Research

For the most recent and comprehensive overviews of my scientific publications, see google scholar.  My CV is out of date.

Governance of AI

Progress in artificial Intelligence (AI) is likely to be one of the most important developments in the coming century. There is a non-trivial chance that we will see, in two decades, developments that transform society, the economy, and international relations. These will pose radical opportunities and challenges. I seek to anticipate these and identify levers for avoiding the risks. I do this work at DeepMind and previously the Centre for the Governance of AI (GovAI) (GovAI's google scholar page). 

For my talks, see allandafoe/ai-talks

For an overview of my perspective, I recommend:


Select work in reverse chronological order:

Reputation, Honor, Provocation, Resolve

Leaders and publics care about reputation and honor. This concern seems to be an important cause of war. Is it? I investigate this through survey experiments, natural experiments, and theory.  

The Liberal Peace

The peace amongst liberal countries is one of the most important phenomena for the wellbeing of humanity. I seek to understand what causes it. 

Methodology

Causal inference is central to social science. Many of our tools depend on implausible parametric assumptions, are fragile, and opaque. I seek to develop tools for causal inference, particularly for observational data, that do not depend on implausible assumptions, are more robust, and more transparent.

Transparency

Science depends on transparency. I work to promote better transparency norms and practices: scientists should share complete replication files, should preregister their analyses, should make their analyses transparent, and should evaluate the robustness of their results to reasonable alternative specifications. 

Other Topics

Unpublished Drafts