Apr 1, 2004 - Considering knowledge sharing as a decisional act founded on two different .... Gundry (2000) affirms that trust is necessary for communication, ...
Trust Mediation in Knowledge Management and Sharing Cristiano Castelfranchi Institute for Cognitive Sciences and Technologies – CNR T3 Group – Trust: Theory & Technology. In Jensen, C., Poslad, S. & Dimitrakos, T. (Eds.) Trust Management. Second International Conference, iTrust 2004, Oxford, UK, March 29 - April 1, 2004. Proceedings Series: Lecture Notes in Computer Science, Vol. 2995
Introduction The aim of this paper is to provide a theory of the role trust plays in knowledge sharing processes, by exploiting a cognitive model of the goals and beliefs of an agent involved in the decision of passing and/or accepting knowledge1, and a related model of the symmetrical or asymmetrical trust relationships within a group (Trust-Nets). This theory is based on the claim that trust is a fundamental mediator in knowledge sharing, so as most authors studying Knowledge Management assert. Anyway they do not analyse the relationship between trust and knowledge circulation theoretically, what is the objective of our work. Considering knowledge sharing as a decisional act founded on two different socio-cognitive actions: to pass knowledge and to accept knowledge, we build a cognitive model of agent’s mind when passing or accepting it, that is a list of his/her goals and beliefs in accordance with which s/he decides if sharing knowledge or not; several of them are trust ingredients. Thus, trust - as mental attitude (a specific set of beliefs and expectations) - comes into play in knowledge sharing process. However, in our analysis trust is not only a subjective disposition (towards others) but it is also an act (the act of trusting somebody) and a social (more or less stable) relationship. This is why we also analyse trust relations not in mental terms but in structural-relational terms, as a net of “channels” for knowledge circulation.2
1. The Knowledge Management perspective and the issue of trust Knowledge Management is a discipline which considers knowledge as the most important organization’s resource and as a strategy whose objective is to collect, manage and put knowledge in action, so that it circulates and develops continuously. Managing and diffusing an enterprise’s knowledge capital determines the innovation of products, processes, organizational models and answers to the customers improving its competitive position. This kind of knowledge concerns creativity, problem solving, disposition to dialogue, uncertainty management and sensitiveness towards cultural context in which the 1
For an extended theory of this – also based on an empirical investigation - see Castelfranchi and Postorino “Trust Relationships and Knowledge Sharing in Organizations” (in preparation). 2 The two reference models we take into consideration in our work are: the socio-cognitive model of trust developed for socio-psychological, organisational and computational studies by Castelfranchi and Falcone, and Knowledge Management studies.
organization operates. We assume as references in this domain two of the most relevant contributions to organisational knowledge studies; those provided by Nonaka and Takeuchi (1996) and Davenport and Prusak (1998). Nonaka e Takeuchi affirm that knowledge creation core is the conversion of tacit knowledge in explicit one and vice versa. The knowledge distinction in tacit and explicit dates back to Polanyi (1967) who defined tacit knowledge as personal, peculiar of a context and therefore hard to be codified comparing it to explicit knowledge that is only the ice-berg peak of the whole knowledge corpus and can be on the contrary codified and transferred through a formal and systematised language. The two “epistemological” (Nonaka & Takeuchi, 1996) knowledge dimensions are complementary. Knowledge can be created and diffused just through their continuous interaction, that they call Knowledge Spiral. Knowledge creation process is also the creation of Ba (Nonaka, Konno & Toyama, 2001), the specific time and space where interactions and interpersonal relationships from which knowledge cannot be separated take place: it is the platform where knowledge concentrates. The four modalities of knowledge conversion are: 1. socialization: conversion of tacit knowledge in tacit one through the building of an interaction field where organisation members can exchange experiences and share mental models. It is the modality of originating ba, where members meet face-toface and share emotions and feelings like care including also commitment, love and trust; 2. externalisation: conversion of tacit knowledge in explicit one through dialogue, using metaphors and analogies. It is the modality of dialoguing ba and of peer-topeer interaction; 3. combination: the elaboration of knowledge through knowledge networks enabling collaboration also thanks to information technology (systematising ba); 4. internalisation: the conversion of explicit knowledge in tacit one through the process of learning by doing “on the site”: in a shared time and space where focused training with colleagues and senior members helps to underline specific mental patterns (exercising ba). In “high care” organisations members tend to bestow knowledge: they form a social network supporting the individual who is not afraid to ask help in performing his/her task because all members share “care” value and are accessible at every level. Members are indulgent towards errors (idem), because sharing the value of learning and developing new knowledge, important not for power gaining but for the satisfaction of a common goal: problem solving and task performing, that is the base of strict cooperation (Castelfranchi, 1998). In this view the relevance of trust in knowledge creation process is evident, even though not analytically explicated. In fact every such step of this ‘knowledge conversion’ requires trust. Sharing knowledge (both in ‘externalisation’ or in ‘combination’) is a specific action, due either to personal purposes (for personal advantages) or to shared objectives and values (cooperation for common goals), or to role duties, or finally to habits and routines. In all these different motivational sets some form of either implicit confidence (in routines and habits, or in role playing) or of explicit trust evaluations (of the others and of the context) is always needed. Analogously, in ‘socialisation’ a trust atmosphere (see §4.) is both the result and the condition of the process; while ‘internalisation’ presupposes trust in the senior member one is adopting as a model, and in the established practices.
