The arrival of “big data” promises new degrees of precision in understanding human behavior. Could it also allow drawing a finer line between “choice” and “circumstances”? In a culture in which individual responsibility continues to be celebrated, this raises questions about new opportunities for institutional design with a stronger focus on individual responsibility. But what is it that can be learned from big data? In this paper I argue that we should not expect a “god’s eye view” on choice versus (...) circumstances from big data. “Responsibility” is a social construct that depends on the logic of different social situations, as well as our epistemic access to certain counterfactuals (e.g., whether an agent “could have acted differently”). It is this epistemic dimension that changes with the arrival of big data. But while this might help overcome some epistemic barriers, it might also create new problems, e.g., because of polluted and hence biased data. This is not just a theoretical problem; it is directly connected to the regulation of insurance. The new developments force us to directly confront questions about mutualist versus solidaristic forms of insurance, and more generally about how much weight to ascribe to individual responsibility, given all we know about unequal background circumstances. (shrink)