p.3 Over the years, experience with the influence diagram has shown
that it is an effective means for communicating with both decision makers and computers. The influence diagram has
proved to be a new "tool of thought" that can facilitate the formulation, assessment, and evaluation of decision
problems.
p.21 The influence diagram paradigm can be used in special forms to frame a decision, to elicit
the often confused knowledge about uncertain quantities that resides in the human mind, and to both explore and characterize
the values that govern the decision. No other current tool possesses its span of application, subtlety of interpretation,
and power of computation. Almost every discussion of human action or knowledge can be elucidated by its use.
However, its most important feature is that its simple appearance belies both its power and its subtlety.
p.47 The influence diagram has also proven to be one of the easiest ways of communicating a decision
problem to a computer.
p.68 An influence diagram can be viewed as an economical scheme for representing conditional independence
relationships and for deducing new independencies from those used in the construction of the diagram. The nodes in
the diagram represent variables in some domain and its topology is specified by a list of qualitative judgements elicited
from an expert in this domain. The specification list designates to each variable v a set of parents judged
to have direct influence on v, and this amounts to asserting that, given its parameter-set, each variable
is independent of its other predecessors in some total order of the variables.
p.125 The construction of statistical models is a skilled activity that draws on three fundamental varieties
of knowledge and understanding. The first is the ability to recognize and name entities,
such as a system (e.g. a nuclear power plant) or a set of statistical units (e.g. a human population), that are the
basic building blocks of the model, and to understand how such entities and subentities relate and interlock
to form the complete system or mosaic represented by a model. The second variety reflects perception
and understanding of variables or attributes that characterize similarities and differences among entities of the same type,
such as the myriad properties of a nuclear power plant recognized by nuclear engineers, or similarly diverse properties of
populations and persons recognized by epidemiologists. Thirdly, a skilled modeler has a substantial
body of theoretical and empirical knowledge, including causal understanding about factors that govern or influence temporal
processes of change. The third variety of knowledge makes possible credible choices of the mathematical representations
of deterministic and probabilistic relations that give inferential power to a model.
p.171-172 This is why expectation is useful as a basic quantity. You can make exactly that collection
of expectation statements about a problem which seem to you to be both directly relevant and also within your capability...
I have never understood why so many people who are prepared to accept probability as basic find it so hard to treat expectation
as basic... If you can find a satisfactory probability formulation, then use it. But otherwise,
you have a choice, and the expectation-based system will typically be more flexible.
