suggested, in which participants share information on

their capacities, schedules, and cost structures. Based on this information, the decision

model addresses the partner’s selection problem in terms of profit maximisation, while

considering various manufacturing and logistics constraints.In Fischer et al. (2004), the virtual enterprise model is based on the concept of

aggregation of small performance units called competence cells. The methodology,

Table 8. Another example of VO suggestions.

Solution o1 o2 o3 o4 o5 o6 o7 Duration

1 c1 c4 c3c2 38

2 c1 c4 c3c2 39


formulated as an optimisation problem, chooses the most capable competence cellsaccording to the core competence. The optimisation conditions consider such factors astime saving, similarity between needed and candidate’s competence, delivery date and itsprobability, and costs.

The work of Camarinha-Matos and Abreu (2005) proposes an approach for benefitsanalysis in collaborative processes for networks of enterprises and introduces the notion ofpast performance as a criterion for future selection. By observing the history of benefits-flow between partners, it is possible to create indicators that can be used to assert partners’collaboration levels, which can be used to select/search partners for future collaborationsituations.

Other recent works started introducing new factors leading to the notion ofpreparedness assessment. For instance, in Holtbrugge (2004), the cultural and strategiccompatibility of partners are seen as a particularly important criterion for partners’selection. A scoring model for partner analysis is proposed, in which several aspects areconsidered, such as the state of cooperation, e.g., harmony among partners, morale,adaptability, and learning.

In Jarimo et al. (2005) the process of partners’ selection is defined as a multi-attributedecision-making problem, in which a hierarchy of attributes is used to characterisepartners. These attributes include elements such as expertise, resources, performance,competence, and economical situation of a partner that are categorised. But it alsoconsiders the ‘network preparedness’, in which the business culture, competition, trust,intelligence, motivation, and infrastructure of a partner are characterised.

An extremely important enabler of collaboration is trust. In Blomqvist et al. (2005), theresearch is focused on the roles of trust and contracts. They examine the potential forbalancing trust and contracting, affirming that contracts alone cannot guaranteesuccessful collaboration, but that the contracting process could be purposefully used toincrease mutual understanding, learning and trust. In this sense, trust both complementscontracts and is a threshold condition for collaboration.

The work of Msanjila and Afsarmanesh (2008) introduces the notion of rational trustin a context of VO breeding environments. Through the definition of a number of trustcriteria based on observable/measurable facts, the method allows for establishing a relativegrading of trustworthiness among enterprises, which can provide a useful indicator forpartners’ selection.

In Haller (2008) a trust management approach based on a Bayesian reputation systemis proposed to help in choosing more reliable partners. In this research, reputation is takenas a trust measure, aggregated from multiple independent trust sources. At the base ofthese measures are elements such as financial, organisational, and operational aspects ofnetwork members, including external and third party entities.

The partners’ search and selection problem is also addressed in Camarinha-Matos et al.(2005) and Camarinha-Matos and Afsarmanesh (2006), in the context of the full process ofVO creation. The proposed framework for VO creation refers to as important issues,usually considered in this process, the elements for search and selection (technical,economical, reliability indicators, preferences); matching algorithms; (multi-criteria)selection criteria; optimisation; assessment (preparedness, etc.), consideration of colla-boration history; external search (if the internal offer is insufficient). Although, partner’sselection is not taken as an ‘optimisation’ problem in a strict sense, instead, it ratherincludes many other factors, some of them of a subjective nature (e.g., personal preferencesand established trust based on previous experience). Furthermore, it is proposed that the


partners’ selection problem is better handled in a context of a VO breeding environment,

for which the membership in this long-term collaborative association ensures the desiredpreparedness to collaborate.

