Centralized decisions that impose risks demand a justification. The most promising way to justify these decisions is in terms of consent of those people who will be affected by them. Attempts to determine how safe is safe enough in terms of justice, rights, or efficiency are unlikely to get us very far. The problem with consent, however, is that actual consent is impossible to obtain for decisions that affect large numbers of persons. For this reason, we need to explore other kinds of consent and the idea of what a reasonable person would agree to. Three models of consent are described, each one more powerful (but more controversial) in terms of its potential for justifying centralized decisions.