Davenport and Prusak (1998) consider knowledge as a changeable mix of structured experiences, values, information about the context and intuitions coming from experience, providing a model where new information and experiences are incorporated. It finds itself in individuals’ heads and is embedded in organisation’s documents, routines, processes, practices and norms. It is different from information but derives from it so as information differs from data but is formed by them. The authors stress the importance of trust in Knowledge Management sustaining that it supports the “knowledge market”. This market is based on reciprocity, reputation and altruism working effectively only thanks to trust, that must be visible (members must see that passing their knowledge people receive real acknowledgement: they must experience reciprocity directly); ubiquitous (because if even one part of internal knowledge market is not trustworthy, the whole market becomes asymmetric and less efficient) and must come from the top (if top managers are not perceived as trustworthy from members but exploiting others’ knowledge for personal incomes a distrust atmosphere will spread in the organisation, while if their image is positive this will influence also the whole organisation’s perceived trustworthiness). Trust must be promoted through frequent face-to-face meetings. Knowledge management is characterised by three processes: generation, codification and transfer of knowledge, that can often be inhibited by friction factors: first of all the lack of trust, whose remedy is the establishment of relationships and personal contacts among people sharing the same culture and language. Other authors writing about Knowledge Management emphasise the importance of trust. Sveiby (1996) considers trust the crux of knowledge sharing and suggest to invest in human resources, encouraging meetings among members working at the same projects, promoting dialogue and pleasant environments where good relationships without fear can grow. Gundry (2000) affirms that trust is necessary for communication, knowledge exchange and information sharing typical of cooperative team work. He underlines the circular relation between knowledge sharing and trust: if trust helps information exchange, people tend also to trust persons who share knowledge more than others. Notice that all these authors make clear statements about the crucial role of trust and correctly focus on the issue, but they are rather generic and vague. Not having a well defined model of trust they cannot account for how and why precisely trust mediates and “is the crux” of knowledge sharing and management. A definitely richer contribution to our theme is provided by Jones and George (1998). They distinguish trust in conditional and unconditional, considering the latter as the ideal condition for knowledge sharing in an organisation. They argue that an agent starts a trust interaction with another behaving as if s/he trusts him/her: it’s a “by default” and blind trust, the first step of a potential interaction that helps a future trust relationship establishment3 more than starting with diffidence; the “null degree” trust is conditional trust: it induces agents to interact until they behave correctly, promoting future interactions until they develop positive feelings and opinions about others’ trustworthiness; the highest form of trust is unconditional trust based on values sharing: the more agents interact the more they understand to share the same values, so they do not trust “by default” as before, but because they experienced directly other’s trustworthiness and his/her adoption of their same values. This kind of trust encourages cooperation not only because it supports colleagues in following a common goal as conditional trust does, but also because it lets 3
Because trust creates trust (Falcone e Castelfranchi, 2001). This default attitude has been proved for ex. by Axelrod to be a good precondition for promoting cooperation even among self-interested agents.
them feel at ease in asking help or knowledge: if all members share the value of cooperativeness, then nobody should feel inadequate or threatened for his/her position of “dependence”. Moreover, the belief of sharing the same values incites the agent to sacrifice his/her personal interest and to trust unconditionally colleagues who become even friends. In the end, since tacit knowledge is embedded in social interactions and transferred by them and since social interactions are supported by unconditional trust, then tacit knowledge sharing is possible until members trust each other unconditionally. This distinction is very interesting, even though the meaning of ‘unconditional trust’ is not so clear: is it trust without evidences justifying it or is trust without control or is trust without counterpart? Anyway, it is sure that values sharing – but only when values are prosocial - facilitates trust relationships: other’s behaviour is more predicable, because the agent can attribute to colleagues the same values s/he adopts. In the end, even though Jones sand George do not explicate it theoretically, it is deducible that unconditional trust is less expensive than conditional one, which needs information, accepts less risk and is always subject to control. We will discuss this issue later.
