Skip to content

Citing causalrl

If you use causalrl in research, please cite the software. The canonical metadata is in CITATION.cff; a BibTeX entry:

@software{coelho_causalrl,
  author  = {Coelho, Raphael},
  title   = {{causalrl}: Causal intervention-selection and causal-RL research tools},
  year    = {2026},
  version = {0.99.0},
  license = {MIT},
  url     = {https://github.com/raphaelrrcoelho/causalrl}
}

Please also cite the primary sources

causalrl implements published methods and is faithful to their originals. When you use a specific capability, cite the paper it comes from — each is attributed inline in the Tour by Task and in the relevant source module. The load-bearing references include:

  • Causal RL taxonomy — Bareinboim, Causal Reinforcement Learning (the 9-task framing).
  • do-calculus / identification — Pearl, Causality; Shpitser & Pearl; Bareinboim & Pearl (gID, JMLR 2015).
  • Transportability — Bareinboim & Pearl (AAAI 2012; NeurIPS 2014).
  • Structural causal bandits / POMIS — Lee & Bareinboim (NeurIPS 2018; AAAI 2019). The POMIS engine is adapted from the MIT-licensed sanghack81/SCMMAB-NIPS2018.
  • Bandits with unobserved confounders — Bareinboim, Forney & Pearl (NeurIPS 2015).
  • Causal imitation — Zhang, Kumor & Bareinboim (NeurIPS 2020).
  • Discovery — Spirtes, Glymour & Scheines; Meek (UAI 1995); Zhang (2008, FCI).
  • Partial identification & sensitivity — Manski (1990); Tan (2006).