Normative Formal Theory

Session Submission Type: Full Paper Panel

Session Description

Typically in political science, there are walls between normative political science and other areas of study. These walls are artificial. In our papers, we hope to show there is a much more fluid relationship between affect, behavior, knowledge, ideas, and formal theory.

In recent years, we have seen significant increases in AI being used as agents on behalf of human principals in several domains including: drone warfare, personal assistants, transportation, stock trading, and political campaigns. This has led to the need to develop and formalize many normative ideas and intuitions in order to ensure AI behavior properly aligns with the values of its human principals. The papers in this panel formalize important normative political ideas which help ensure AI behavior aligns with human values.

Additionally, the papers in this panel address populism and privilege issues such as, “is Rikerian populist democracy possible?” and “will society divide into those with privilege and those without it, even if society starts out in an egalitarian fashion?”. What follows are brief descriptions of each paper.

Traditionally, formal decision theory has taken the normative decision criteria as given, with the analysis focused on empirical uncertainty of the decision making situation. Recently, there has been increasing interest in the question of how to decide when there is uncertainty about the normative decision criteria. The typical approach to this problem has been to apply expected value theory to this conundrum. In their paper, Franz Dietrich and Brian Jabarian show this is just one of four reasonable approaches to the conundrum.

With more and more political decisions being automated (e.g. drone warfare, political campaigning with bots on Facebook and Twitter, etc.), it is reasonable to expect automated political decision making to play a larger role in politics in the coming decades. Two of the major schools of thought in decision theory are causal decision theory (CDT) and evidential decision theory (EDT). (To observe the difference between the two theories, consider a two player prisoner’s dilemma where the two players are known to act independently, but nevertheless have tended to act similarly. EDT would recommend cooperation (because the agents are likely to behave in the same manner) while CDT would recommend defection (because the agents behave independently and the behavior of one player does not cause the behavior of the other.) In his paper, Daniel Kokotajlo demonstrates in novel ways that CDT agents are much more susceptible to being exploited by other players in games.

In his classic “Second Discourse”, Jean-Jacques Rousseau argued that envy inevitably leads to inequality. In his paper, Alexander Motchoulski formally models Rousseau’s argument from the “Second Discourse”. Motchoulski shows that Rousseau’s argument is comparable to a three person prisoner’s dilemma. When formalized, Rousseau’s argument demonstrates inequality will inevitably arise with the formation of a privileged high class and exploited lower class, even when all players start out equal.

In his magnum opus, William H. Riker argued that the intransitivity of majority preference made the creation of a general will and populist democracy impossible. Several political theorists, including Robert Dahl, Duncan Black, and Christian List have compiled normative arguments for majority rule as the best democratic decision rule on two alternatives. These compilations typically include five kinds of arguments: (social) contractual, social choice, utilitarian, epistemic, and statistical (central tendency). The compilers, like Dahl, typically contend that the intransitivity of majority preference makes generalization of the arguments to multiple alternatives difficult, if not impossible. In his paper, Mahendra Prasad contends that if we revive an interpretation of majority rule as consent-of-the-majority, an interpretation which is commonly found in the works of Locke, Rousseau, and Rawls, then we can overcome the intransitivity problem and generalize the arguments to multiple alternatives. Prasad’s work suggests that a general will and Rikerian populist democracy are possible.

Sub Unit

• DIVISION 4: Formal Political Theory

Individual Presentations

• Decision Under Normative Uncertainty - Brian Jabarian; Franz Dietrich, Paris School of Economics & CNRS

• The Origin and Inevitability of Inequality: A Model of Rousseau’s 2nd Discourse - Alexander Motchoulski, University of Arizona

• Why a Popular Will Can Exist: Generalizing Majority Rule’s Five Arguments - Mahendra Prasad, University of California, Berkeley

Chair

• Steven J. Brams New York University

Discussant

• David Wiens University of California, San Diego