2. What trust is: a socio-cognitive model The Socio-Cognitive model of trust (for a more complete presentation see Castelfranchi and Falcone 1998; 2000; 2002; Castelfranchi & Tan, 2001; Falcone and Castelfranchi, 2001) is based on a portrait of the mental state of trust in cognitive terms (beliefs, goals). This is not a complete account of the psychological dimensions of trust: it represent the most explicit (reason-based) and conscious form. The model does not account for the more implicit forms of trust (for example trust by default, not based upon explicit evaluations, beliefs, derived from previous experience or other sources) or for the affective dimensions of trust, based not on explicit evaluations but on emotional responses and an intuitive, unconscious appraisal. The word trust means different things, but they are systematically related with each other. In particular, three crucial concepts have been recognised and distinguished not only in natural language but in the scientific literature. Trust is at the same time: • a mere mental attitude (prediction and evaluation) towards an other agent, a simple disposition; • a decision to rely upon the other, i.e. an intention to delegate and trust, which makes the trustor “vulnerable”; • a behaviour, i.e the intentional act of trusting, and the consequent relation between the trustor and the trustee. In each of the above concepts, different sets of cognitive ingredients are involved in the trustier’s mind. The model is based on the BDI (Belief-desire-intention) approach for modeling mind, that is inspired to the Bratman’s philosophical theory. First of all, in the trust model only an agent endowed with both goals and beliefs can “trust” another agent. Let us consider the trust of an agent X towards another agent Y about the (Y's) behaviour/action α relevant for the result (goal) g when: X is the (relying) agent, who feels trust; it is a cognitive agent endowed with internal explicit goals and beliefs (the trustier); Y is the agent or entity which is trusted (the trustee); X trusts Y about g/α and for g/α.
In the model Y is not necessarily a cognitive agent (for instance, an agent can -or cannottrust a chair as for as to sustain his weight when he is seated on it). On the contrary, X must always be a cognitive agent: so, in the case of artificial agents we should be able to simulate these internal explicit goals and beliefs. For all the three notions of trust above defined (trust disposition, decision to trust, and trusting behaviour) we claim that someone trusts some another one only relatively to some goal (here goal is intended as the general, basic teleonomic notion, any motivational representation in the agent: desires, motives, will, needs, objectives, duties, utopias, are kinds of goals). An unconcerned agent does not really "trust": he just has opinions and forecasts. One trusts another only relatively to a positive expectation, i.e. for something s/he wants to achieve, that s/he desires. If x does not have any concern and goal, she cannot really decide, nor care about something (welfare): she cannot subjectively «trust» somebody. Second, trust itself consists of beliefs. 2.1 A Cognitive Anatomy of Trust Since Y’s action is useful to X (trust disposition), and X has decided to rely on it (decision to trust), this means that X might delegate (act of trusting) some action/goal in his own plan to Y. This is the strict relation between trust disposition, decision to trust, and delegation. The model includes two main basic beliefs (evaluations) (we are considering the trustee as a cognitive agent too): - Competence Belief: a sufficient evaluation of Y's abilities is necessary, X should believe that Y is useful for this goal of its, that Y can produce/provide the expected result, that Y can play such a role in X’s plan/action. - Willingness Belief: X should think that Y not only is able and can do that action/task, but Y actually will do what X needs (under given circumstances). This belief makes the trustee's behaviour predictable. Another important basic belief for trust is: - Dependence Belief: X believes - to trust Y and delegate to it - that either X needs it, X depends on it (strong dependence), or at least that it is better to X to rely rather than do not rely on it (weak dependence). In other terms, when X trusts someone, X is in a strategic situation: X believes that there is interference and that his rewards, the results of his projects, depend on the actions of another agent Y. Obviously, the willingness belief hides a set of other beliefs on the trustee’s reasons and motives for helping. In particular, X believes that Y has some motives for helping it (for adopting its goal), and that these motives will probably prevail - in case of conflict - on other motives, negative for it. Notice that motives inducing to adoption are of several different kinds: from friendship to altruism, from morality to fear of sanctions, from exchange to common goal (cooperation), and so on (figure 2). This is why, for example, it is important to have common culture, shared values, the same acknowledged authorities between trustier and trustee. 2.2 Internal vs Environmental Trust Another important characteristic of the socio-cognitive model of trust is the distinction between trust ‘in’ someone or something that has to act and produce a given performance
