<p>Theory-deduced predictions might change agents’ beliefs, and thus also agents’ behavior. Since agents react to their beliefs by modifying their behavior to obtain their goals, they might react to a belief inspired by a theory-deduced prediction by modifying their behavior to obtain their goals, and this may have implications for the theory and its predictive success. In this paper, I first theorize this phenomenon. I disqualify past formulations of so-called reflexive predictions and advocate my account of self-defeating and self-fulfilling reactivity. I then examine the implications of three kinds of self-defeating reactivity, weak, strong, and vicious, for predicting. I conclude that self-defeating reactivity makes it impossible to predict, at least in some cases. Finally, I rethink whether self-defeating and self-fulfilling reactivity is exclusive to the human/social sciences or to states of affairs where human beings/social actors are involved. Here, I conclude that while reactivity is not exclusive to the human/social sciences, it is exclusive to cases where agents are involved. Thus, it is exclusive to cases where human beings/social actors are involved only <i>de facto</i>.</p>