Scholar
Ryan Carey
Google Scholar ID: 9U1CpcAAAAAJ
University of Oxford
Citations & Impact
Citations
271
H-index
9
i10-index
9
Publications
19
Co-authors
1
list available
Contact
Publications
Resume (English only)
Academic Achievements
- "Human Control: Definitions and Algorithms" (UAI 2023): Studies definitions of human control (e.g., corrigibility, alignment), their guarantees for human autonomy, and associated algorithms.
- "Reasoning about Causality in Games" (Artificial Intelligence Journal 2023): Introduces structural causal games as a unified framework for causal and game-theoretic reasoning.
- "Path-Specific Objectives for Safer Agent Incentives" (AAAI 2022): Addresses how to optimize objectives without undesirable means (e.g., user manipulation).
- "A Complete Criterion for Value of Information in Soluble Influence Diagrams" (AAAI 2022): Provides a complete graphical criterion for value of information in multi-decision influence diagrams.
- "Why Fair Labels Can Yield Unfair Predictions" (AAAI 2022): Shows how unfairness can be incentivized even with perfectly fair labels, with graphical conditions.
- "Agent Incentives: A Causal Perspective" (AAAI 2021): Presents sound and complete graphical criteria for four types of agent incentives.
- "Incorrigibility in the CIRL Framework" (AIES 2018): Analyzes how Cooperative Inverse Reinforcement Learning may fail to prevent incorrigible behavior.
Research Experience
- Research Fellow at the Future of Humanity Institute.
- Research Intern at DeepMind.
- Research Intern at OpenAI.
- Founder of the EA Forum (Effective Altruism Forum).
- Co-founder of the Causal Incentives Working Group, which applies causal models to AI safety.
