what-when-how
In Depth Tutorials and Information
it reaches a level when it can no longer be reduced. Irreducible uncertainties are
inherent uncertainties due to the natural complexity of the subject matter. We can
distinguish the following types of uncertainty (Walker, 2003):
CausalUncertainty
: When scientists draw causal links between different parts of
the system, or between a specific input and an output, there is an uncertainty in the
causal link. For instance, the relationship between air pollution concentration and
respiratory problems is associated with causal uncertainty, given that the same air
pollution concentrations can result in diferent levels of respiratory problems. his
occurs because other, sometimes unknown, factors can influence the causal link.
here is also the important difference between correlation and causation, in that
an existing correlation does not necessarily indicate causation. Another source of
causal uncertainty is the existence of feedback loops in a system. Causal uncertainty
is strongly dependent on the “mental map” of the person drawing the linkages.
Measurement Uncertainty
: When measuring physical or social phenomena,
there are two types of measurement uncertainty that can arise. he irst is the
reliability of the measurement, and the second is its validity. Reliability refers to
the repeatability of the process of measurement, or its “precision,” whereas validity
refers to the consistency of the measurement with other sources of data obtained in
a diferent ways or its “accuracy.” he acceptable imprecision and inaccuracy in the
case of different subject matters can be very different. For instance, the acceptable
inaccuracy for a weather forecast is different from the inaccuracy of measurements
for the leakage rate of a nuclear waste containment casket, given the different levels
of risk involved. herefore, deining the acceptable uncertainty in measurements is
a rather subjective decision.
SamplingUncertainty
: It is practically impossible to measure all parts of a given
system. Measurements are usually made for a limited sample and generalized over
the entire system. Such generalization beyond the sample gives rise to sampling
uncertainty. Making an inference from sample data to a conclusion about the entire
system creates the possibility that error will be introduced because the sample does
not adequately represent that system.
FutureUncertainty
: he future can unfold in unpredictable ways, and future
developments can impact the external environment of a system or its internal struc-
ture in ways that cannot be anticipated. his type of uncertainty is probably one
of the most challenging, given that there is little control over the future. However,
it is possible to anticipate a wide range of future developments and simulate the
effect of particular decisions or developments in a system across these potential
futures. In sociotechnical systems, the effects of new technologies often cannot
be adequately determined
apriori
. Collingridge (1980) indicates that, historically,
as technologies have developed and matured, negative effects have often become
evident that could not have been anticipated initially (automobile emissions or
nuclear power accidents and waste disposal). Despite this ignorance, a decision
has to be made today.
