NON-MONOTONIC REASONING FOR INTELLIGENT PAYMENT SERVICES

The payment agent in IAWPS employs possibilistic deduction to infer if a settlement service is a feasible solution (i.e., candidate) according to its preference (e.g., cheap, less risky, meeting deadline, etc.). When a settlement service is evaluated, its characteristics are encoded as possibilistic formulae and added to the payment agent’s personalized knowledge base as assumptions. If the payment agent’s knowledge base entails a settlement service as a candidate service (i.e., ), will be added to the set of feasible solutions. This scenario is similar to the symbolic approach of finding a feasible plan or schedule through assumption based reasoning (Kraetzschmar, 1997). In MXFRMS (Kraetzschmar, 1997), each possible plan is added to the agent’s knowledge base as assumption. If the plan is consistent with the knowledge base where the constraint about the planning process is stored, the plan is considered feasible. However, for the payment agent, some of the characteristics of a service may lead to its consideration as a candidate service, whereas other service qualities may remind the agent to exclude it from consideration. This is a general problem in multiple criteria decision making (Zeleny, 1982). In logical term, inconsistency () may exist in the resulting possibilistic knowledge base . Therefore, the nontrivial possibilistic deduction is used so that the payment agent can draw sensible conclusions based on the most certain part of even though some of its elements are contradictory to each other. To choose the optimal settlement option(s) from , the valuation will be compared for each . A candidate settlement service that receives the highest valuation will be chosen as the settlement service since the payment agent is most certain that it should be a candidate service.