Decomposition is a common strategy for dealing with the complexity of multiattribute decision problems. A cognitively demanding task is broken down into tasks requiring simpler, perhaps easier, judgments which can then be aggregated. But individual judgments can be inconsistent in systematic or random fashion and when aggregated there is the possibility of propagation of this inconsistency. In this paper inconsistency in the form of random error is investigated in the context of additive decomposition of multiattribute utility. The process of aggregation of random error is studied and a comparison made with random error in holistic estimates of multiattribute utility. Conditions under which decomposition improves the consistency of the multiattribute utility estimate are presented and discussed.
- Multiattribute Utility
- Random Error