thanks to its internal characteristics, and the global trust in the global event or process and its result which is also affected by external factors like opportunities and interferences. Trust in Y (for example, ‘social trust’ in strict sense) seems to consists in the two first prototypical beliefs/evaluations identified as the basis for reliance: ability/competence (that with cognitive agents includes knowledge and self-confidence), and disposition (that with cognitive agents is based on willingness, persistence, engagement, etc.). Evaluation about external opportunities is not really an evaluation about Y (at most the belief about its ability to recognize, exploit and create opportunities is part of our trust ‘in’ Y). We should also add an evaluation about the probability and consistence of obstacles, adversities, and interferences. Trust can be said to consist of or better to (either implicitly or explicitly) imply the subjective probability of the successful performance of a given behaviour α, and it is on the basis of this subjective perception/evaluation of risk and opportunity that the agent decides to rely or not Y. However, the probability index is based on, derives from those beliefs and evaluations. In other terms the global, final probability of the realization of the goal g, i.e. of the successful performance of α, should be decomposed into the probability of Y performing the action well (internal attribution) and the probability of having the appropriate conditions (external attribution) for the performance and for its success, and of not having interferences and adversities (external attribution). This decomposition is important because: the trustier’s decision might be different with the same global probability or risk, depending on its composition (for example for personality factors); trust composition (internal Vs external) produces completely different intervention strategies: to manipulate the external variables (circumstances, infrastructures) is completely different than manipulating internal parameters. 2.3
Degrees of trust
The idea that trust is gradual quantity is common (in common sense, in social sciences, in AI). However, since no real definition and cognitive characterization of trust is given, the quantification of trust is quite ad hoc and arbitrary, and the introduction of this notion or predicate is semantically empty. On the contrary, we claim that there is a strong coherence between the cognitive definition of trust, its mental ingredients, and, on the one side, its value, and on the other side, its social functions and its affective aspects. More precisely the latter are based on the former. In our model we ground the degree of trust of x in y, in the cognitive components of X's mental state of trust. More precisely, the degree of trust is a function of the subjective certainty of the relevant beliefs. We use the degree of trust to formalize a rational basis for the decision of relying and betting on y. We also claim that the "quantitative" aspect of another basic ingredient is relevant: the value or importance or utility of the goal g. In sum, the quantitative dimensions of trust are based on the quantitative dimensions of its cognitive constituents. For us trust is not an arbitrary index with operational importance, without a real content, but it is based on the subjective certainty of the relevant beliefs which support each other and the decision to trust (Figure 1).
Self-Confidence Know How
MOTIVATION
Ability MORALITY
INTENDS & PERSISTS
FEAR of AUTHORITY
COMPETENT
PREDICABILITY Will Do UNHARMFULL
INTERNAL
DECISION to trust y. to count on y
OPPORTUNITIES
Figure 1 If we call DoTXYτ the degree of trust of an agent X about Y on the task τ=(α,g) we have: DoTXYτ = DoCX[OppY(α,g)] * DoCX[AbilityY(α)] * DoCX[WillDoY(α,g)] where: - DoCX[OppY(α,g)], is the degree of credibility of X's beliefs about the Y’s opportunity of performing α to realize g; - DoCX[AbilityY(α)], the degree of credibility of X's beliefs about the Y’s ability/competence to perform α; - DoCX[WillDoY(α,g)], the degree of credibility of X's beliefs about the Y’s actual performance; DoCX[WillDoY(α,g)] = DoCX[IntendY(α,g)] * DoCX[PersistY(α,g)] (given that Y is a cognitive agent) In any circumstance, an agent X endowed with a given goal, has three main choices: i) to try to achieve the goal by itself; ii) to delegate the achievement of that goal to another agent Y; iii) to do nothing (relatively to this goal), renouncing. Consider the simplified scenario in which only (i) and (ii) are the possible choices we have the Figure 2:
U(X)p+ Success
To do by itself Failure U(X)p-
Delegation
U(X)d+
Success Failure U(X)d-
Figure 2 Where if U(X) is the agent X's utility function, more specifically: U(X)p+, the utility of the X’s success performance; U(X)p-, the utility of the X’s failure performance; U(X)d+, the utility of a successful delegation (the utility due to the success of the delegated action);
U(X)d-, the utility of a failure delegation (the damage due to the failure of the delegated action). In the above scenario, in order to delegate we must have: DoT XY τ * U(X) d + + (1 - DoT XY τ ) U(X) d - > DoT XX τ * U(X) p + + (1 DoT XX τ ) U(X) p where DoTXXτ is the selftrust of X about τ. More precisely, we have: U(X) p + = Value(g) + Cost [Performance(X)], U(X) p - = Cost [Performance(X)] + Additional Damage for failure U(X) d + = Value(g) + Cost [Delegation(X Y)], U(X) d - = Cost [Delegation(X Y)]+ Additional Damage for failure where is supposed that it is possible to attribute a quantitative value (importance) to the goals and where the costs of the actions (delegation and performance) is supposed to be negative. There is threshold for the decision to trust y (which entails a risk and a bet). We evaluate both the external attribution of trust, i.e. the environmental favoring conditions and opportunities for Y’s successful action, and the internal attribution of trustworthiness to y (trust in y). This trust in Y has two facets: competence and willingness. Each of them is based on and analyzed in terms of other beliefs about y (for example Y’s motives or know how). The final decision about relying on Y’s action will be taken on the basis of the strength of those beliefs about y and about the environment.
