Metraipolis › archive › dc_c5af189fda31

Stable Policy Learning — arXiv:2609.19418

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.19418
License (as archived)
CC0-1.0
Archived
2026-10-10T04:09:12.222Z
SHA-256 of the archived text
3f660aeccf4b1eeed72db68f780a567ff3e628fb25e26edbb9e33fc202e6d218
Stable URL of this page
https://metraipolis.com/documento/dc_c5af189fda31

How to cite

Stable Policy Learning — arXiv:2609.19418 — arXiv (en), https://arxiv.org/abs/2609.19418 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c5af189fda31 (Metraipolis archive dc_c5af189fda31, sha256 3f660aeccf4b1eeed72db68f780a567ff3e628fb25e26edbb9e33fc202e6d218)
BibTeX
@misc{dcc5af189fda31,
  title = {Stable Policy Learning — arXiv:2609.19418},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.19418},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c5af189fda31; sha256 3f660aeccf4b1eeed72db68f780a567ff3e628fb25e26edbb9e33fc202e6d218},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.19418. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

In evidence-based policymaking, typically one experimental sample is observed, then a learned policy recommendation is implemented at scale. Policies learned from the experimental data can perform well in expected welfare, yet random sampling in the experiment can produce recommendations with poor welfare outcomes. In this paper, we ask: how should policy learning algorithms balance expected welfare against sampling risk? Our main contribution is to show that algorithmic stability plays a central role in characterizing and navigating the tradeoff. Intuitively, if a policy learning algorithm's recommendation remains stable when one experimental unit is replaced, then that algorithm has limited sampling risk. We propose a method for policy learning called policy-vote bagging, which learns treatment decisions on many subsamples then averages their votes into treatment probabilities. Relative to using one subsample, averaging across subsamples preserves expected welfare and improves expected utility for a risk-averse researcher. We derive sharp bounds linking estimation accuracy, subsample size, and welfare variation, including an exact guarantee under CARA utility.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.LG).
Categorie arXiv: Econometrics (econ.EM) ; Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML).
Pagina del lavoro: https://arxiv.org/abs/2609.19418 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-10T03:48:18.393Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).