Causal Models

Course at the 6th Tsinghua Logic Summer School, 2026

Instructors: Bonan Zhao and Tadeg Quillien (University of Edinburgh)

Location: tbd

Time: tbd

Teaching assistants: Wenlong Zheng (zhengwl21 at mails dot tsinghua dot edu dot cn) and Xiangyuan Ji (xy-ji23 at mails dot tsinghua dot edu dot cn)

Course description

This course explores formal approaches to causal reasoning. We begin with an introduction to causal graphical models, examining how Causal Bayes Nets and Structural Causal Models can be used to represent and reason about causal structures. These lectures emphasize the formal distinction between ‘seeing’ and ‘doing’.

We then discuss more recent approaches that represent causal knowledge with structured programs. We examine probabilistic program induction as a framework for causal reasoning, treating the acquisition of causal knowledge as a search problem over program spaces. Drawing on formal methods including probabilistic context-free grammars (PCFGs), approximate Bayesian inference, and adaptor grammars, we explore how structured representations enable few-shot learning and generalization.

The course concludes with an introduction to counterfactual reasoning and some of its applications to reasoning about causal responsibility.

Pre-requisites: Basic probability theory; Propositional and first-order logic.

Course material

Day 1: introduction to causal models

Day 2: inference and learning in causal models

Day 3: program induction

Day 4: approximate inference

Day 5: counterfactuals, conclusion

Homework and exams

(tbd)