[JLJ note: Katsenelinboigen has some great ideas here which transfer from the field of economics. Playing
the positional game of chess is similar to economic planning in an uncertain environment.]
p.15 I want to underscore the extensive parallels between decision making in chess and in an economy.
p.50 Devising methods of a system's operation under the conditions of uncertainty is much more complicated.
Two basic approaches to this problem have been charted. The first one attempts to reduce the problem of uncertainty
to probability, and the second calls for the creation of flexible structures.
p.52-53 I stressed the need to create potential in adaptive systems. In this context, potential
signifies a certain state that improves the system's position - a position that, on the one hand, certainly affects development
but without knowing the probability of the influence it exerts but, on the other hand, purports to be more certain than chaos.
My aspirations have to do with the creation of such potentials. The impact exerted by the potential
upon future development, as manifest in the respective degree of uncertainty, is called the degree of predisposition.
p.62 Potential is a measure of the system's predisposition to development. This approach
is different from the probabilistic one that deals, albeit in terms of frequencies, with fully objective connections between
the present and future states of the system. A potential forms a structure aimed at inducing the events in the environment
in which the system is immersed to the system's advantage, preparing the system to channel unexpected outcomes in a way that
is favorable to the system, and absorbing or reducing the shocks of unexpected events harmful to the system.
p.66 Ackoff (1981) suggested that the planning process be divided into three stages, each assigned its respective
criterion: ideals, objectives, and goals. That is how he defines them:
1. Goals: those ends that we can expect to attain within the period covered by the planning.
2. Objectives: those ends that we do not expect to attain within the period planned for but which
we hope to attain later, and toward which we believe progress is possible within the period planned for.
3. Ideals: those ends that are believed to be unattainable but toward which we believe progress
is possible during and after the period planned for (p.63).
It seems to me other important categories termed aims and targets can be added to the above groups. These
categories arise from the need to introduce intermediate goals directly interconnected with each other... Aims, in turn, differ
from targets as far as the completeness of the set end. Whereas aims are specified in terms of all the important parameters
pertaining to a given stage (for instance, profit), targets refer to individual parameters that are currently considered particularly
valuable (for instance, output growth).
p.67 Objectives of the second stage are less remote than ideals, and the means of reaching them are visible.
However, this should not be at all construed to imply that from our position in the present we know precisely how to create
a program for obtaining such objectives. Therefore, it is important that we create initial conditions at the second
stage that would facilitate the most efficacious development of the system in an unknown future. In other words,
these initial conditions should be such as to make it possible to steer the future development of the system in a desirable
direction, that is, create the greatest predisposition of the system toward future development in the direction of
the chosen ideal. A structure that measures up to the requirements we set for the conditions in question
was called a potential; and the process of its formation, potentiating.
p.70 The subjective element arises not because the set of positional
parameters and their valuations are formed based on a player's intuition. Rather, the choice of relevant parameters
[in an evaluation function] depends on the actual executor of the position, that is, the particular strengths and weaknesses
of a given player. The role of the executor becomes vital because the actual realization of the position is not known beforehand,
so future moves will have to be made based on the contingent situation at hand.
p.71 the way to improve chess algorithms is to modify the straightforward approach to positional style and
attempt to generate satisfactory, if inexact, solutions by posing more limited goals and methods of their achievement... To
sum up, a positional style (a style posing limited goals), incorporating composites as goals for a given stage, makes the
search more economic as compared with the total search, which takes into account all the established parameters.
p.72-73 Positional style, however, emphasizes the creation of preconditions necessary for
a combination. It does not negate combinational style; it merely postpones it until the position is ripe for combinational
play.
p.132 One important variant of a program-like class of methods is the robustness principle introduced by
Jonathan Rosenhead, Martin Elton, and Shiv Gupta (1972). It pertains to cases where information regarding the various strategies
is available but the implementation takes place under uncertainty. This principle embodies the idea of flexible structures.
A class of problems for which (profit-maximizing) optimality is particularly inappropriate is that of strategic
planning, a special case of decision-making under uncertainty. A plan consists of a number of currently preferred
future decisions, and only one decision or group of decisions which must be irrevocably implemented at this stage. The subsequent
preferred decisions which constitute the plan can each be reviewed in the light of up-to-date information when their commitment
times arrive. In this situation an appropriate criterion for the initial decision is one based on the degree of useful flexibility
for future decisions which will be maintained.
Robustness is such a criterion of flexibility in achieving near-optimal solutions
in conditions of uncertainty (p.428).[see also Rosenhead, 1980]
p.138 the principle essentially aims to augment the potential at a given stage, that is, to increase
its impact upon the development of the system directing it along the course selected based on one's beliefs. Thus,
from the functional point of view, uncovering predispositions means finding a state that will ensure that the system
is prepared to face an uncertain future.
p.138 There are two modes to form predispositions and preserve this holistic effect: reactive and selective.
The first method is based on a set of actions that represents isolated rules, heuristics. The second method involves a search
for a scalar value denoting the system's predisposition.
p.252-253 The analogy between heuristics in chess and firm [business] behavior is quite cogently formulated
in a brochure by Rudyard Istvan (1984) where he says:
Rules of thumb are guiding principles that, while never true because of oversimplification,
point reliably to the probable direction of action. Experience builds rules by remembered results of trial and error.
All good chess programs depend on rules of thumb to simplify calculations. Some
moves are not examined, because they almost never pay. Very possibly, the brilliantly innovative winning move will be excluded
by such a process. But the frustration of novice players matched against the simplest chess programs demonstrate their power.
Most of the major ideas in business strategy are conceptual rules of thumb about economic
competition... (p.5)
