Abstract

Frank Knight's Risk, Uncertainty and Profit (1921) separates a measurable uncertainty, which grouping and insurance can turn into a fixed cost, from an unmeasurable one, which they cannot. Two readings of what makes the second kind unmeasurable divide later scholarship. LeRoy and Singell (1987) argue that the line does not turn on whether people can form probabilities but on whether an insurance market can exist, and place uncertainty where moral hazard and adverse selection keep a risk from being insured. Langlois and Cosgel (1993) answer that for Knight uncertainty arises where cases cannot be sorted into classes of like cases, so that no frequency can be had, and judgment must take over. Both are readings of Knight's text. This article asks what a document written at the moment of decision adds to them. This article reads Chapters VII and VIII of Knight's book closely and then two synthetic decision records published for study by WSTiger, a studio of the RISCON family: an automated engine set in 2035 that sizes orders by the confidence its agents report, and a trading desk set in 2026 whose long and short timeframes point opposite ways. Both records are modern, written for the platform, and report no real firm or event. Read beside Knight, they place the missing class at the moment of decision and the threat to any later class in the decider's own interest in how the case is described. Knight's Chapter VIII names both difficulties in one chain. The records show them at work at different times, and they cannot settle which reading of Knight is right.