A Decision Support System for Trust Formalization - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 54

factor of the vendors and to deal with variable dependencies in the system by decoupling them using
human expertise in the form of linguistics rules (Oussalah, Nefti, & Eltigani, 2007).
The general trust model proposed in this section is composed of five modules. Four modules will
be used to quantify the trust measure of the four factors identified in our trust model (existence, affili-
ation, policy, and fulfilment) and the fifth module will be the final decision maker to quantify the trust
factor as illustrated in Figure 3.
The inputs of the existence module are the physical existence, people existence, mandatory registra-
tion variable, and the output of the affiliation module. Indeed, as explained earlier in this section, the
mandatory registration variable is dependent on the third party endorsement and membership variables
of the affiliation module. We also note here that the physical existence variable is composed of three
sub-variables, which are the telephone number, the fax number, and the physical address. For the af-
filiation module, the inputs are the third-party endorsement, membership, and portal variables. For the
policy module, the inputs are the customer satisfaction, privacy, and warranty variables. Finally, the
fulfilment module has as inputs the delivery, payment methods, and community comments variables.
The decision maker has as inputs the outputs of the four modules which are Existence_Trust_Index,
Fulfilment_Trust_Index, Policy_Trust_Index and the Affiliation_Trust_Index. The output of this module
is the trust factor of the merchant's Website. In our model, this trust factor will be determined by the
aggregations of the trust indices of all modules.
In the fuzzification phase, two membership functions described by the labels “Low” and “High” will
be used for each variable related to each module. However, we introduce a third membership function
“Average” for the outputs. For the decision maker module, we use all three membership functions for
the inputs and output corresponding to Low, Average, and High degree of trustworthiness of the Web-
site. These membership functions are represented by Gaussians defined by the centre and the standard
deviation parameters. The output values are normalized within the interval (0, 1), with 1 for full trust
Figure 4. Output of the finale decision maker module