This paper proposes and applies a formal theoretical model of an automatic (prosocial and antisocial) caring mechanism: the informational affective tie mechanism (iATM) model. Novel in the formalization is the factoring in of information extraction concerning the behavioral type of agents interacted with, jointly with the contexts of these interactions and the attention they attract. Empirical support comes from five very different data sources: experimental findings, econometric results, model-based brain scanning (fMRI) findings, additional neurobiological evidence, and translational and evolutionary evidence. Applications address: the impact of time pressure and cognitive load; the endogeneity of different behavioral response patterns (like tit-for-tat); social preference drift and tipping points in collective action; why behavioral survey questions can be problematic; spread of caring through affective networks, an uncertainty-based link between social-, risk-and time-preferences; happiness and identity; and, the neglected political economic role of communities (next to centralized authorities and markets). The endogeneity of caring preferences sharply contrasts with the standard assumption in economic theory of stable (mostly selfish) preferences. Moreover, the provision of a neurobiological underpinning moves the iATM model away from the standard as-if approach towards an as-is approach. Although the focus is on humans, some attention will be paid also to the model's relevance for studying other species.