This paper represents a collaboration between a policy researcher and a behavioural scientist who studies cooperation. Our goal was to develop a shared understanding of one particular policy topic, the reforms to the UK system of disability benefits initiated during the last term of the New Labour Government and accelerated under the Conservative-led administrations since 2010. These reforms introduced much stronger focus on conditionality and assessment, aiming to reduce the cost of the benefit by identifying and removing ‘cheaters’ or ‘undeserving’ recipients from the system. The reforms have failed by even their own stated goals. Here, we seek to understand why they seemed appealing and intuitively likely to succeed. We argue that humans are vigilant cooperators, sensitive to cues of need in others, but also highly susceptible to the idea that others are cheating. This vigilance is particularly marked where they lack a reassuring stream of direct personal evidence to the contrary. The vigilance of human cooperative psychology makes ideas of greater conditionality and punishment easy for politicians to conceive of and sell. However, set against this, there are principles that can be used and successfully appealed to in advocating greater generosity in welfare systems. These include the fundamental social similarity of recipients and non-recipients, and the idea that resources are not generated individually but represent the common windfall of a whole group.