As referred to before, there is now an increasing trend in research towards

incorporating aspects of a ‘soft nature’ in partners’ selection. Nevertheless, existing

research on collaboration preparedness, based on soft aspects, mainly those exploring abehavioural dimension, seem to still be scarce. One contribution in this direction is, for

instance, given in Camarinha-Matos and Macedo (2007), which establishes adependency of the joint behaviour from the underlying value systems prevalent in the

network. In Westphal et al. (2007) the problem of collaboration performance is

addressed, using aspects, such as flexibility, reliability and commitment, which can beconsidered as traits of an organisation’s character. In Romero et al. (2007) the

definition of guidelines for governance rules and bylaws for behaviour regulation is

attempted. In Conte et al. (2004), a collaboration readiness methodology composed ofmotivation, capability and interoperability assessments is presented. Another early

example is Caballero et al. (2000) that presents a methodology to evaluate enterprise

candidates to become members of virtual industrial clusters, based on a methodologythat integrates a set of benchmarking tools to evaluate an enterprise’s competencies and

infrastructures.

5. Conclusions

Collaboration can be highly beneficial and even a survival factor for industrial companies.

But it can also be risky, being important to assess the readiness of potential partners.Although most works in the past were focused on ‘hard’ factors such as competency

matching or technological preparedness, the success of a collaborative process depends on

several other factors of a ‘soft’ nature such as organisation’s character, willingness tocollaborate, or the affectivity/empathy relationships. A preliminary approach to handle

such elements was introduced.A belief network was proposed to model the behavioural aspects of organisations

working in collaboration, according to a set of support definitions provided. Based

on available evidence of organisations’ characters, this model provides predictions on

collaboration preparedness. Extending the model to a decision network alloweddecision making using our collaboration preparedness concept, which allows

selecting the candidates to invite for a new collaboration opportunity. Using this

framework, the decisions can take into account the broker’s personal preferences,which can be incomplete, imprecise and quite subjective, allowing for a higher

flexibility of our preparedness model. As a result, the proposed approach can cope

with the underlying uncertainty that always characterises a partners’ selectionprocess.

The preliminary results show that this assessment approach is feasible and promising.

Nevertheless, further research is needed towards the development of a full assessmentmodel for collaboration readiness, which is the subject of our ongoing research.

Acknowledgement

This work was funded in part by the European Commission through the ECOLEAD project.


Note

1. Given a set A¼ {a, b, c}, the power set of A, or the set of all subsets of A, is 2A¼ {{}, {a}, {b},{c}, {a, b}, {a, c}, {b, c}, {a, b, c}}.

References

Afsarmanesh, H. andCamarinha-Matos, L.M., 2005. A framework formanagement of virtual organi-

zation breeding environments, In: L.M. Camarinha-Matos, H. Afsarmanesh and A. Ortiz, eds.

Collaborative networks and their breeding environments. New York: Springer, 35–48.

Blomqvist, K., Hurmelinna, P., and Seppanen, R., 2005. Playing the collaboration game right –

balancing trust and contracting. Technovation, 25 (5), 497–504.

Boutilier, C., 2003. On the foundations of expected utility. In: Proceedings of the nineteenth annual

conference on uncertainty in artificial intelligence (UAI-03), 9–15 August 2003, Acapulco,

Mexico, 98–106.Caballero, D., Molina, A., and Bauernhans, T., 2000. A methodology to evaluate enterprises to

become members of virtual industry clusters, In: L.M. Camarinha-Matos, H. Afsarmanesh

and R.J. Rabelo, eds. E-Business and virtual enterprises. Norwell, MA: Kluwer Academic

Publishers, 443–454.

Camarinha-Matos, L.M. and Abreu, A., 2005. Performance indicators based on collaboration

benefits, In: L.M. Camarinha-Matos, H. Afsarmanesh and A. Ortiz, eds. Collaborative

networks and their breeding environments. New York: Springer, 273–282.Camarinha-Matos, L.M., et al., 2005. Towards a framework for creation of dynamic virtual

organizations, In: L.M. Camarinha-Matos, H. Afsarmanesh and A. Ortiz, eds. Collaborative

networks and their breeding environments. New York: Springer, 69–80.Camarinha-Matos, L.M. and Afsarmanesh, H., 2006. Creation of virtual organizations in a breeding

environment. In: Proceedings of INCOM’06, 17–19 May St. Etienne, France, 583–591.Camarinha-Matos, L.M. and Macedo, P., 2007. Towards a conceptual model of value systems in

collaborative networks, In: L.M. Camarinha-Matos, et al., eds. Establishing the foundation of

collaborative networks. Boston, MA: Springer, 53–64.Charniak, E., 1991. Bayesian networks without tears. Artificial Intelligence Magazine, 12 (4),

50–63.Cheng J. and Greiner R., 2001. Learning Bayesian belief network classifiers: algorithms and system.

