Work

Human authorship in artificial intelligence-assisted research: A counterfactual boundary test

Year: 2026
Type: article
Abstract: Artificial intelligence (AI) can already participate in literature retrieval, question generation, coding, analysis, interpretation, and manuscript production, making fixed task-based rules about what should remain human increasingly unstable. This conceptual article introduces a Counterfactual Boundary Test for distinguishing why human involvement is retained in AI-assisted research. The test asks whether the rationale for human involvement would disappear if the relevant AI limitation disappeared. If so, the ground is capability-contingent; if the rationale persists because the work continues to be substantively attributed to human authors, the ground is human-authorship-related; other ethical, legal, fiduciary, political, or social grounds may remain distinct. Current hallucination and verification risks illustrate the first case, whereas comprehension, critical evaluation, justified acceptance, and accountability provide a parsimonious diagnostic synthesis for examining the second. The framework does not claim that all future science must remain human. Instead, it clarifies which rationales may move with technological capability and which arise specifically under human-authored science. These indicators do not replace the cumulative International Committee of Medical Journal Editors (ICMJE) authorship criteria. A worked citation-validation example illustrates selective revision of controls, and a specialist-coauthor case explains justified reliance.
Language: English
Open Access status: gold
Estimated APC paid: $500
OpenAlex ID: W7236563722
FWCI:
Cites: 18
Cited by:
Related to: 10