Probabilism through a transform: An improved accuracy argument for Probabilism

Abstract Standard accuracy arguments for Probabilism require acceptable scoring rules to be strictly proper. This requirement makes these arguments question‐begging. In this paper, I offer an alternative, better argument for a version of Probabilism. It avoids the strict propriety requirement, hence, my argument is not question‐begging. Moreover, dropping strict propriety enables agents to use a wider variety of scoring rules, giving them more freedom in how to represent their uncertainty. My argument utilizes and improves upon an old argument by Dennis Lindley for the conclusion that representations of uncertainty are rationally admissible if and only if they are probabilistic through a transform. This means that these representations must be implicitly probabilistic through a transform function that is uniquely determined by the agent's scoring rule. This conclusion is preferable to Probabilism, as it is both more permissive in how agents can represent their uncertainty and relies on weaker and less questionable premises.