Decisions
© Copyright 2009, 2012 Herbert J. Bernstein
Decision making is a very complex subject. See the
Wikipedia article on
Decision Making as a starting point for further
study.
- What is a decision?
- A decision is a choice among alternatives
- e.g. turn left vs. turn right
- e.g. get up vs. stay in bed
- e.g. paint that wall bright white, beige, cream, off-white, ....
- e.g. get married, stay single, have children, don't, ...
- e.g. invest in mortgage back securities or get a CD
- e.g. hire this person or that person
- e.g. keep working for your boss or quit
- e.g. produce trucks or produce cars
- e.g. start a war or make peace
- Someone or something must be the decision maker (the decider?)
- Multiple people or entities (the stakeholders) may be impacted by a
decision
- Not making a decision is itself a decision
- What makes a decision "good"?
- "good" may not be quantifiable
- "bad" may not be quantifiable
- What actually happens after the decision may be very different
from what we thought would happen before we made the decision
- Suppose we can quantify the decision
- The elements of the decision may be deterministicly quantifiable
or may be random variables.
- If elements of the decision are random, we may need statistical
techniques to make the decision.
- Possibly some greater benefit is derived from the chosen
alternative than from the other alternatives?
- Possibly the losses (the costs) from the chosen alternative are
less than from the alternatives
- Our choices may be influenced by and influence choices made by
others.
- Strategies in Making Decisions
- Ignore the problem
- Maybe it will go away
- Maybe it won't
- von Neumann and
Morgenstern Game Theory
- Cost-Benefit
Analysis
- Reduce all elements of the problem to dollars
- Try to maximize the ratio of overall benefit to total cost
- Try an alternative, then try another, ... see how they come out
- Get a time machine
- Model (simulate) the system and try the alternatives in the
simulation
- Models may be physical or
conceptual (logical).
- Diagrams can help in conceptual models Process
maps
- Models are intended to extract the salient features of the system
- Models may miss important aspects of the system
- Models may be qualitative or quantitative.
- Qantitative models may be deterministic or
stochastic (based on random variables).
- To make a decision, we try to maximize (or minimize) an objective function (e.g.
profit)
- For uncertain variables, probability distributions are used
- Instead of making one decision, we may need to make a continuing series
of decisions based on changing varaibles, see
Statistical_process_control.
- For linear sets of inequalities, the simplex method is
used