Workshop announcement

Beyond single-agent decision theory

The current decision-theoretic foundations of statistics and machine learning are insufficient for addressing some of the key challenges facing science and society today. First, there are pressing concerns about the social impact of artificial intelligence and machine learning, regarding issues such as fairness, inequality, and value alignment. Single-agent decision theory is insufficient for conceptualizing the underlying conflicts of interest between different agents, or the value alignment issues resulting from divergent objectives.
Second, there is a perceived replication crisis of empirical research, which might be due to p-hacking or publication bias. This crisis has motivated proposed solutions such as pre-registration of statistical analyses and reforms of the publication system. Single-agent statistical decision theory again cannot make sense of these problems and solutions, as it does not allow for conflicts of interest between different parties, private information, or dynamic inconsistency.

Workshop format

The goal of this workshop is to provide a venue for extended in-depth interaction, and the development of research agendas, formulating “big open questions,” with collaboration across disciplinary boundaries, to further the foundations of statistics and machine learning. This workshop will involve speakers from econometrics and economic theory, computer science, and philosophy. The workshop will start with a series of tutorial lectures. The goal of these lectures is to bring all participants on the same page, especially with regards to the relevant state of research in other disciplines. These will be followed by talks on frontier work. In these talks, the speakers will give an overview of their own work, and other related frontier work. The workshop will conclude with an open-ended panel discussion on big open questions.

Confirmed speakers

  • Isaiah Andrews (Econometrics) Website
  • Celestine Mendler-Dünner (Computer Science) Website
  • Nika Haghtalab (Computer Science) Website
  • Lily Hu (Philosophy) Website
  • Carina Prunkl (Philosophy) Website
  • Jann Spiess (Econometrics) Website
  • Ana-Andreea Stoica (Computer Science) Website


Tutorial lectures

Monday, May 22

11:45 Maximilian Kasy
Opening remarks
Slides, Recording

12:00 Carina Prunkl
Algorithms and social epistemology
Slides, Recording

14:15 Celestine Mendler-Dünner
Performative Prediction
Slides, Recording

16:00 Jann Spiess
Econometrics with Misaligned Preferences
Slides, Recording

Tuesday, May 23

11:30 Lily Hu
Causal Inference and the Problem of Variable Choice
Slides, Recording

14:15 Isaiah Andrews
Correcting for Selective Publication and Attention
Slides, Recording

16:00 Ana-Andreea Stoica
Diagnosing Algorithmic Inequality in Social Networks
Slides, Recording

Wednesday, May 24

11:30 Nika Haghtalab
Foundations of learning and incentives
Slides, Recording

13:45 Guided tour of Oxford for speakers

Frontier talks

Wednesday, May 24

16:00 Ana-Andreea Stoica
Interventions for mitigating Algorithmic Inequality in Social Networks

Thursday, May 25

11:30 Jann Spiess
Explanations with a purpose: regulating black-box algorithmic decisions
Slides, Recording

14:15 Carina Prunkl
Noise - a flaw in algorithmic judgment?

16:00 Celestine Mendler-Dünner
Algorithmic Collective Action in ML
Slides, Recording

Friday, May 26

11:30 Nika Haghtalab
Multi-objective learning: A unifying perspective on collaboration, fairness, and robustness
Slides, Recording

14:15 Isaiah Andrews
A Model of Scientific Communication
Slides, Recording

16:00 Lily Hu
Do Causal Diagrams Assume a Can Opener?
Slides, Recording

Saturday, May 27

10:00 Group discussion
Speakers only. Radcliffe Humanities building.