Integrated Assessment (IA) is the pursuit of a research program generated by the limitations of traditional forms of risk management. This claim can be justified by the following argument. Over the last decades, analysts and practitioners have brought to maturity a large array of tools for risk management. Most of them rely on combining judgments of utility with judgments of probability. This is the approach of the Rational Actor Paradigm (RAP). With many environmental problems, however, RAP‐based tools have run into considerable practical and theoretical difficulties. In response to these difficulties, a series of alternative approaches to practical risk management and to the theoretical understanding of risk have been elaborated. They try to embed the rational choices of individual actors studied by RAP into a broader framework of social rationality. This task can be approached by distinguishing situations where an actor holds unambiguous judgments of preference and probability from situations characterized by ambivalent judgments. RAP can handle the former, but not the latter. Problems whose management requires a combination of widely differing scientific disciplines are especially likely to involve ambivalent judgments of probability. The study of such problems constitutes the research program of integrated assessment. It involves three main tasks: developing IA models which can represent ambivalent expectations and evaluations, developing IA models which use such representations to study non‐marginal changes of social systems, and developing procedures of participatory IA which enable researchers to engage in an iterative exchange with various stakeholders.