Search

Standard decision theory has some striking consequences. First, if you have any irrational preferences, it does not make sense to ascribe credences to you---decision theory treats you as having no opinions at all. Second, you should always prefer to look at more information before making a decision if the information is free, even if you think the information is likely biased.These consequences seem wrong. Your preferences are sometimes irrational, but you clearly still have opinions about things. If we could only ascribe credences to agents with perfectly rational preferences, decision theory couldn't give any useful advice to imperfect agents like us. Furthermore, people sometimes prefer not to look at more information before making a decision. For example, when grading a paper, I prefer not to know the student's name. This behavior seems perfectly rational, even laudable.Does this mean that decision theory is broken? No. I argue that we can fix its problems. The key to ascribing credences to non-ideal agents is to start by looking at comparative probability judgments, like thinking it's equally likely to be sunny and rainy tomorrow. Comparative probability is tied to preferences in a straightforward way---you think that sunshine and rain are equally likely if you are indifferent between betting on them. If sunshine and rain are the only two possibilities for the weather tomorrow and you think they're equally likely, you assign probability .5 to both. I show that if your preferences satisfy some minimal constraints, we can extend this procedure to fix your entire credence function while allowing many of your preferences to be irrational (Chapter 2).The key to explaining why it can be rational to reject free information is to recognize that we can be uncertain about how we will react to evidence. If I were certain that I always respond to evidence in a perfectly rational way (by 'conditionalization'), then perhaps it would be rational to look at my students' names before grading papers. Indeed, certainty that one will react to evidence in a perfectly rational way is an assumption built into Good's well-known theorem about the value of information. But this assumption might fail. I know that I'm not always rational---for example, I might give too much weight to the fact that George got an `A' on the first paper and treat this as better evidence than it is that his current paper deserves a high mark. Once I take this possibility into account, I might be better off ignorant (Chapter 3).Uncertainty about how we will react to evidence has other consequences as well. For example, it can lead us to make sequences of choices which, taken together, yield sure loss. Many decision theorists have claimed that such choices always indicate irrationality. I argue that this bit of conventional wisdom needs revision (Chapter 4).One upshot is that by widening the scope of decision theory to include non-ideal agents, we enable the decision theorist to give vindicating explanations of common phenomena, like having opinions without being fully rational and avoiding information before making a decision. Another upshot is that a more sophisticated decision theory is relevant for designing beneficial AI systems, since existing ideas for doing so assume an implausibly strong conception of rationality (Chapter 5).

Among many things, Frank Ramsey is famous for his invention of a logical device called the Ramsey sentence. The Ramsey sentence of a scientific theory is a conjunction of the theory’s propositions where the theoretical vocabulary is existentially quantified out. The sentence appears in an obscure paper titled “Theories”. This paper is extremely difficult because Ramsey likely did not finish it before his death and it mostly consists of a mathematical model of scientific theories and a labyrinthine toy example. Consequently, philosophers have proposed many radically different and incoherent interpretations of the Ramsey sentence and Ramsey’s philosophy of science. An important fact interpreters have ignored is that Ramsey had intended “Theories” to be a chapter in a book centered on a revised version of his formal theory for decision-making under uncertainty. The principal accomplishment of my dissertation is to leverage this fact to provide the first complete and philosophically satisfying interpretation of “Theories”, including Ramsey’s toy example and the Ramsey sentence.Ramsey’s theory for decision-making is incompatible with his philosophical commitments in “Theories” and contemporary papers. In my dissertation, I revise that theory to make it consistent with those commitments. I combine this with the formal model he describes in “Theories” to provide a unified reconstruction of Ramsey’s intended model for scientific theories. I use the reconstructed model to explain Ramsey’s toy example, and I furnish a complete interpretation of Ramsey’s arguments in “Theories”. With these parts, I develop a detailed account of the laws and chances derived from scientific theories and a novel interpretation of the Ramsey sentence. This has several important philosophical consequences.First, Ramsey did not use the Ramsey sentence in the ways philosophers have previously proposed. Philosophers view the Ramsey sentence either as a tool for eliminating theoretical terms or for asserting the existence of theoretical properties. Neither of these claims is correct. Instead, I argue that the Ramsey sentence is a communication and deliberation device aimed at representing the propositions an individual believes with her behavioral dispositions. Informally, it expresses as new propositions the mixture of rules a person uses to guide her behavior so that she may deliberate and communicate those rules with herself and other people. This suggests the Ramsey sentence can be used to read off the content of an agent’s belief from their behavior.Second, Ramsey formulated a novel and interesting type of non-reductionistic scientific anti-realism. That is, for Ramsey theoretical propositions are fictitious and yet not eliminable for observational propositions. Theoretical propositions are fictitious because their meaning depends on theoretical laws, which, being universally quantified, are also fictitious. But theoretical propositions cannot be eliminated because they entail conceptual possibilities that are richer than those given by the observation language.Third, Ramsey held laws and chances to be rules of deference for degrees of belief. His account of laws and chances foreshadows pragmatic, subjectivist accounts of laws and chances from later in the twentieth century. Degrees of belief are representations of preferences over gambles involving observable propositions. A chance then is a rule for preferring some gambles over others, and a law is a chance that assigns unity to its degrees of beliefs. This prefigures the views that chances and subjunctive conditionals are reducible to properties of degrees of belief.

