Experience in the DITTY project taught us that both points of strength of the proposed DSS framework are effective in practice.
The main limitation of the proposed DSS framework is the discrete approach with respect to the control options and the external factors. When control options and external factors vary in continuous sets, gridding of these sets suffers from a number of drawbacks. First of all, there is no guarantee that the final decision is the truly optimal one, since evaluation of points outside the considered grid could in principle change the outcome of the decision process. This problem could be alleviated by increasing the density of the grid, but this solution is not always possible in practice, especially when model simulations are very time consuming. In this respect, another opportunity offered by the proposed DSS structure is to exploit the feedback branch for the refinement of the solution. One could start from a sparse grid of the set of control options, and then refine iteratively the solution in the most promising zone. Another limitation of the proposed DSS framework is that it heavily relies on the availability of models of the cause–effect relations in the system under study. If models with these characteristics are not available or cannot be developed, the DSS structure of Fig. 1 is not applicable. In this respect, another limitation is that the DITTY-DSS structure was not conceived for model development. Moreover, the accuracy of the DSS outcomes is strictly dependent on the accuracy of the models. Although model uncertainty is considered and presented in the DSS outcomes, inaccurate models produce a high level of uncertainty, thus hindering decision makers from trusting the DSS results. Finally, another limitation could be that the mechanism underlying the DSS structure of Fig. 1 is not completely automatic, rather it requires user's participation and supervision. Although participation of decision makers in the decision process is a prerequisite for developing effective DSSs, we experienced that their involvement in highly technical parts, such as guiding the refinement of the solutions or providing the AHP pairwise comparisons, could repulse them from using the tool.