In: E. Stroulia and S. Matwin, eds., AI 2001, Lecture Notes in Computer Science, vol. 2056.

Berlin, Heidelberg: Springer, 141–151.

Cheng, J., Bell, D.A., and Liu, W., 1997. An algorithm for Bayesian belief network construction

from data. In: P. Smyth and D. Madigan, eds., Proceedings of AI and STAT’97, Fort

Lauderdale, Florida, 83–90.Chu, X.N., et al., 2000. Partners selection for virtual enterprises. In: Proceedings of the IEEE third

world congress on intelligent control and automation, vol. 1, 28 June–2 July Hefei, PR China,

164–168.Chun, R., 2005. Ethical character and virtue of organizations: an empirical assessment and strategic

implications. Journal of Business Ethics, 57 (3), 269–284.Conte, M., et al., 2004. The ARICON VE readiness assessment approach in the new product

development (NPD) context, In: L.M. Camarinha-Matos, ed. Virtual enterprises and

collaborative networks. Norwell, MA: Kluwer Academic Publishers, 499–506.

Fischer, M., Jahn, H., and Teich, T., 2004. Optimizing the selection of partners in production

networks. Robotics and. Computer-Integrated Manufacturing, 20 (5), 593–601.

Friedman N., 1997. Learning belief networks in the presence of missing values and hidden variables.

In: D. Fisher, ed., Proceedings of the Fourteenth International Conference on Machine Learning.

San Francisco, CA: Morgan Kaufmann, 125–133.


Goldie, P., 2004. What people will do: personality and prediction. Richmond Journal of Philosophy,

(7), 11–18.Gothlich, S.E., 2003. From loosely coupled systems to collaborative business ecosystems. Paper no.

573, University of Kiel, Germany. Available from: http://www.bwl.uni-kiel.de/grad-kolleg/de/

kollegiaten/goethlich/Business%20Ecosystems%204b%201,5Z.pdf [Accessed 17 March 2008].Gupta, P. and Nagi, R., 1995. Flexible optimization framework for partner selection in agile

manufacturing. In: Proceedings of the 4th industrial engineering research conference, 21–25

May Nashville, TN, 691–700.

Haller, J., 2008. A Bayesian reputation system for virtual organizations, Negotiation, auctions, and

market engineering; lecture notes in business information processing. Berlin: Springer, 171–178.

Available from: http://www.springerlink.com/content/m78322g8571741u2/ [Accessed 4

November 2008].Holtbrugge, D., 2004. Management of international strategic business cooperation: situational

conditions, performance criteria, and success factors. Thunderbird International Business

Review, 46 (3), 255–274.Horvitz, E.J., Breese, J.S., and Henrion, M., 1988. Decision theory in expert systems and artificial

intelligence. International Journal of Approximate Reasoning, 2 (3), 247–302.Jarimo, T., et al., 2005. Hierarchical multi-attribute decision support approach to virtual

organization creation, In: L.M. Camarinha-Matos, H. Afsarmanesh and A. Ortiz, eds.

Collaborative networks and their breeding environments. New York: Springer, 135–142.Jensen, F.V., 1996. Bayesian networks basics. In: L.C. Van Der Gaag, ed., AISB Quarterly

(Newsletter of the Society for the Study of Artificial Intelligence and Simulation of Behaviour,

Special Theme: Bayesian Belief Networks), 94, Springer, 9–22.Martinich, J.S., 1997. Production and operations management: an applied modern approach.

