Satisficing
Games and Decision Making: With Applications
to Engineering and Computer Science
by Wynn
C. Stirling
Cambridge University Press, New York,
2003
268 pp., illus. 20 b/w. Trade, $ 80.00
ISBN: 0-521-81724-2.
Reviewed by John Knight
User-Lab, Birmingham Institute of Art
and Design
John.knight@uce.ac.uk
Wynn C. Stirling is professor of electrical
engineering at Brigham Young University.
The book outlines an alternative approach
to decision making that aims to integrate
"the way humans make decisions into
machines [in order to solve] subtle situations
where group and self interest conflict".
The book has nine sections that take the
reader from Rationality, through Locality,
Praxeology, Equanimity, Uncertainty, Community,
Congruency, Complexity and Meliority.
There is a useful appendix with introductory
texts to some of the background theory.
It should be of interest to "engineers,
computer scientists and mathematicians
working in artificial intelligence and
expert systems". It is not an easy
read but is rewarding to anyone with an
interest in how people and machines make
decisions. Chapters follow a consistent
format starting with background theory
with algebraic proofs, then a description
of Stirlings approach that ends
with examples of decision-making in games.
Stirling suggests that human decision
making is often based on what is "good
enough" rather than what is "optimal".
Starting with Maupertuis "principle
of least action" (p. 1), the author
describes definitions of the optimal and
notes how context and complexity make
it a difficult notion to sustain. He concludes
that "good enough" is an intelligent
response to meeting diverse needs given
the intractable nature of complex problems.
While based on game theory, he is concerned
with bias and that it "is built of
one basic principle: individual self interest"
(p. 19).
The first chapter deals with models of
rationality. The author outlines "the
concept of preference ordering" that
orders options by reflexivity, antisymmetry,
transitivity, and linearity. Choice is
underpinned by "substantive rationality"
that assumes the decision-maker is in
possession of all the options and that
they make an impartial and informed choice.
Alternatively "procedural rationality"
focuses on the quality of the decision-making
process and the rules that are employed
in it. He notes, however, that heuristics
mean that the "solution cannot be
considered in anyway optimal" (p.
8). A third approach is described as "bounded
rationality" that "involves
the exigencies of practical decision-making
and takes into consideration . . . the
constraints that exist in real-world situations"
(p. 10). This idea introduces Herbert
Simons notion of satisficing that
aims to meet expectations rather that
"optimization". The author seeks
a middle ground between the "hyperrational"
and the ad hoc and offers two "desirable
attributes" of adequacy, (knowing
the decision is an optimal compromise)
and sociality that it accommodates others
needs. The chapter concludes by looking
at a number of ways of accommodating group
needs including the "Pareto equilibrium".
These attributes are meant to provide
a "general relaxing of substantive
rationality . . . while maintaining the
integrity of inherent principles . . .
[and bringing] rigour to procedural logic"
(p 11).
The third chapter deals with the "science
of efficient action" (Praxeology)
and concerns "practical activity
and human conduct" (p. 60). The chapter
looks at trade-offs, errors, and abduction
on the basis that "satisficing requires
the formation of dichotomous trade-offs"
(p. 82). Equanimity is discussed in the
following chapter, concluding that satisficing
"should conform to fairness"
(p. 73) based on the availability of resources,
certainty (that the decision is good enough)
and that it does not "foreclose optimality".
Adequacy and consistency are introduced
given that "people may be motivated
by aspirations that are higher than their
ability" (p. 81) and "that it
is essential . . . that due care be given
to the relative weighting" of options
(p. 82).
Chapter Five investigates uncertainty
and quotes Pliny the Elder, "The
only certainty is uncertainty" (p.
89). A number of uncertainty domains are
described including bluffing, the prisoners
dilemma, and the impact of other decision
makers power. The book then returns
to group decision-making and quotes Margolis,
"A decision-maker in a group will
possess a social utility and a private
utility" (p. 140). This strategy
leads on to a discussion of congruency
that deals with agreement seeking, negotiation,
and compromise and touches on ethics and
"social welfare" (p. 161). The
penultimate chapter introduces complexity
suggesting that "[o]ne who suffers
from uncertainty is frustrated by a lack
of knowledge; one who suffers from complexity
is frustrated by a lack of know-how"
(p.169).The final chapter looks at meliority
arguing that "the trouble is, that
humans are quite capable of operating
with a double standard, it is not clear
how to endow an artificial decision-making
entity with such an ability" (p.
205). Stirling has written a fascinating
book on a uniquely human subject: How
do we (and machines) do the right thing.
The book is not aimed at popularising
the subject but is worth the investment.