“Coherent Care” by abramdemski

Decision theory requires a shift from the traditional advice-giving model toward a design-oriented framework to effectively evaluate decision procedures. By treating decision procedures as components within a decision problem, one can apply calibration conditions—specifically observation and subjective state calibration—to determine which problems are well-defined for an agent. This approach resolves classic paradoxes, such as the Smoking Lesion and Transparent Newcomb, by identifying when a decision procedure is miscalibrated for a given scenario. While Updateless Decision Theory (UDT) proves optimal for observation-calibrated problems, it necessitates a prior, distinguishing it from Causal and Evidential Decision Theory. Ultimately, rational agency relies on a coherent web of purposes, where decision procedures avoid internal conflict, providing a more robust foundation for understanding how minds function across diverse, complex environments.