We propose to analyze population-scale prospective data on maladaptive drinking behavior and the behavioral and contextual matrix in which it is embedded. Our goal is to discover those implicit or """"""""silent"""""""" variables that are sufficient to predict and influence behavior change. Thus, it fits in with the control-theoretic view of behavior, as part of the identification stage. The approach is based on very modern non-linear dimensionality-reduction techniques. We pioneered use of these techniques in the analysis of personality data. The dimensionality-reduction techniques will be tailored to codify trends in behavior, social context, social status, collateral relationships, and other potential influences. When available, longitudinal data will be used to cluster individuals according to maladaptive alcohol drinking behavior. Our interdisciplinary team of computational and social scientists will place them in context of drinking behavior control systems and will use them to identify those individuals at risk. We will seek to propose a framework for incorporating our implicit variables into active behavior modification regimes.
