Beslutningsregler

Beslutningsregler are principled prescriptions for making choices under uncertainty. They specify how information about the current state, data or signals should be translated into a decision. The term is used across decision theory, statistics, economics, operations research and artificial intelligence to describe the rule or criterion guiding action.

A decision rule maps an information set to an action or a distribution over actions. Its objective is typically to maximize an objective function such as expected utility, minimize expected loss, or minimize expected cost. There is a distinction between normative rules, describing how rational agents should decide, and descriptive rules, describing how people actually decide. Rules can be deterministic, yielding a single action for each input, or stochastic, assigning probabilities to different actions.

Common types of decision rules include: expected utility maximization, which selects the action that maximizes the expected value of a utility function; Bayes decision rules, which minimize posterior expected loss; and classification-oriented criteria such as maximum a posteriori (MAP) and maximum likelihood (ML). Robust rules such as minimax or maximin are used in adversarial or highly uncertain settings. Threshold rules decide actions when a signal crosses a predefined value.

In sequential decision problems, decision rules are implemented as policies, often derived through dynamic programming and the Bellman equation, to maximize cumulative expected utility. In practice, decision rules appear in rule-based systems, machine learning classifiers, and multi-criteria decision analysis, blending formal optimization with qualitative judgments.

Applications span medicine, finance, operations, risk management and quality control, where explicit decision rules help standardize choices, reduce ambiguity, and provide transparent justifications for actions.