This collection of papers examines applications of microeconomic theory to practical problems. More specifically, I identify frictions between theoretical results and agent behavior. I seek to resolve these tensions by either proposing mechanisms to more closely capture the theoretical environment of interest or extending the model to more closely approximate the world as individuals perceive it. In the first chapter, "Compensation without Distortion,'' I propose a new mechanism for compensating subjects in preference elicitation experiments. The motivation for this tool is the theoretical problem of incentive compatibility in decision experiments. A hallmark of experimental economics is the connection between a subject's payment with their actions or decisions, however previous literature has highlighted shortcomings in this link between compensation and methods currently used to elicit beliefs. Specifically, compensating individuals based on choices they make increases reliability, however these payments can themselves distort subjects' preferences, limiting the resulting data's usefulness. I reexamine the source of the underlying theoretical tension, and propose using a stochastic termination mechanism called the "random stopping procedure'' (RSP). I show that the RSP is theoretically able to structurally avoid preference distortions induced by the current state of the art protocols. Changing the underlying context subjects answer questions—by resolving payoff uncertainty immediately after every decision is made—the assumed impossibility of asking multiple questions without creating preference distortions is theoretically resolved. To test this prediction, I conduct an experiment explicitly designed to test the accuracy of data gathered by the RSP against the current best practice for measuring subject preferences. Results show that RSP-elicited preferences more closely match a control group's responses than the alternative. In the second chapter, "School Choice and Class Size Externalities,'' I revisit the many-to-one matching problem of school choice. I focus on the importance of problem definition, and argue that the "standard'' school choice model is insufficiently sensitive to relevant characteristics of student preferences. Motivated by the observation that students care about both the school they attend, and how over- or under-crowded the school is, I extend the problem definition to allow students to report preferences over both schools and cohort sizes. (Cohort size is intended as a generalization of school crowding, relative resources, or other similar school characteristics.) I show that, if students do have preferences over schools and cohort sizes, current mechanisms lose many of their advantageous properties, and are no longer stable, fair, nor non-wasteful. Moreover, I show that current mechanisms no longer necessarily incentivize students to truthfully report their preferences over school orderings. Motivated by the observation that students care about both the school they attend, and how over- or under-crowded the school is, in "School Choice and Class Size Externalities'' I extend the standard school choice problem to incorporate both of these elements. I show that, if students do have preferences over schools and cohort sizes, current mechanisms are no longer stable, fair, nor non-wasteful. In response, I construct an alternative matching mechanism, called the deferred acceptance with voluntary withdrawals (DAwVW) mechanism, which improves on the underlying (unobserved) manipulability of "standard" mechanisms. The DAwVW mechanism is deterministic and terminates, more closely satisfies core desirable matching properties, and can yield substantial efficiency gains compared to mechanisms that do not consider class size. In the third chapter, I provide an overview of the history of decision experiments in economics, describe several of the underlying tensions that motivate my other projects, and identify alternative potential solutions that have been proposed by others to these problems. In this project, I add context to the larger field of experimental economics in which my research is situated. In addition to the mechanisms discussed by prior literature reviews, I incorporate and discuss recently developed payment and elicitation methods, and identify these new approaches' advantages and drawbacks.