Adaptive Conjoint Analysis. Training Data: Knowledge or Beliefs? A Logical Perspective of Preferences as Beliefs

摘要

The foundational model of conjoint analysis is to model consumer purchase preferences by means of utility functions. Analysts run surveys and interviews to obtain a basic set of training data, typically user preferences on which the utility function is mapped. The utility theory trust the training data as knowledge while there is large literature emphasizing that users preference may change, may be incomplete and sometimes inconsistent. This paper argues on a logic-based model of conjoint analysis, particularly by proposing an alternative model of preferences as belief instead as fully trust knowledge. We adopt the categorical beliefs approach but the quantitative, probabilistic approach may be considered too. In the context of adaptive conjoint analysis, we identified three kinds of beliefs, describe a mechanism of mapping answers to beliefs and provide the basis on belief update when new information occurs. Future work on our logic-based framework will focus obtaining an optimal logic-based preference aggregation including by relaxing Pareto efficiency in Arrow's aggregation framework as well as researching on non-prioritized belief revision in adaptive conjoint analysis.