What a Data Release Must Preserve for Off-Policy Evaluation

Key points

• In off-policy evaluation, merging decision-time histories that matter differently for a specified policy value or comparison can make that evaluation unidentified, even with observed actions and outcomes and known behavior and target policies.1
• The same radius becomes a release-design objective when each record may be refined into at most $m$ categories.1
• When behavior does not vary within a released record, we give a finite-data repair algorithm whose maximum evaluation-weight radius is at most twice the finite-data optimum.1
• For a fixed action-outcome distribution within a released record, a distinct sharp-identification problem remains.1

Sources (1)

• [1]What a Data Release Must Preserve for Off-Policy Evaluation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 02:27 AM

  We characterize which distinctions must be preserved: identification holds if and only if the action-specific evaluation weights are constant within each released record, and the coarsest exact release groups histories by these weights.

  In off-policy evaluation, merging decision-time histories that matter differently for a specified policy value or comparison can make that evaluation unidentified, even with observed actions and outcomes and known behavior and target policies.

Extractive summary: sentences quoted from the sources.