3. Mental ingredients in knowledge sharing and the role of trust In our perspective, knowledge sharing is both a state resulting from a process/activity and the process/activity in itself. This process entails on the subjective, cognitive site two fundamental decisions and actions: 1. to transfer, to pass a piece of knowledge (PassK); 2. to accept a given piece of knowledge (AccK). In the former case, agent x decides whether passing or keeping knowledge; in the latter case, s/he decides whether accepting or rejecting it. (Castelfranchi, 2001) Trust attribution supporting the decision to share knowledge is external and internal, as in the classic model of trust. Internal attribution is constituted by all agent x’s beliefs about himself/herself and his/her own knowledge and about agent y. External attribution is constituted by all agent x’s beliefs about the environment where interaction takes place, that is organisation intended as: • a structure with rules, routines, procedures, represented by an authority A and defended by a controller C (that can coincide with A too) who has the right and the power to sanction positively or negatively workers’ behaviour; • the whole of members working in the same organisation, who are potential sources of information and of reputation about colleagues (so about x and y too); • a set of sets of roles, infrastructures, and of practices, data and procedures, that should be appropriate for an efficacy and safe work. Since in our perspective there are two crucial subjective and interpersonal moves for the knowledge sharing process, we examine them in terms subjective evaluations and expectations (that is: beliefs) about each others, about k, and about the organisational
context. It will be clear that several of them are typical trust beliefs i.e. the beliefs that constitute trust as mental disposition. 3.1 Predictions about K-sharing K-Sh is the result of two different socio-cognitive operations: • passing or not-passing K (PassK); • accepting or rejecting K (AccK). Those moves determine the spreading of K in the social net (Castelfranchi, 2001) and their respective decisions are influenced by different factors, but both of them also depend on trust. PassK is mainly influenced by centrifuge/outputting trust: the more X trusts Y the more s/he will be prone to pass her/his K to Y. More precisely, X should trust Y not generically, but for specific features relevant for passing K to Y: 1. X believes that Y is able to understand K, to understand the use of K; 2. X believes that Y will appropriately use K (for the Org purposes and advantage); 3. X trusts that Y will not make worse (and possibly improve) Y’s attitude towards X (Y will reciprocate, or be grateful, or more friendly close, Y will increase trust towards X, etc.) but also 4. X believes that Y trust him/her enough to AccK (belief about centripetal trust) Plus other necessary beliefs: 5. X believes that Y needs K 6. The belief 1 – belief of competence – and also 2 often implies other two beliefs. If X doesn’t experience Y’s competence directly, it means that X relies on Y’s reputation, that is X trusts other members of the organization as sources of information about Y. 7. The belief 4 about Y’s honesty (and so about his/her reciprocation too) implies that X believes that Y has an “Assumption of Normative Pertinence” and also a Normative Goal (that is Y knows the rules and believes that they regard him/her too, so s/he has the goal to adopt and respect them) 8. X relies on the organization in general, because s/he believes that if Y has been employed it means that s/he has got the necessary competencies to work in that organization or the necessary formation to understand k. 9. X trusts that Y has not k so that s/he depends on X to get it 10. X believes that if Y shows not to have k, it’s true (s/he doesn’t deceive). 11. X trusts (or better: x believes, but s/he actually has no experience to which refer, so s/he hopes) that Kshar is an advantage, improving his/her work (this will be a reciprocation, even though not immediate) 12. X trusts Organization Authority, that is s/he adopts the rules of the Organization, so s/he trusts other members because s/he believes that they also adopt them and trust (have fear of) Authority and her capability to check and sanction. 13. X trusts of course the Controller too, that is his/her capability to check and evaluate who respects or disobey rules.
