What Artificial Intelligence Cannot Replace in Ecology?

Juline Rodrigues da Conceição

Artificial intelligence is becoming increasingly integrated into ecological research, from literature synthesis and hypothesis generation to statistical modelling and data analysis. As AI systems become more autonomous, discussions about ecology's future are often framed as a competition between human and machine intelligence. Yet this perspective overlooks a more fundamental question: what aspects of ecological reasoning can be delegated to AI, and which aspects remain the responsibility of the researcher? To explore this issue, 11,749 ecology publications indexed in Scopus were analyzed using text mining and Structural Topic Modelling (STM). Publication trends revealed a rapid expansion of AI-related research, particularly following the COVID-19 pandemic and the public release of ChatGPT in late 2022. Topic dynamics indicated growing attention to AI applications, while discussions centered on statistical methodology, inference, and quantitative reasoning became comparatively less prominent within literature. These patterns raise important questions about how technological innovation may reshape scientific reasoning within ecology. Drawing on philosophy of science and discussions of the epistemic limits of AI, this article argues that the central challenge posed by AI is not the replacement of researchers, but the preservation of scientific judgment in increasingly automated workflows. A framework based on three sequential principles for the responsible use of AI in ecology is therefore proposed: (i) Critical Evaluation Capacity, (ii) Intellectual Authorship of Questions and Hypotheses, and (iii) Quantitative Ecology Literacy and Methodological Transparency. Together, these principles clarify which responsibilities can be delegated to intelligent systems and which remain essential to the ecologist's role in producing reliable ecological knowledge.