The current state of the art in the development of decision support system methodologies is focused around the human design of bespoke systems, which are specifically tailored to the particular problem solving environment in hand.
The team at Nottingham and others are carrying out pioneering research at the interface of OR and Computer Science to underpin the development of automated systems to build and design search methodologies. This is a particularly challenging goal which is being addressed in the almost complete absence of a mathematical and theoretical understanding of how to build intelligent systems which are capable of automatically building new systems. This initiative will establish a major cross-institutional effort to explore such issues. A deeper understanding of how and why heuristics work would feed into and inform the development of systems which can automatically and intelligently build and select search methodologies.
Burke EK, Hyde M, Kendall G, Ochoa G, Ozcan E, Woodward J (2009) Exploring hyper-heuristic methodologies with genetic programming. In: Mumford C, Jain L (eds) Computational Intelligence, Intelligent Systems Reference Library,Volume 1, III, 177-201.
Real-world intractable problems demand the use of heuristics if progress is to be made in reasonable time. Therefore, the practical importance of heuristics is unquestionable, and how heuristics are produced then becomes an important scientific question. Many of the current heuristics in use today are the result of years of study by experts with specialist knowledge of the domain area. Therefore, one may pose the question;
Instead of getting experts to design heuristics, perhaps they would be better employed designing a search space of heuristics (i.e. all possible heuristics or a promising subset of heuristics) and a framework in which the heuristics operate, and letting a computer take over the task of searching for the best ones.
This approach shows a clear division of labour; Humans, taking on the innovative and creative task of defining a search space. Computers take on the chore of searching this vast space. Due to the fact that humans often still need to play an important part in this process, we should strictly refer to this methodology as a semi-automated process. One of the advantages of this methodology is that if the problem specification were to change, the experts who engage in hand designing heuristics, would probably have to return to the drawing board, possibly approaching the problem from scratch again. This would also be the situation with the search for automatically designed heuristics, with one important difference. As the search process is automated this would largely reduce the cost of having to create a new set of heuristics. In essence, by employing a method automated at the meta-level, the system could be designed to tune itself to the new problem class presented to it.
A paradigm shift has started to occur in search methodologies over the past few years. Instead of taking the rather short term approach of tackling single problems, there is a growing body of work which is adopting the more long term approach of tackling the general problem, and providing a more general solution.