Caspar Oesterheld
My name is Caspar Oesterheld (click here for (anglicized) pronunciation). I am a Computer Science PhD student at Carnegie Mellon University, supervised by Vincent Conitzer. I'm interested in foundational topics in theoretical computer science, game theory and AI safety.
Working papers
- Caspar Oesterheld and Vincent Conitzer: "Can de se choice be ex ante reasonable in games of imperfect recall?"
Peer-reviewed publications
- Caspar Oesterheld, Johannes Treutlein, Roger Grosse, Vincent Conitzer, Jakob Foerster: "Similarity-based Cooperation". NeurIPS 2023.
- Caspar Oesterheld*, Johannes Treutlein*, Emery Cooper, and Rubi Hudson: "Incentivizing honest performative predictions with proper scoring rules". UAI 2023. Also presented at the CHAI 2023 Asilomar workshop. [Poster]
- Caspar Oesterheld, Abram Demski, and Vincent Conitzer: "A theory of bounded inductive rationality". TARK 2023. Also presented at OptLearnMAS 2021, FUR 2022, LOFT 2022.
- Vojtech Kovarik, Caspar Oesterheld, and Vincent Conitzer: "Game Theory with Simulation of Other Players". IJCAI 2023.
- Emanuel Tewolde, Caspar Oesterheld, Vincent Conitzer, and Paul Goldberg: "The Computational Complexity of Single-Player Imperfect-Recall Games". IJCAI 2023.
- Vincent Conitzer, Caspar Oesterheld: "Foundations of Cooperative AI". AAAI 2023 Senior Member Blue Sky Track.
- Caspar Oesterheld and Vincent Conitzer (2022): "Safe Pareto improvements for delegated game playing". Short version published at AAMAS 2021. Published in JAAMAS 36. Also presented at the 2019 EAF research retreat, the Cooperative AI Workshop at NeurIPS 2020, and CredibleCommitments.WTF. [Talk at AAMAS (15 min), Poster]
- Scott Emmons, Caspar Oesterheld, Andrew Critch, Vincent Conitzer, and Stuart Russell: "Symmetry, Equilibria, and Robustness in Common-Payoff Games". ICML 2022 (Spotlight). Also presented at GAIW 2021.
- James Bell*, Linda Linsefors*, Caspar Oesterheld*, and Joar Skalse*: "Reinforcement Learning in Newcomblike Environments". NeurIPS 2021 (spotlight presentation). [Poster]
- Johannes Treutlein, Michael Dennis, Caspar Oesterheld, and Jakob Foerster: "A New Formalism, Method and Open Issues for Zero-Shot Coordination". ICML 2021.
- Caspar Oesterheld and Vincent Conitzer: "Decision Scoring Rules". WINE 2020. Also presented at the 2019 EAF research retreat, CMID 2020, GAMES 2020/1, DIMACS Forecasting 2021, GAIW 2021, COMSOC 2021. [Extended version, Poster, talk at CMID 2020 (40min)]
- Caspar Oesterheld and Vincent Conitzer: "Minimum-regret contracts for principal-expert problems". WINE 2020.
- Caspar Oesterheld and Vincent Conitzer: "Extracting Money from Causal Decision Theorists". The Philosophical Quarterly 71 (4), pages 701–716, October 2021. Listed in Oxford University Press' "Best of Philosophy" 2021 collection. Also presented at the IJCAI-PRICAI 2020 AI Safety workshop, GAMES 2020/1.
- Caspar Oesterheld: "Approval-directed agency and the decision theory of Newcomb-like problems". Synthese 198, pp. 6491–6504, November 2021. Special Issue on Decision Theory and the Future of Artificial Intelligence. Shared second prize winner of the AI alignment prize.
- William MacAskill, Aron Vallinder, Caspar Oesterheld, Carl Shulman, and Johannes Treutlein: "The Evidentialist's Wager". The Journal of Philosophy 118 (6), pages 320–342, June 2021.
- Caspar Oesterheld: Robust Program Equilibrium. Theory and Decision 86, pages 143–159, February 2019. Also see my notes on "Testing εGroundedFairBot in a Transparent Prisoner's Dilemma Tournament".
- Caspar Oesterheld: Formalizing preference utilitarianism in physical world models. Synthese 193, pages 2747–2759, September 2016.
Teaching
In 2020, I received the Duke CS Department's Outstanding TA award.- Fall 2019: TA for Duke CS 230 – Discrete Mathematics for Computer Science – Instructor: Prof. Bruce Donald
- Spring 2020: TA for Duke CS 323 – Computational Microeconomics – Instructor: Prof. Vincent Conitzer
- Fall 2020: TA for Duke CS 590.7 – Computational Microeconomics – Instructor: Prof. Vincent Conitzer
- Fall 2022: Co-instructor with Vincent Conitzer and Tuomas Sandholm for CMU CS 15-784 – Foundations of Cooperative AI
Reviewing
EC 2021, GAIW 2021, ICML 2022 ("Top 10% Reviewer"), GAIW 2022, SAGT 2022, NeurIPS 2022, Minds and Machines, NeurIPS ML Safety Workshop 2022, UAI 2023 ("Top Reviewer"), ICML 2023, AIES 2023, NeurIPS 2023, CSCW 2024
Last update: 25 September 2023