The article "Risk, Ambiguity, and the Savage Axioms" by Daniel Ellsberg explores the distinction between "measurable uncertainty" or "risk" and "unmeasurable uncertainty," which Knightian uncertainty cannot be represented by numerical probabilities. Ellsberg questions whether people always behave as if they assign numerical probabilities to uncertain events, even in situations where statistical information is irrelevant or impossible to obtain. He discusses the implications of the Savage axioms, which propose that rational individuals can infer numerical probabilities from their choices, and argues that these axioms may not hold in certain circumstances. Ellsberg presents examples where individuals' choices violate the Savage axioms, such as in situations of high ambiguity where the decision-maker has limited or conflicting information. He suggests that in such cases, people may not be acting based on numerical probabilities but rather on other criteria, such as the credibility and reliability of their information. Ellsberg concludes that while the Savage axioms are useful for prediction and normative guidance in many situations, they may not always be applicable, and alternative decision rules based on nonprobabilistic descriptions of uncertainty may be more appropriate in certain contexts.The article "Risk, Ambiguity, and the Savage Axioms" by Daniel Ellsberg explores the distinction between "measurable uncertainty" or "risk" and "unmeasurable uncertainty," which Knightian uncertainty cannot be represented by numerical probabilities. Ellsberg questions whether people always behave as if they assign numerical probabilities to uncertain events, even in situations where statistical information is irrelevant or impossible to obtain. He discusses the implications of the Savage axioms, which propose that rational individuals can infer numerical probabilities from their choices, and argues that these axioms may not hold in certain circumstances. Ellsberg presents examples where individuals' choices violate the Savage axioms, such as in situations of high ambiguity where the decision-maker has limited or conflicting information. He suggests that in such cases, people may not be acting based on numerical probabilities but rather on other criteria, such as the credibility and reliability of their information. Ellsberg concludes that while the Savage axioms are useful for prediction and normative guidance in many situations, they may not always be applicable, and alternative decision rules based on nonprobabilistic descriptions of uncertainty may be more appropriate in certain contexts.