What is Decision Analysis

Decision science was developed in the 1960s and 1970s at Harvard, Stanford, MIT, Chicago, Michigan, and other major universities (see Bibliography). It is generally considered a branch of the engineering discipline of Operations Research, but also has links to economics, mathematics and psychology.  We make a distinction between decision science – the study of how to make more effective decisions and decision analysis – using decision science to analyze a particular decision.  Logical Decisions makes it easy to use decision analysis for comparing alternatives.

Types of problems addressed by Decision Analysts

Decision analysis practitioners generally work or two broad classes of problems. The first class involves sequential decisions where uncertainties and probabilistic dependencies play a large role. The second class involves one-time decisions where a group of alternatives must be compared on the basis of multiple (and possibly competing) goals and objectives. The primary difficulty of these types of decisions is in creating a “value model” that allows explicit comparisons between alternatives that differ in many ways.

Sequential decisions are modeled using tools called “decision trees” and “influence diagrams”. These tools allow decision analysts to explicitly organize the sequence of decisions and the uncertainties related to the decision.

Multi-objective decisions are modeled by creating a “multi-measure utility function” that allows an alternative’s overall desirability to be computed based on how it performs on a set of evaluation measures. The study of how to analyze multi-objective decisions is called “multi-attribute utility theory” or MAUT.  In a decision analysis, MAUT is implemented using a set of techniques that are collectively called Multi-Objective Decision Analysis (MODA).

Logical Decisions is designed to address multi-objective problems, not sequential decisions. Several excellent software tools have been developed to create decision trees and influence diagrams for sequential decisions. See the links section if you would like to look at their web pages.

The decision analysis process

The essence of decision analysis is to break complicated decisions down into small pieces that you can deal with individually and then recombine logically. A key goal of decision analysis is to make a clear distinction between the choices that you can make (the alternatives), the characteristics of the alternatives (quantified by the measures) and the relative desirability of different sets of characteristics (preferences). These distinctions let you clearly separate the objective and subjective parts of your decision. The alternatives and the way they are quantified using the measures are relatively objective. Even if there are uncertainties in the levels of the measures, it is usually possible to come to an agreement about how to characterize those uncertainties.

On the other hand, the relative importances (weights) of the different measures, the interactions between them, and attitudes toward risk are inherently subjective. Reasonable people can have wide disagreements on these subjects. You can’t generally eliminate these subjective parts of a decision. Logical Decisions provides methods for logically dealing with both the objective and subjective parts of a decision while keeping them well separated.

The decision analysis method for multi-objective problems can be described by the following steps:

1) Identify the alternatives to be ranked.
2) Clarify the goals and objectives that should be met by choosing the top-ranking alternative.
3) Identify measures to quantify how well the alternatives meet the goals and objectives.
4) Quantify the level for each measure for each alternative.
5) Quantify preferences about different levels of the measures.
6) Rank the alternatives by combining information from steps (4) and (5).
7) Do “sensitivity analysis” to see the effects on the results of changes in measure levels or preferences.

Bibliography

An annotated list of books and articles relevant to decision analysis and MAUT in particular.

Links to other Decision Analysis Sites

Copyright © 2021 Logical Decisions

 7,598 total views,  4 views today