Validation of Multiattribute Utility Procedures
Abstract
Validation topics were examined through both a real-world and laboratory setting. In the laboratory experiments we taught subjects additive and multiplicative value functions via outcome feedback. We found that standard MAUM procedures recovered the taught functions. We also found behavioral differences between value and utility elicitation gauge the validity of alternative multiattribute utility elicitation techniques. The results indicate that subjects learned the value functions very well, independently of whether the problem involved two or four attributes, equal or unequal weights, additive or multiplicative functions. Riskless value and risky utility elicitation methods were able to identify the structural properties of the taught models (additive vs. multiplicative, sign of the interaction parameter) quite well, although risky methods generated a tendency towards multiattribute risk aversion in additive models. Furthermore, for the simple models (e.g. additive, equal weight, and multiplicative equal weight), the standard elicitation methods were able to recapture the taught model parameters quite well. The ability of multiattribute utility techniques to recover value functions decreased, however, when models became very complex (e.g. multiplicative unequal weights).
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1982
- Accession Number
- ADA129972
Entities
People
- Detlof Von Winterfeldt
- Gregory M. Griffin
- Richard S. John
- Ward Edwards
Organizations
- University of Southern California