ABSTRACT
Lore has it that a fundamental principle of Bayesian rationality is for decision makers to never turn down the offer of free information. To see why single-person decision-making is not immune to negative-valued information, this chapter considers more carefully Savage's argument that it is immune. Good's principle appears in Savage's discussion in Foundations of Statistics of the differences between a basic decision problem and a derived decision problem. The chapter focuses on probability assessments, which for Bayesians are understood as a person's partial beliefs. The first dent to this folklore comes from the theory of games, where some strategic interactions can result in a player being better off having less information. According to the canonical theory of synchronic decision making under risk, a perfectly rational person is one whose comparative assessments of a set of consequences satisfies the recommendation to maximize expected utility.
