In this chapter, Densham spoke at length about the components required to develop a decision support system capable of handling and utilizing spatial information. He seemed to approach this from the development perspective – how can this type of system be designed to be useful and useable for both programmers and decision-makers. As Dipto already mentioned, two decades have passed since this book was written and GISystems have advanced considerably since – many packages now could be considered at least many aspects of what Densham describes as important in a SDSS. This discussion also reminds me of some statistical debates, where many statisticians have argued the need for a decision-theoretic approach to analysis. In a full Bayesian data analysis (with spatial data or not), a loss function should be specified that relates to the decision made from the results of the analysis – the loss function should capture the “consequences” of making a given decision. An example could be a randomized clinical trial between two drugs (A and B), and based on the results of the trial, drug B is deemed superior based on some criteria. The decision to use one drug over the other will negative consequences (adverse reactions, financial repercussions) and these could be represented in the loss function probabilistically. Presumably one set of repercussions would be worse than the others, and ultimately assist in a more holistic decision-making process beyond the usual statistical analysis performed. I don’t think I’ve ever seen this actually attempted in the literature yet, but with more data and better software/systems, analyses could move in this direction. I think there are probably many areas where researchers, scientists, and many others call for systems that aid in decision-making rather than simply performing a given task or analysis.
References:
-Kathryn