Intentional action and nearly certain success

Many philosophers have argued that a necessary condition on an action's being intentional is that the agent has the ability to alter the probabilities of the relevant outcome. These philosophers would hold that this condition is what allows us to deny that an agent, for example, intentionally rolls something other than five fives with a set of dice, despite that agent's being virtually sure that this will be the outcome of the roll. The current paper uses some examples to cast this explanation, and the necessity of the associated condition, into doubt. It then suggests that what actually differentiates intentional from unintentional action – in the examples that falsify the hypothesis about control – is to be found in the agent's representation of the processes by which the relevant outcomes are produced. In particular, the agent must represent the outcome as happening because of what she does.