The Unasked Experiment: How Scientific AI May Narrow the Space of Discoverable Knowledge

The increasing participation of artificial intelligence in scientific research shifts some consequential epistemic decisions from human investigators to automated selection procedures. Systems may propose hypotheses, rank candidate experiments, allocate scarce resources, interpret observations, and report findings. Yet the adequacy of a completed experiment does not, by itself, establish the adequacy of the choices that excluded other experiments. This conceptual paper examines the scientific status of identifiable experimental alternatives that were considered but not executed. Its central distinction separates justified prioritization, insufficiently warranted exclusion, unrecoverable selection histories, and possibilities that were never generated at all. Drawing on active learning, Bayesian experimental design, autonomous scientific discovery, causal inquiry, and research-integrity literature, the paper proposes an Experimental Exclusion Warrant (EEW): a bounded, claim-relative specification of a non-execution decision and the evidence necessary to reassess it. Three illustrative tables and two conceptual figures differentiate the object of inquiry, its strongest conventional alternatives, and a prospective information-matched evaluation. Four adversarial thought experiments show why successful prediction need not settle explanatory adequacy, why missing logs do not prove a missed opportunity, and why ordinary research auditing may already suffice. The contribution is a limited methodological synthesis, not a discovery of an unknown selection bias, a novel mathematical theorem, or an empirically validated intervention. No new experiments were conducted. The proposed account warrants further consideration only insofar as it makes particular exclusion decisions more contestable than competent existing methods and remains capable of concluding that no additional framework is needed.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24293213
Primary Topic
Philosophy of Biology and Scientific Explanation
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

The Unasked Experiment: How Scientific AI May Narrow the Space of Discoverable Knowledge

Rogério Figurelli
2 links
Zenodo (CERN European Organization for Nuclear Research)
Philosophy of Biology and Scientific Explanation
preprint

The Unasked Experiment: How Scientific AI May Narrow the Space of Discoverable Knowledge

Rogério Figurelli
preprint en

Abstract

The increasing participation of artificial intelligence in scientific research shifts some consequential epistemic decisions from human investigators to automated selection procedures. Systems may propose hypotheses, rank candidate experiments, allocate scarce resources, interpret observations, and report findings. Yet the adequacy of a completed experiment does not, by itself, establish the adequacy of the choices that excluded other experiments. This conceptual paper examines the scientific status of identifiable experimental alternatives that were considered but not executed. Its central distinction separates justified prioritization, insufficiently warranted exclusion, unrecoverable selection histories, and possibilities that were never generated at all. Drawing on active learning, Bayesian experimental design, autonomous scientific discovery, causal inquiry, and research-integrity literature, the paper proposes an Experimental Exclusion Warrant (EEW): a bounded, claim-relative specification of a non-execution decision and the evidence necessary to reassess it. Three illustrative tables and two conceptual figures differentiate the object of inquiry, its strongest conventional alternatives, and a prospective information-matched evaluation. Four adversarial thought experiments show why successful prediction need not settle explanatory adequacy, why missing logs do not prove a missed opportunity, and why ordinary research auditing may already suffice. The contribution is a limited methodological synthesis, not a discovery of an unknown selection bias, a novel mathematical theorem, or an empirically validated intervention. No new experiments were conducted. The proposed account warrants further consideration only insofar as it makes particular exclusion decisions more contestable than competent existing methods and remains capable of concluding that no additional framework is needed.

Zenodo (CERN European Organization for Nuclear Research)
Philosophy of Biology and Scientific Explanation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon