MishMash

What is uncertainty?

A couple of recent activities have forced me to explore my discomfort with with the way the word uncertainty is used in flood risk management. One of these activities was a refinement and extension of the wiki on methods for handling uncertainty in flood risk management that I have been involved in putting together as part of the Flood Risk Management Research Consortium. The other is an ongoing attempt to write a paper, which also deals with, among other things, uncertainty. I find writing incredibly hard, not least because I can't stop thinking about things like this.

I have assumed, since I first thought about it, that certain and its various cognates (certainty, uncertain, uncertainty, ...) referred originally to a state of mind, and that through a process of metonymic transfer they had come to be used to refer to the object of that state of mind: the thing about or because of which one was, or was not, certain. A trawl through the OED, to which the University subscribes, presumably hoping that doing so will result in a net productivity gain among its employees, suggests that this is not the case.

For all of these words, the first definition (and I presume order is significant) given by OED is related to something being "determined, fixed, settled". For certainty, the mental state which I assumed was the primary meaning appears only in the fourth, "the quality or fact of being (objectively) certain", and fifth, "the quality or state of being subjectively certain; assurance, confidence; absence of doubt or hesitation" definitions.

The etymology given for certain makes things clearer

a. OF. certain (= Pr. certan, Sp. and It. certano), repr. late L. or Romanic type certan-us, certan-o, f. cert-us determined, settled, sure, orig. pa. pple. of cern-ere to decide, determine, etc. The sense-development had taken place already with L. certus.

(OED)

So the root meaning of certain would appear to be "having been decided/determined", and the metonymic transfer, if anything, proceeded in the opposite direction from that which I had supposed, from the fact of my having decided something to my state of mind having done so.

In launching myself into the OED, I was looking for justification for using uncertainty in a constrained sense to refer only to a state of mind. I didn't find it, which leaves my original problem unsolved. To design software which supports a more explicit treatment of uncertainty in flood risk management requires that we address all of the many aspects of uncertainty, and these go far beyond the numerical-computational treatment provided by the tools of quantitative uncertainty analysis. To do so, or even to explain the need to do so, we must be able to distinguish these many aspects in both thought and language, in turn requiring the use of a vocabulary which recognises these distinctions. Using uncertainty to refer indisciminately to both the object of a state of mind and that state itself renders clarity of thought, and with it communication, impossible.

As far as the laws of mathematics refer to reality, they are not certain; and as far as they are certain, they do not refer to reality.

(Albert Einstein)

True objective certainty is only possible in the context of formal systems such as mathematics and logic. Everything else, including the results of modern science, is subjective. The use of certitude, according to the OED, has since Hobbes been increasingly restricted to "the quality of being subjectively certain" [my emphasis], and is thus the opposite of doubt. Science allows us to develop justified certitude, which might tend towards objective certainty but cannot ever get us there. Flood risk management draws heavily on modern-scientific methods and results; this injects a certain amount of "objectivity" into the process, but in the end most decisions still have much more to do with certitude, subjective certainty, than with objective certainty.

Perhaps, then, I can use doubt and certitude to refer to the states of mind of participants in the decision-making process and uncertainty, in the manner familiar to most of my potential readers, to refer to the objects which, by their uncertainty, inspire doubt.

Khatibi on flood forecasting and warning, model categorisation

An interesting article from Rahman Khatibi [thanks Rich] about the role of flood forecasting and warning in sustainability of flood defence.

The core of the paper, if I understand correctly, is

1. The idea that structural defence ("hard", "preventative", "bounded", "finite" measures) tend to lead to a positive feedback cycle where value at risk increases as probability of flooding is reduced.
2. Looking at flood forecasting and warning as part of the overall system of response to flood risk encourages a more balanced system where the natural negative feedback (risk leads to people living elsewhere) isn't so severely suppressed.

I like this idea, and it fits with my earlier comment that

Systems without negative feedback tend to spiral out of control. [Hamish Harvey]

Khatibi argues that particular types (categories) of model are a better fit to particular types of physical system. Intuitively apparent, but establishing with any confidence what those appropriate combinations are would surely require a very large amount of data.

… research activities are promoted by  the NFWC to produce prescribed Pf values [measures of the probability that an accurate forecast is made and disseminated] for each category of modelling techniques. [Rahman Khatibi]

How well can these values be defined? How clear a categorisation of physical system type is possible? Many different categorisation schemes are clearly possible, such as

… characterising these systems based on the significance of a particular term(s) in the Saint-Venant equations … using the lead-time as the basis for classification of catchment systems … there are other possible approaches [Rahman Khatibi]

This all suggests that a faceted classification system is needed, and this in turn provides a potential combinatorial explosion in the number of Pf values to be calculated.

Update 2006-02-24: broken links fixed.