14. X believes that Y would like to be well-evaluated by X (that is Y’ s goal of good reputation). 15. X believes that Y has the goal to be trusted by X (that’s why s/he will use correctly k, s/he will reciprocate…). 16. X is self-confident: he trusts his/her knowledge (or competence…). 17. X trusts that k is adequate, important or necessary to (solve) a given situation or to be transferred to Y. This is a question of self-confidence too, because X trusts his/her ability to evaluate how, when, how much and with whom sharing k. 18. Eventually X trusts the sources of information from which s/he received k. 19. X trusts that s/he will not be replaced by another worker (because of transferring his/her professional value, that is just k!) or at least to keep on being professionally acknowledged for his/her competencies, even though other people have them. 20. X trusts that s/he will receive a professional acknowledgement for his/her SharK, because s/he adopts the rule of K sharing. AccK is mainly (but not only) influenced by centrifuge trust too but in another way; it depends on how much X trusts Y as for being a competent and reliable source and as for being loyal and good-willing towards X (not having reasons for deceiving X or for inducing X in error; how much Y is in competition with X, etc.) 1. X trusts that Y is well informed and competent 2. X trusts that Y is trustworthy, not deceiving: it means that Y believes that if X gives k then X believes that k is true. 3. X trusts that Y is not hostile with X, that means also that X believes that Y trust him/her enough to PassK (belief about centripetal trust) These seem to be the three fundamental evaluations, the three specific aspects of X’s trust in Y relevant for AccK. They influence X’s trust in K passed by Y: K is a reliable K. Moreover an additional belief is necessary for AccK: 4. X believes that K is useful, is pertinent for some of his/her goal (not irrelevant). As we can see, trust evaluations and expectations about the other agent, the organisation and its rules and structures, one-selves, and k are a necessary ground for the knowledgesharing process and a crucial component of the mental attitudes that it requires in the organisational actors. This is why is not only reasonable to assume that trust is a precondition for knowledge sharing (and a result of it), but, more precisely, that it is a mediator, a catalyst of the process: it is a mental and interpersonal (cognitive, dispositional, and relational) precise condition for the two crucial steps in the organisational flow of knowledge. Thus, trust is also something that a policy of knowledge management must seriously take into account, build, and defend, by clearly knowing its nature and dynamics. 3.2 The knowledge sharing circularity It is interesting to notice that in order to share “new” knowledge it is necessary to share previous knowledge, or at least that x must believe to have some knowledge in common with the agent he passes knowledge to). First of all, this is necessary because some shared knowledge is presupposed for communicating and for understanding (for example a shared vocabulary and ontology, a shared encyclopaedia, a shared context about work, roles, organisation, etc.).
Second, because (as we will see in section 3.2) most of x’s beliefs for PassK are in fact common and even mutual beliefs between x and y; x and y believe the same (and believe that the other believes so). For example, x and y must believe that k is useful for y’s goal, that k is valid, etc. The knowledge sharing circularity (Figure 3 a) can be either an obstacle or an incentive. If x believes that y does not share a certain cultural or education background with him/her, then s/he can hesitate to pass knowledge to him/her or above all to accept k from him/her. Notwithstanding, if x perceives to share knowledge with y, she is more stimulated to transfer and accept other knowledge by him/her (it functions like that in the case of values sharing which is in a circular relationship with knowledge sharing). The relationship between trust and knowledge sharing is circular too (Figure 3 b): in order to trust y, x must either have information about y, helping him/her to evaluate y’s trustworthiness or having knowledge in common with him/her that encourages the establishment of a trust relationship so as values sharing; on the other hand, in order to share knowledge, it is necessary to have a trust relation or atmosphere.
KNOWLEDGE SHARING
KNOWLEDGE SHARING
TRUST
KNOWLEDGE SHARING
Figure 3 (a, b)
4. Trust as a Net and its predictions Let us now analyse trust relationships not in mental terms (which represent the fine grain, the micro-level) but at a macro-level in structural-relational terms, as “channels” for knowledge circulation (Busch et al., 2001). In fact, trust is not only a mental disposition founding decisions of reliance, delegation, collaboration; as we said at the beginning, it also is an action and an (established or attempted) social relation. Trust relationships around an individual within a given community connecting her/him with other members creates a sort of trust-sociogram, that we call TrustNet. Let us characterise the main features and properties of such a network. First of all, links in the net can be unilateral and asymmetric, or bilateral. They can be more or less strong/weak depending on the strength, intensity of trust (see § 2.3); links can be more or less permanent, stable, since this sociogram can either characterise a temporarily social situation: x has to rely upon somebody or choosing somebody for cooperation, or it can characterise a structure of acquaintance, of stable social “relations”. 4.1 A local perspective Let us first consider the TrustNet only from x’s local and unilateral perspective: where x is the “source” of several trust attitudes/arrows towards others, i.e. only centrifuge/outputting arrows.
