A direct reputation model for virtual organisations formation

Recently, a large number of new collaborative, networked organisations have emerged, having as motivator the explosive progress in computer networks and communication systems, but also as a reaction to market pressures that demand customised, high quality products and services at lower costs and, at the same time, shorter production and marketing times. Promising grater flexibility, resource optimisation and responsiveness in competitive open environments, Virtual Organisations are presented as an example of this trend that has pervaded not only business domains but other areas such as e-science. What distinguishes Virtual Organisations from other forms of organisation is the full mutual dependence of their members to achieve their goal and therefore the need for cooperation. However, open environments, where Virtual Organisations are embedded, involve organisations and individuals that do not necessarily share the same objectives and interests, which they might not know in advance, and where they might not trust each other, but should work together and help each other to achieve a common goal. One of the key omissions in the computational representation of Virtual Organisations relates to the need to take into account more subjective facets like the reputation of the individuals, that helps to cope with the heterogeneity, the autonomy and the diversity of interests among their members. This thesis addresses the introduction of reputation into Virtual Organisations, by providing a reputation model based on the adaptive evaluation of direct experiences to identify trustworthy members of a Virtual Organisation. This thesis makes three main contributions. First, it provides the means for designers of VOs to identify situations where reputation is required. This is needed since members of a Virtual Organisation, which are also self-interested, must cooperate with each other, and reputation represents a way to identify the possible deceivers and the more attrac tive partners in the environment. Second, it provides a reputation model for Virtual Organisation partners that does not require individuals to interact extensively to detect possible abusive or honest partners. Third, it provides a comparative analysis of current reputation systems used in e-commerce and Multi-Agent Systems, and of the new model proposed.