The case study used here to illustrate the notion voodoo decision theory examines a non-probabilistic decision theory for severe uncertainty called info-gap decision theory, which is claimed to be new, and what is more, particularly suitable for dealing with a severe uncertainty that is characterized by
- A vast (e.g. unbounded) uncertainty space.
- A poor point estimate that can be substantially wrong.
- A likelihood-free quantification of uncertainty.
Indeed, its advocates hail it as an approach to robust decision under severe uncertainty that is radically different from all current theories for decision under such conditions.
The fact of the matter is, however, that its robustness model is a simple radius of stability model (circa 1960). This means that, like any robustness theory that is based on a radius of stability model, info-gap decision theory is not designed to seek robustness against severe uncertainty, but rather (local) robustness against small perturbations in the nominal value of the parameter of interest.
The question therefore arising is this:
How can an approach for tackling a severe uncertainty of this type, that is so patently flawed, possibly be advocated by senior risk analysts/scholars and academic research centers?
The objective of this project is to address this intriguing question through a comprehensive critique of this theory.
In a nutshell, the main thesis that will be argued here is that this is due to the fooled by robustness effect, where local robustness is mistaken for global robustness. In the case of info-gap decision theory, this blurring of the distinction between local and global robustness leads to the misconstruction of local robustness as robustness to severe uncertainty.
The case of info-gap decision theory is further exacerbated by the failure to appreciate that info-gap’s robustness model is in fact a very simple instance of Wald’s famous Maximin model (circa 1940), hence, a very simple robust optimization model.