Difficult actions, such as quitting smoking, pose a paradox for rational confidence. Statistical evidence suggests that failure is likely, leading to the intuition that agents should have low confidence in success. Yet successfully quitting smoking requires sustained high confidence to maintain perseverance. This creates a tension between epistemic and practical rationality—should one be less confident based on evidence or more confident to be able to succeed?

In response to this paradox, I introduce a novel account, Confidence Dualism, which distinguishes between epistemic confidence (grounded in degrees of belief) and practical confidence (grounded in degrees of acceptance).

First, I show why orthodox monistic-epistemic approaches to rational confidence struggle to resolve the paradox. Some argue that epistemic rationality is a form of instrumental rationality. Yet this leads to the incommensurability problem and fails to explain how one can simultaneously have epistemic reasons for low confidence and practical reasons for high confidence.

Then, by extending Bratman’s model of the agent’s cognitive background of deliberation to include degrees of acceptance, I argue that practical confidence is grounded in degrees of acceptance. This enables agents to rationally pursue difficult actions without epistemically overestimating success, thereby resolving the paradox of difficult action.