Counterfactual Inference under Thompson Sampling
Olivier Jeunen · 2025
Read Counterfactual Inference under Thompson Sampling by Olivier Jeunen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-08773.txt.
Description
Jeunen wrote in twenty twenty-five about causal inference for recommender systems that learn by Thompson sampling. Recommenders choose content step by step under uncertainty. Thompson sampling picks actions probabilistically to balance trying new options against using proven ones. Offline evaluation asks what would have happened under a different policy. Standard estimators need the probability of each action taken, called its propensity. Those probabilities are not readily available under Thompson sampling.
Jeunen derives exact formulas for the propensities that are cheap to compute. The formulas cover a variety of parameter and outcome distributions. With them, off-policy estimators become usable in Thompson sampling settings. That enables unbiased offline evaluation of recommender systems. Applications extend to online advertising and personalization more broadly. The work fills a practical gap in causal inference for adaptive systems.
Catalog record
| CREATOR | Olivier Jeunen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-08773