9781118041581

Summary 569

SUMMARY

Decision-Making Principles

1. Any new information source is potentially valuable in influencing
   forecasts of uncertain events and thus guiding better decisions.
   a. The decision maker should acquire additional information if and only
   if its expected value (in making better decisions) exceeds its cost.
   b. The decision maker should not commit needlessly to a single course
   of action for the foreseeable future. By crossing bridges only when he
   or she comes to them, the decision maker can expect to make better-
   informed decisions.


2. The logic of Bayes’ theorem shows that any probability forecast is based
   on a combination of the decision maker’s previous information (his or
   her prior probabilities) and newly acquired information.
   a. The greater the initial degree of uncertainty or the stronger the new
   evidence, the greater the subsequent probability revision.
   b. Information is valueless if it results in no probability revisions or, even
   with such revisions, it does not change the individual’s optimal decisions.


3. Although most business and government decision makers rely on
   informal prediction methods, evidence shows that these methods are
   prone to error and bias.


4. Optimal search involves sequential decisions in which the manager seeks
   out and evaluates alternatives from which he or she ultimately will choose.
   The manager’s search propensity increases the greater the marginal benefit
   of search and the smaller the marginal cost.


5. In search situations, the decision maker is better off the greater the
   number of alternatives from which to choose.

Nuts and Bolts

1. Calculation of revised probabilities is accomplished using a joint
   probability table (with rows listing the test results and columns listing the
   uncertain outcomes) or employing Bayes’ theorem.


2. New information (such as a test result) appears at the beginning of the
   decision tree, prior to the main decision. A decision square follows each
   possible test outcome.


3. As always, the decision tree is solved by averaging out and folding back.
   The expected profit computed at the beginning of the tree measures
   the expected benefit from making decisions based on the acquired
   information.

c13TheValueofInformation.qxd 9/26/11 11:02 AM Page 569