Stephan Jagau

I am a postdoctoral scholar at the Institute for Mathematical Behavioral Sciences at UC Irvine and a member of the EPICENTER (Research Center for Epistemic Game Theory) at Maastricht University. I did my PhD at CREED (Amsterdam School of Economics) and at the Tinbergen Institute under the supervision of Theo Offerman and Matthijs van Veelen.

I am an Economic Theorist and Experimentalist interested in Epistemic and Psychological Game Theory, Decision Theory, and Evolutionary (Game) Theory.

At UC Irvine's Experimental Social Science Laboratory, I am co-directing an initiative to virtualize lab experimentation using Kubernetes and build an Experimental Social Science Research Network. The virtual lab allows hosting classical and online experimenters for large numbers of experimenters, blurring the lines between the traditional lab setup and platforms such as oTree-Hub. Mail to sjagau[at] if you are interested in becoming a user or a contributor.



Common Belief in Rationality in Psychological Games: Belief-Dependent Utility
and the Limits of Strategic Reasoning.
(with Andrés Perea)
Journal of Mathematical Economics (2022) 100, 102635. Paper

Defaults, normative anchors, and the occurrence of risky and cautious shifts.
(with Theo Offerman) Journal of Risk and Uncertainty (2018) 56(3), 211-236. Paper

A general evolutionary framework for the role of intuition and deliberation
in cooperation. (with Matthijs van Veelen) Nature Human Behaviour (2017) 1(8), 0152.
Paper SI News & Views (Adam Bear and David Rand)

Working Papers

The Fundamental Theorem of Epistemic Game Theory: The Infinite Case. Full Paper Pitch
Abstract: In many applications of game theory, we let infinite strategy sets stand in for large finite strategy sets, expecting that results will back-translate from the infinite to the finite setting. Transfinite eliminations of non-best replies in some infinite games raise concerns whether this will always work. This paper gives a novel, measure-theoretic characterization of common belief in rationality via two alternative procedures. One is the usual transfinite elimination of non-best replies. The other one, elimination of non-best replies and supporting beliefs, entirely avoids transfinite eliminations. Thus, transfinite eliminations of non-best replies never require transfinite depths of reasoning, and back-translating infinite-game results to finite settings always works. Instead, what transfinite eliminations reveal is a deficiency of non-best reply elimination as a general description of strategic rationality.

Work in Progress

Additive Context-Dependent Preferences. Extended Abstract Pitch
Abstract: "Money equals utility" is a much criticized 'axiom' that is central across a vast range of economic experiments and theory. Subjecting this 'axiom' to experimental testing requires an empirically tractable theory of context-dependent preferences. This paper presents novel behavioral foundations for additive context-dependent preferences and state-dependent expected utility. Crucially, these behavioral foundations do not require empirically implausible comparisons of alternatives across different states. Moreover, they can handle any state-dependent, multi-alternative decision problem. In particular, no diversity-type assumptions are used. A central application is to direct utility measurement in games, enabling a causal understanding of how e.g. risk- and social-preferences affect strategic choice.

The Incompatibility of Admissibility and Equilibrium. (with Christian Bach)
Abstract: Iterated admissibility and equilibrium are two fundamental solution concepts for normal-form games. This note shows that they entail logically conflicting requirements on players' beliefs. Strategic reasoning that combines equilibrium and iterated admissibility is impossible for all but a very small class of normal-form games. For this restricted class, the resulting "admissible equilibrium" is shown to select fully mixed undominated perfect equilibria.

Virtualizing the Lab with oTree and Kubernetes. (with John Duffy) Project Presentation
Abstract: In recent years, experimental economists have increasingly explored designs that reach beyond the confines of the traditional experimental economics lab. Examples include lab-in-the-field experiments, large-scale experiments, and online experiments. This project revamps the traditional lab server to meet the challenges of state-of-the-art experimentation. We virtualize standard software for running experiments (Python/oTree) using Docker and embed it in a server cluster orchestrated by Kubernetes. The containerization of experimental software brings two major benefits. Firstly, it allows us to manage large numbers of parallel experimenter-workspaces and to safely deploy computerized experiments to any location across the web. This not only blurs the lines between lab-, field-, and online-experimentation. Also, it enables teams of experimenters to pool resources in a joint lab beyond the limits of a single university. Secondly, the containerized software is deployed fully independently of experimenters’ individual data. This allows us to build an archive of past versions and variants of experimental software from which exact replications of past experiments can be deployed within a matter of minutes.

Big Bucks, no Whammies: Understanding equilibrium selection
in experimental stag-hunt games.

Abstract: Even 44 years after Harsanyi and Selten's seminal work on payoff- and risk-dominance, the behavioral mechanisms driving experimental equilibrium selection remain poorly understood. Specifically, most existing designs fall short of separating "me"- and "we"-drivers in coordination decisions on the one hand and context-factors such as learning, communication, and complexity on the other hand. This experiment uses a no-feedback, choice-list style design that allows to systematically chart the changing attitudes of individuals in symmetric 2x2 coordination problems across the payoff space. Payoffs within choice lists are designed to vary within orthogonal subspaces of "Nash"- and "Pareto"-incentives, allowing for an independent exploration of "me"- and "we"-drivers in coordination decisions.

Linear Psychological Games. (with Andrés Perea)

Decisions in the Face of Discoveries. (with Burkhard Schipper)

Selection in the Lab. (with John Duffy)