New York: John Wiley & Sons.Mowery, D.C., Oxley, J.E., and Silverman, B.S., 1998. Technological overlap and interfirm

cooperation: implications for the resource-based view of the firm. Research Policy, 27 (5),

507–523.Msanjila, S.S. and Afsarmanesh, H., 2008. Trust analysis and assessment in virtual organization

breeding environments. International Journal of Production Research, 46 (5), 1253–1295.Muskin, J.B., 2000. Interorganizational ethics: standards of behavior. Journal of Business Ethics,

24 (4), 283–297.

Norsys, 1997. Application for belief networks and influence diagrams users guide, Norsys Software

Group. Available from: http://www.norsys.com [Accessed 17 March 2008].

Oxford Dictionary of English, 2003. C. Soanes and A. Stevenson, eds. second edition, ISBN

0-19-8613474. Oxford, UK: Oxford University Press.Pearl J., 1996. Decision making under uncertainty. In: ACM Computing Surveys, 28 (1), 89–92.

Petersen, S.A. and Divitini, M., 2002. Using agents to support the selection of virtual enterprise

teams. In: Proceedings of fourth international bi-conference workshop on agent-oriented

information systems (AOIS-2002), at AAMAS’02, 16 July 2002, Bologne, Italy, 98–112.

Available from: http://ftp.informatik.rwth-aachen.de/publications/CEUR-WS/vol-59/9

petersen.pdf [Accessed 12 March 2000].

Renooij, S. and Gaag, L.C., 1998. Decision making in qualitative influence diagrams, Proceedings of

the eleventh international FLAIRS conference, 18–20 May 1998, Sanibel Island, Florida. Menlo

Park, CA: AAAI Press, 410–414.

Romero, D., et al., 2007. Towards governance rules and bylaws for virtual breeding environments,

In: L.M. Camarinha-Matos, et al., eds. Establishing the foundation of collaborative networks.

Boston, MA: Springer, 93–102.Rosas, J. and Camarinha-Matos, L.M., 2008a. A collaboration readiness assessment approach. In:

Proceedings of the 8th IFIP international conference on information technology for balanced

automation systems – Basys 2008. 23–25 June 2008, Porto, Portugal. Boston, MA: Springer,

77–86.


Rosas, J. and Camarinha-Matos, L.M., 2008b. Modeling collaboration preparedness assessment. In:L.M. Camarinha-Matos and H. Afsarmanesh, eds. Collaborative networks – referencemodelling. Boston, MA: Springer, 227–252.

Sandelands, L.E. and Stablein, R.E., 1987. The concept of organization mind. In: S. Bacharach

and N. DiTomaso (Eds.), Research in the sociology of organization. Greenwich, CT: JAIPress, Vol. 5, 135–161.

Szolovits, P., 1995. Uncertainty and decisions in medical informatics. Methods of Information in

Medicine, 34, 111.Viswanadham, N. and Gaonkar, R.S., 2003. Partner selection and synchronized planning in

dynamic manufacturing networks. IEEE Transactions on Robotics and Automation, 19 (1),

117–130.Walsh, J.P. and Ungson, G.R., 1997. Organizational memory, In: L. Prusak, ed. Knowledge in

organizations. Boston, MA: Butterworth-Heinemann, 177–212.

Webber, J., 2006. Virtue, character and situation. Journal of Moral Philosophy, 3 (2), 193–213.Westphal, I., Thoben, K.D., and Seifert, M., 2007. Measuring collaboration performance in virtual

organizations, In: L.M. Camarinha-Matos, et al., eds. Establishing the foundation ofcollaborative networks. Boston, MA: Springer, 33–42.

Wilkinson, I.F., Freytag, P., and Young, L., 2005. Business mating: who chooses whom and getschosen? Industrial Marketing Management, 34 (7), 669–80.

Zheng, Q. and Zhang, X., 2005. Automatic formation and analysis of multi-agent virtual

organization. Journal of the Brazilian Computer Society: Special Issue on Agents Organizations,11 (1), 74–89.


an approach to assess collaboration readiness

Documents