Q
Y
U X
W
Z
Centrifuge Trust
Figure 4 This kind of structure allows for some predictions. The strength of x’s trust in y is a predictor of x’s choosing y for reliance, that is of x’s counting upon y. However, this is not a perfect predictor, i.e. the differential amount of x’s trust in y (compared with x’s trust in z or w) does not completely determine x’s choice. There are also other factors. How much is x dependent on y, z or w (Sichman et al. 1998)? How much does x need each of them? And which is the cost of relying upon y rather than upon z or w? In other words, not always the chosen partner is the most trusted. Although more risky certain relationships can be preferable, that is more convenient. In fact in our model describing the role of trust in decision (see Figure 2) an equation predicts whether x will rely upon y or not, on the basis of her/his degree of trust, but also of utility and risk acceptance. 4.2 A more integrated view of TrustNet Let’s now come back to the general characterization of TrustNet for expanding our theory of links by adopting a less local and unilateral view. They can be: 1. Unilateral: X trusts Y but Y does not trust X. 2. Bi-directional: X trusts Y and Y trusts X. Bi-directional trust relationships can be distinguished in various kinds: Bilateral: X trusts Y and Y trusts X in a completely independent way; each of them might even ignore if and how the other trusts her/him. Reciprocal: X trusts Y also because (X believes that) Y trusts X. Mutual (bi-reciprocal): X trusts Y also because (X believes that) Y trusts X, and vice versa, and both of them assume this. 3. Unbalanced: A link is unbalanced when the trust of X for Y is (perceived as) quite broader or stronger than Y’s trust in X. It’s balanced when their trust is of similar force.
X
Y
UNBALANCED
In case of Bilateral trust link (and especially for reciprocal and mutual one) we combined (add) the dimension of the two inverse arrows in a sort of broadness of a trust “channel” between X and Y.
Q Y
U X
W
Z
Trust Channels
Our claim is that: the TrustNet gives us a map of the important “channels” along which private or implicit knowledge becomes common and explicit, i.e. the channels along which knowledge is shared and circulates.(Falcone & Castelfranchi, 2002) • First bronze prediction: the larger the channel the larger the flux of knowledge and its rapidity. • Second bronze prediction: the more channels are bilateral and balanced the more uniformly shared knowledge is. This is also a measure of a trust “atmosphere” (see below). A poor net (few links), subtle channels, unilateral attitudes proportionally reduces knowledge passing and accepting, thus the building of a collective capital. 4.3 “Trust Atmosphere” In this perspective it is possible to provide an operational notion and a model of the so called “Trust Atmosphere”. We define it as: the diffuse perception (beliefs based on experience) within a group or organization G of the fact that “every body trusts everybody “ or better that: the (great) majority of the members of G have multiple and mutual relations of trust. Multiple means that each member (in average) has more than one centrifuge trust relation. We mean that there is not a trust atmosphere if simply each member has a specific and unique trust relations (although bilateral) in G: it is not enough that everybody trusts somebody. It is necessary to have that everybody trusts more than one agent in G, and is trusted by more than one agent. In other words, no one of the following structures would represent a real trust atmosphere.
Mutual means Bilateral + reciprocally known and reinforced by the trust of the other: i.e. each agent knows that the other trusts him and trusts the other also because of this.
Obviously the notion of Trust Atmosphere is a gradual notion: it depends both on the percentage of agent in G that have reciprocal and multiple Trust relationships; on the degree of their trust; on the density, i.e. on the number of reciprocal trust arrows per node. The maximum will be when “everybody trusts very much everybody”. This disposition should be due to the diffuse perception of those relationships. i.e. the members of G share this representation of the G TrustNet, and share this representation because have occasions for observing trust acts among the others, towards oneself, and to listen to positive reputation spreading. Moreover, something additional seems crucial for a climate or atmosphere of trust: the fact that trust relationships will mainly be of the “unconditional” kind or better that they are “trust by-default” 4 and implicit, affective forms of disposition. The last point is – till now – more difficult to be modeled (Castelfranchi, 2000).
5. Concluding remarques Trust has been identified in knowledge management literature as a crucial condition for the creation of an effective collective cognitive capital. However, no analytic model has been proposed about why and how this mediation works. We have proposed a cognitive model of trust, mainly in terms of beliefs (evaluations and expectations), to understand the precise mechanism of such a catalytic role of trust. We decomposed the knowledge sharing process in two basic decisions/actions of the participants: passing k and accepting k. We have carefully analysed the mental attitudes involved in those moves and thus in the knowledge sharing process. On such a basis, we have shown how trust ingredients are necessarily present and required while passing or accepting knowledge. After this microanalysis of relevant cognitive presuppositions of knowledge management, we adopted a more coarse macro-level perspective, looking at trust not only as a psychological disposition but as a social relationship. We introduced the idea of a TrustNet with more or less strong and bilateral links, and on such a basis we introduced some prediction about knowldge circulation along the trust “channels”. On the basis of this analysis, we would like to conclude that: a) A socio-cognitive view of trust is necessary for modelling trust issues in organisations; b) while caring of making knowledge capital explicit and circulating, an organisation should care of what the beliefs of the actors about k, about the organisation values, authority, and infrastructure, about each-others are, and what they expect and feel on the basis of such beliefs. In knowledge management organisations should monitor and build the right expectations in their members. Knowledge management entails a cognitive, affective, and structural “trust management” in organisations. Knowledge sharing in fact is a phisiological (frequently non fully aware) process in any organisation and teamwork, it is supported by a plurality of individual and social motives and values, but in anycase presupposes some form of trust among the agents and between the agent and the group/organisation.
