Abstract
Decision support systems (DSS) become more and more complex to deal with decision-making problems related to the current conjuncture. Several agent-oriented methodologies have been proposed to design distributed decision support systems and improve the decision process. This paper presents a comparative study of these existing methodologies (based on a set of relevant criteria like the adaptability, reliability and decision objective) and identifies a set of issues, notably the lack of faults tolerance mechanism and the absence of the formal verification. Hence, we propose a work-in-progress approach (also based on multi-agent paradigm) called ADS^2 adapted for distributed decision support systems. We also present how ADS^2 will allow to remove the global control, to reinforce the adaptability and to reduce developers burden thanks to the utilization of model-driven engineering settings.