Professor Ahti Salo initiated the research group on Decision Modeling and Foresight Methodologies, which is based at the Systems Analysis Laboratory of the Helsinki University of Technology.  Decision analysis as well as decision making and risk management have been among the key issues that the group wanted to make progress on. They were run in collaboration with the Helsinki School of Economics as of 1995. The activities have been enabled through basic and applied research projects funded by organizations such as the National Technology Agency (Tekes), the Academy of Finland, Ministries of the Finnish Government, industrial firms and the European Union.

With respect to decision modeling, the most relevant focal research topic is the modeling and exploitation of incomplete information in decision support processes. The topic is motivated by the realization that information on the performance of decision alternatives or the relative importance of the decision criteria can be difficult, impossible or prohibitively expensive to acquire. It is thus pertinent to examine how useful and defensible recommendations can be developed on the basis of the information that can be obtained through decent efforts.

The research group is directed by Prof. Ahti Salo of the Department of Mathematics and Systems Analysis, in Aalto University. The group develops decision analytic methods and tools; it also deploys these in the context of applied research projects carried out in collaboration with decision and policy makers on topics in relation to science, technology and innovation. In terms of methodology, particular attention is given to participatory approaches for technology foresight and assessment, as well as methods for addressing uncertainties in innovation management (e.g., prioritization of research themes). The research group consisted of Prof. Salo, four full-time researchers, three Master’s Thesis students, and two part-time research assistants.

In recent years, the main projects in relation to S&T decision and policy making were “robust portfolio modeling in innovation management” (a research project on the use of portfolio approaches in technology foresight, funded by Tekes and industrial firms); “A participatory exploration of business opportunities for the environmental cluster of Ylä-Savo” (carried out together with Savonia Polytechnic); “A resource allocation model for standardization activities at a telecommunication company” (Nokia); and “Decision Analysis in Project Portfolio Optimization” (Academy of Finland).

Elaborating in further details pertaining to the above projects, the methods being developed and utilized were built on the well-established frameworks for value tree analysis and some hierarchical weighting models.  Four important methods can be investigated:

• PAIRS & PRIME: these methods are to accommodate incomplete information about the model parameters by way of set inclusion.  For example, the lower and upper bounds may be placed on the alternatives’ scores and criteria weights may be constrained through linear constraints. With the assistance of relevant dominance concepts and decision rules, such information can be synthesized to convey a) which alternatives can be definitely recommended on the basis of all feasible combinations of model parameters; and b) what alternatives are supported by decision rules that can transform incomplete information into corresponding decision recommendations.
• Preference Programming: methods are promising as they support interactive learning processes. They also can reduce the costs of information elicitation and may increase decision modeling commitment to the particular decision support process.
• Rank Inclusion in Criteria Hierarchies (RICH): the method extends preference programming methods to the analysis of incomplete ordinal information. A related decision support tool, called RICH Decisions, also emerged on the basis of RICH to be employed in the selection of risk management methods and the development of priorities for a research program.  Such practices were found being applied to an energy utility and a program in Scandinavia, respectively.   Moreover, the RICHER method, or so called RICH with Extended Rankings, offers even more flexible preference elicitation modes.  It applies such modes of RICH to the comparison of alternatives.  Thus, for any given subset of alternatives, the model may specify a subset of rankings that these alternatives may assume in relation to a single evaluation criterion, several criteria, or even all criteria.
• Robust Portfolio Modeling (RPM): the development came from the works at the juncture of preference programming and multi-criteria project portfolio selection.  The method makes it possible for foresight manpower to determine a) which core projects are included in all non-dominated portfolios; b) which exterior projects are not included in any non-dominated portfolios; c) which intermediate projects are included in some but not all non-dominated portfolios.  Based on the analyses, the model can be advised to select core project while rejecting trivial ones.  Furthermore, subsequent information elicitation efforts can be focused on intermediate projects, which helps reduce the costs of information elicitation. Contrasting to the earlier literature on robustness, RPM is considered unique in that it offers decision recommendations about individual projects instead of offering a ‘single’ optimal portfolio on some selected robustness measure, e.g. max-min. This makes it appropriate for interactive group decision support processes where considerations that are less amenable to formal modeling efforts can be addressed through judgmental considerations, e.g. project interactions.   Pragmatically, a wide range of RPM projects have been carried out in various contexts, including road asset management, formulation of a product strategy in a high-tech firm, scanning of innovation ideas, development of a strategic research agenda, and ex post evaluation of an innovation program.  In relation to RPM, the group is actively working on the development of decision support tools for the computation (RPM-Solver) and internet-based dissemination of RPM results (RPM-Explorer).