Stochastic Process Models of Preference Change

This chapter gives an informal summary of a research program aiming to develop and test stochastic process models of preference change. What does it take to develop a formally precise and descriptively valid model of persuasion? Any such model should specify formally concise definitions of hypothetical constructs such as preferences or attitudes. The chapter reviews weak order and semiorder models of preferences that are grounded in decision theory. Such a model should also spell out how hypothetical constructs relate to observable behavior, such as feeling thermometer ratings. The chapter reviews response processes that, in some cases, accommodate within and across respondent heterogeneity in overt behavior. The model should furthermore specify formally what it means to change one’s preference over time and how that change relates to the persuasive environment. The chapter treats preference change as a continuous time stochastic process on a graph of preference states. The most innovative feature of the approach is to model the (perceived) persuasive environment itself also as a hypothetical construct that is not directly/objectively observable by the researcher. Last but not least, the chapter discusses how to accommodate partisan differences, how to incorporate respondents with immutable preferences, and the possibility that respondents may tune in and out of a persuasive campaign. The emphasis of the chapter lies in explaining key conceptual ideas grounded in decision theory and mathematical psychology.