Trust Modeling Based On Associative Mind-Maps

In our daily life, we make the preference of trust almost every day. With the increasingly popularity of internet, internet-oriented researches, businesses and other activities are occurring more and more frequently—raising the issue of trust prominently. Thereby how to build confident trust model under the virtual environment becomes a hot research topic in broad areas, such as psychology, sociology, economics, computer science, and so forth.Due to dynamic environmental interaction and other uncertain factors, interactive entities of transaction need to know their trust relationship. Generally speaking, trust includes not only the identity of entity, but also the behavior of entity. Accordingly, traditional security mechanisms, such as encryption and access control, can be used to provide authorization and authentication—solving the issue of trust on identity. However, those approaches can not deal successfully with the trust on entities' behavior because of uncertain guarantee on provided services. Many current trust models are mainly based on the experience of previous transactions, while the trust relationship is assessed by other analogy entities.However, these trust models are only restricted into specific applications, because they usually focuses on the relationships between computers and computers, or the relationships between computers and users. Meanwhile, these models rarely take account of the psychological changes from the users' side. In other words, they ignore the reason why the outcome of such trust would bring forward.Therefore, this thesis is aimed at proposing a different trust model, in which data mining methods are facilitated to help study the trust of entities' interaction. Based on dialogue between humans in the research of trust relationship, MIT researchers proposed SOM (Self Organizing Map). Although SOM is efficiently to understand the psychological changes in the process of dialogue, the overall trust model is not well studied. Specifically, SOM only considers about situational trust on a point of changed view of human beings. However, neither did their model measure the trust between humans, nor did it explain how to build such mathematical model, such as renewable equation.Inspired by the idea from SOM, this thesis comprehensively describes human trust, and establishes the trust model by considering each person as self-organizing mind-maps. Specifically, under the structure of self-organizing mind-maps, each person has its own views on the world map, so called "self map", together with the views on other people, so-called "other map". Moreover, map of each node denotes a favorite expression over one item. While conversations are facilitated as training set in order to update these maps. After some conversations, conclusion of trusting the speaker or not is drawn. Self-contradictions or disagreements of the speaker might make listener stop the talk.In order to measure the trust relationship between entities, the model not only takes advantage of previous transaction experience, but also uses the assessment over other entities. This model also takes account of the psychological change of users—ensuring a broader range of applications.The model is implemented in Java and tested in some scenarios, to determine whether the behavior of this model performs as expected as a real human being does. It is run in command line after specifying configuration files.At current stage, this thesis only consider conversations in simple forms such that each sentence only refers to preference, such as like or dislike, over one particular item. It also raises the need of further improvement in future, such as how to extend the conversation to universal human conversation; how to extend the conversation into common human experience, as well as the improvement on SOM model itself.