Towards an Equational Calculus of Interventions
summary
Presents a typed equational theory that rewrites interventions in probabilistic programs into equivalent do‑free programs by replacing mechanisms with a free variable Ŝ or by conditioning plus normalization, proving observation–intervention exchange as a local rewrite and constructing a final program that plugs a single unknown observational subprogram O into ordinary inference; provides a formal equivalence proof for the rules, explicitly does not claim completeness, and sketches automation and a semantic model as future work.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:42Causal queries in probabilistic programs
- 2:47Generalizing interventions
- 3:49Latent variables and experiments
- 4:53Do calculus for programs
- 5:57Equational rules
- 7:22Observation-intervention exchange
- 8:46Program simulating interventions
- 9:28Future work and conclusions
- 10:14Q&A
related talks
49mSemantic Foundations for Probabilistic Programming
24mA Friendly Introduction To Causality
01h 06mIntroduction to Causal Graphical Models: Graphs, d-separation, do-calculus
01h 07mAn Introduction to Potential Outcomes, DAGs and Single-World Intervention Graphs
24mHigher order programming with probabilistic effects: A model of stochastic memo...
01h 01mIntroduction to Experimental Design
01h 08mAn Introduction to Potential Outcomes, DAGs and Single-World Intervention Graphs
17mProgrammatic Strategy Synthesis: Resolving Nondeterminism in Probabilistic Progr...