This chapter argues that the distinction between ‘good’ and ‘bad’ violence, which legitimates monopolies of violence in territorial states, is not based on meaning or fear but on the role of the sacred in the management of violence. The irruption of religious violence in politics is not so much a return to a form of pre-secular forms of the sacred, but rather indicates that the self-organizing mechanism of violence that gave rise to the sacred has become ineffective. In the individualist modern world violence has increasingly lost its capacity to give rise to new institutions, to new religions or new forms of social organization. The progressive disappearance of the sacred in its protective function that the state has developed to protect us against our own violence is not only an institutional challenge but also requires re-thinking the relations between citizens, states, and ethics in an increasingly borderless world.
If data-driven agency is a form of agency based on what machines have learned, it seems important to understand the nature and limit of the type of knowledge that can be mechanically obtained from digital data. After reviewing some of the popular claims made about big data this chapter explores some of the differences in the use of big data and machine science in the natural sciences and in the social domain. It insists in particular on the fact that in the natural sciences what constitutes data and how it should be interpreted are under the collective jurisdiction of specialists of the domain whose authority is recognized by governments, funding agencies and the general public, while in the social domain the data is often claimed to be simply ‘found’ though it is explicitly sought for a variety of reasons. It is not however ‘crafted’ in the sense of being validated and authenticated by the community of concerned researchers. In consequence, anyone who has the necessary technical competence gains the authority to interpret the data and declare what the data proves. Finally, the chapters analyzes some aspects of machine learning and science that tend to encourage the faulty interpretation that ‘data is enough’.