Falsifiable Lifting: Testing Whether Agents Know When Their Observation Space Is Incomplete

This record archives the public preprint draft and source package for “Falsifiable Lifting: Testing Whether Agents Know When Their Observation Space Is Incomplete.” The paper proposes an evaluation framework for observation-space insufficiency in AI agents: whether agents know when their current observation is insufficient, what additional observation would distinguish compatible hypotheses, and what evidence would justify accepting a hidden explanation. This Zenodo record is intended as a public timestamped preprint archive while formal preprint submission is pending Public draft repository:https://github.com/hamlet-lab/falsifiable-lifting