Techniques for Generating/Eliciting a Hierarchy or Network of Objectives

When structuring an ill-defined problem, eliciting the underlying objectives is fundamental. The task is subjective, but it certainly help to follow a formalized strategy. Evans p. 45 lists some existing approaches.

Clemen and Reilly (2017) p. 52 ask: "How do we first separate means and fundamental objectives and then construct the fundamental-objectives hierarchy and the means-objectives network?" and suggest four guiding questions, techniques, for organizing means and fundamental objectives:

• Why Is That Important? (WITI)
• How can this objective be achieved?
• What do you mean by that?
• Of what more general objective is this an aspect?

Buede (1986) proposes two structuring methods: top down and bottom-up approaches.

• Top-down method: it's objective-driven (close to Keeney and Raiffa 1976); "the analyst begins by ascertaining the global objectives of the decision maker and proceeds to a value structure by subdividing the objectives, sub-objectives, and so forth until a final set of attributes is obtained."
• Bottom-up approach: it's alternative-driven. "The analyst begins by questioning the decision maker for a reasonable set of alternatives, each of which might solve the problem. Once the alternatives are defined, the analyst generates a value structure by probing the decision maker for the major differences between the identified alternatives. The analyst then categorizes these differences into groups corresponding to objectives so that a hierarchical value structure can be systematically constructed. The identified differences comprise the set of attributes."

Manheim and Hall (1967) reject the method which hinge on too much mathematization of decision components (e.g. cost-benefit analysis, and utility theory: von Neumann and Morgenstern) because they "it tends to obfuscate the issues of choice by concealing them in the mathematics of utility". Which is reasonable. The method is:

1. goal fabric analysis: list all the known goals for the project and then identifying the various relations among the goals.
2. utilize the goal fabric analysis to rank the alternatives. This entails mapping each new alternative onto the goal fabric (i. e., predicting the performance of the alternative with respect to some of the goals) and then, using this mapped information and the structure of the goal fabric, comparing the new alternative with one previously ranked, to fit the new one into the ranking.

"The method operates on only two alternatives at a time. Any attempt to formulate a list of goals runs into problems of consistency, overlap and varying degrees of detail of the goals. These problems are usually approached by trying to state all the goals in a uniform way. In the method we propose, however, this is precisely what is not done: the list of goals can contain overlap and different degrees of detail. We propose analyzing the list to identify explicitly all the relations among these "non-uniform" goals. The goal analysis is intended to structure the goals by identifying the relations among them that are relevant to evaluation of the alternatives." There are four relations of importance (the first two guide expansion of the goals list in order to clarify the vague goals)

1. specification: entails explaining in more detail what we mean by the general goal. 
2. means-end: describes how a goal can be accomplished
3. value-wise dependence: are those goals that can be evaluated only in conjunction with other goals.
4. value-wise independence: can be evaluated on their own, without regard to any other goals

Reference
Evans, G. (2017) Multiple Criteria Decision Analysis for Industrial Engineering. CRC.
Clemen and Reilly (2014) Making Hard Decisions with DecisionTools. CENGAGE
Keeney (1992), Value Focused Thinking. Harvard, p. 57ff
Keneey and Raiffa (1976) p. 31ff
Buede (1986) Structuring Value Attributes. Interfaces, 16. 2
Manheim and Hall (1967) Abstract representations of goals. MIT