4
‘By-default’ means that the rule of the agent is “if you have no specific reasons for distrust and suspicion, be confident” and not the other way around: “if you have specific reasons for trust, be confident, otherwise be suspicious”.
Acknowledgements This research has been carried out within the PAR research project of University of Siena and the TICCA project (Trento ITC) as for the trust net and knowledge management, and within the European Alfebiite project as for the basic model of trust. I would like to acknowledge Rosella Postorino’s contribution: during her thesis (I was the supervisor of), she was both collecting data in a real business organization (I thank dott. Thomas Schael for supervising her in this activity) and discussing and working with me about knowledge management theories and trust. Many thanks also to Rino Falcone and Maria Miceli for discussions and comments.
References Busch P.A., Richards, D., Dampney, C. Mapping tacit knowledge flows within organisation X. Department of Computing, Macquarie University, Sydney, Australia Email: {busch, richards, cdampney}@ics.mq.edu.au (2001) Castelfranchi, C., (1998) Modelling Social Action for AI Agents. Artificial Intelligence, 103, 157-82. Castelfranchi, C. (2000). Affective Appraisal vs Cognitive Evaluation in Social Emotions and Interactions. In A. Paiva (ed.) Affective Interactions. Towards a New Generation of Computer Interfaces. Heidelbergh, Springer, LNAI 1814, 76-106 . Castelfranchi, C. (2001) Towards a Cognitive Memetics: Socio-Cognitive Mechanisms for Memes Selection and Spreading. Journal of Memetics.Evolutionary Models of Information Transmission, 5. http://www.cpm.mmu.ac.uk/jom-emit/2001/vol5/castelfranchi_c.html Castelfranchi C., Falcone R., (1998) Principles of trust for MAS: cognitive anatomy, social importance, and quantification, Proceedings of the International Conference on Multi-Agent Systems (ICMAS'98), AAAI Press, pp.7279. Castelfranchi C., Falcone R., (2000), Trust and Control: A Dialectic Link, Applied Artificial Intelligence Journal, Special Issue on "Trust in Agents" Part1, Castelfranchi C., Falcone R., Firozabadi B., Tan Y. (Editors), Taylor and Francis 14 (8), 799-823. Castelfranchi, C. Falcone R. (2002), Social trust Theory, in "The Open Agent Society" Jeremy Pitt (Ed.) John Wiley & Sons, Ltd . Castelfranchi C. and Yao-Hua Tan (eds.) (2001) Trust and Deception in Virtual Societies , Kluwer Academic Publisher, Dordrecht,, ISBN 0-7923-6919-X T. H. Davenport and L.Prusak, Working Knowledge: How Organizations Manage What They Know, Boston, MA: Harvard Business School Press, 1998 Falcone R., Castelfranchi C. (2002), The socio-cognitive dynamics of trust: does trust create trust? In Trust in Cybersocieties: Integrating the Human and Artificial Perspectives R. Falcone, M. Singh, and Y. Tan (Eds.), Springer Verlag) Grundy, J, (2000) Trust in Virtual Teams. www.knowab.co.uk/wbwtrust.pdf Jones, G. R., George, Jennifer M.. 1998. "The Experience and Evolution of Trust: Implications for Cooperation and Teamwork." Academy of Management Review 23 531-546 Nonaka,I. Konno, N. & Toyama,R. (2001) Emergence of “Ba”. In Nonaka, I. & Nishiguchi, T. (eds.) Knowledge Emergence. Social, Technical, and Evolutionary Dimensions of Knowledge Creation. Oxford Univ. Press, N.Y. Nonaka, I., Takeuchi, H., Umemoto, K., (1996) “A theory of organisational knowledge creation” International Journal of Technology Management 11(7/8) :833 – 845 Polanyi, M., (1967) The tacit dimension Routledge & Kegan Paul London U.K. Sichman, J.S., Conte, R., Castelfranchi, C., Demazeau, Y. (1998) A Social Reasoning Mechanism Based On Dependence Networks. In M. Hunhs and M. Singh (Eds.) Readings in Agents. Morgan Kaufmann, S. Francisco, 416-21. Sveiby, K.E. (1996) What is Knowledge Management? Internal Structure Intitiatives. Build Knowledge Sharing Culture. 3M, USA. www.sveiby.com.au/KnowledgeManagement.html