Comparing the predictive validity of alternatively assessed multi-attribute preference models when relevant decision attributes are missing

Multi-attribute decision problems require that trade-offs between conflicting objectives be made when evaluating choice alternatives. To overcome the shortcomings inherent in unaided intuitive decision maker evaluations, decision theorists have prescribed formal procedures to improve decision making. Such procedures elicit a descriptive model of choice for the alternatives being considered. Although these procedures are founded on the belief that the subject's preferences are well defined and error free, psychological studies have shown that errors do arise in the elicitation process. Thus, the formal preference elicitation procedure employed should be structurally capable of encoding meaningful preferences in the presence of these response errors. This study employs a simulated decision making environment to compare the predictive validity of alternatively assessed models when relevant attributes are omitted from the elicitation process and preference responses are not error free. The study found that meaningful preference predictions are sensitive to the elicitation procedure employed.