My research interests are diverse.  In brief, my main research thrust may be described as ''applied social choice'' (less generally, one could say my main research work is in the areas of legislative institutions and voting theory).  Methodologically, this involves formal analytic methods such as game theory and social choice theory, as well as computational models (specifically agent-based models).  Substantively, I am interested in the origins and consequences of legislative rules and procedures, in particular their self-organized and endogenous nature.

I have recently become interested in empirical and statistical analysis, specifically the use and application of machine learning and statistical inference to political science questions.

Below is an incomplete list of papers and projects, in various forms of completion.  Please feel free to email if interested.

      Gambling on issues [with John Gasper] : We develop a formal model where a set of candidates engage in a series of debates in order to learn about the public responsiveness to issues. After each debate, we assume that a poll is released.  Candidates’ statements can hence be thought of as an investigative process into the voters’ preferences (ie, we let candidates explore which issues are popular).  Our model of candidate inference borrows from the decision theoretic literature on multi-armed bandits.  At the core of the model is a fundamental and interesting tension: there is an incentive for a candidate to take a stance early (claiming an issue), yet there is an opposing incentive to free-ride off of others (and hence not make a mistake).

      Fair tests of set-valued social choice predictions:   In this paper I discuss and suggest ways to test if data supports various deterministic, set-valued social-choice solution concepts.   Many (most) common solution concepts in social choice literature are deterministic and set valued.  This presents challenges for testing such solution concepts with empirical data, as essentially the likelihood function is zero for a non-trivial portion of the alternative-space.  This paper surveys various ways to bring existing social choice solution concepts closer to statistical testing.  In it I show that an intuitive solution to the problem gives unsatisfactory results and I develop a Bayesian technique (based on the theory of random data) that allows for testing of theories that yield deterministic set-valued predictions.

      Robustness of Ideal Point-Scaling Procedures:  Extant methods for scaling legislators' ideal points rely on the ideological space being Euclidean. In this paper, I investigate how restrictive this assumption is. In particular, I examine two notions of ‘robustness’: (1) if voting is done in accordance to a ‘proximity’ which is non-metrizable, what can be said of existing scaling methods based on roll-call votes? Simulations are preformed in various topologies in which the triangle inequality does not hold and the results of various methods (OC, NOMINATE, etc.) are compared. Additionally, (2) scaling in multi-dimensional Euclidean space is compared to scaling in less structures spaces.

      A Note on Contestation-Related Solution Concepts:  This note introduces a family of new solution concepts based on the contestation relation (Schwartz 1990) and relates them to existing refinements of the Banks set.  I examine the relation of the introduced solution concepts to a popular (but enigmatic) solution concept, the Tournament Equilibrium Set (TEQ, Schwartz 1990).  Additionally, the connection between the contestation relation and general solution concepts is discussed.

      Scoring Methods and Banks Stability-  Nature Versus the Agenda Setter:  This work introduces new scoring methods for use in  social choice situations.  The new methods incorporate information relevant to the Banks set.  We then compare them, as well as traditional scoring methods  to investigate what information (if any) they convey about the robustness of alternatives in the Banks set.  Broadly speaking, an agenda setter may use  summary scores of alternatives to help schedule ``robust'' agendas:  when agenda setters are uncertain about the about the set of alternatives to be chosen from, how should she schedule agendas?

      Information Aggregation Properties of Agenda Procedures:  Agenda procedures are an important aspect of political decision making and considerable variation exists in the kinds of agendas deliberative bodies employ. This paper compares different agendas and evaluates them on their ability to aggregate information. I develop a model of voters voting strategically in committees and use this model to compare different agenda forms. The analysis allows for a comparison of various institutional practices (allowing for abstention, order of the status quo on the agenda, etc.)

      Voting by Adaptive Agents in Multicandidate Election:  This paper presents a computational model that incorporates a principal of robustness and is used for comparing voting rules.  I focus on voter behavior under different voting rules and model the behavior of different populations of agents voting in multiple elections over time who revise their voting strategies based on electoral outcomes.  The model employed relaxes rational-choice assumptions about voters' behaviors  and provides a direct comparison of different voting rules based on (1) voter strategies, (2) social utility and (3) outcomes.  Results from computational experiments suggest that approval and plurality voting rules generate similar, relatively intuitive behaviors, instant runoff consistently selects social-utility maximizing candidates, whereas  Borda Count often encourages pathological strategies to be used by voters and ``worst-case'' candidates to be elected.