This paper defines Preference Formation as a structural process through which pref-erences are produced through continuous interaction with an AI-mediated environment.In contrast to models that treat preference as an internally given or autonomously se-lected property, this study establishes that preferences are not chosen; they are formedunder conditions of repeated exposure.The paper identifies the observed structural conditions of preference formation,including repeated exposure, convergence of available elements, stabilization of inter-action patterns, and gradual transformation. It then formalizes Preference Formationas a dynamic and cumulative process, distinguishing it from both choice and fixedproperties, and expressing it as a time-dependent structural relation.The structural properties of Preference Formation are specified as cumulativity,convergence, asymmetric reversibility, and recursivity. These properties demonstratethat preference is progressively stabilized through repeated interaction while remainingpath-dependent and continuously reconfigurable.The paper further establishes the structural relationship between Preference For-mation and Algorithmic Selection. While selection defines the conditions of exposure,preference formation determines how exposure becomes internally structured into rel-atively durable orientation. This interaction forms a recursive loop linking externalstructuring and internal formation.Finally, the implications of Preference Formation are outlined as a structural pro-gression from preference to behavior, evaluation, and score formation. This progressionpositions Preference Formation as the internal condition through which orientation be-comes behaviorally stable and externally legible within AI-mediated environments.Accordingly, Preference Formation is defined as a fundamental internal mechanismwithin AI-Scored Societies, completing the structural model in which preferences arenot given, but continuously formed.
Kawazoe Tsutomu (Wed,) studied this question.