Scientific judgment cannot be replaced by artificial intelligence

reflections on the education of future researchers

Resumo

Resistance to the use of artificial intelligence (AI) in scientific writing is often grounded not in technophobia but in concerns about the role of intellectual effort in the development of scientific reasoning, particularly among non-native English-speaking researchers. Writing for international peer-reviewed journals is a demanding and formative process through which authors develop conceptual clarity, argumentation, and an identifiable authorial voice. This article examines the practical use of AI during the preparation of responses to reviewers in manuscripts submitted to international journals. Two real cases are analyzed: one involving conceptual clarification about stroke-related mobility measurement and another concerning statistical reasoning and variable selection in regression analysis. In both situations, AI produced grammatically fluent but generic responses that failed to address the underlying scientific logic of the reviewer’s critique. In contrast, the final author responses required explicit articulation of methodological decisions and theoretical reasoning. These examples illustrate that while AI may assist with language refinement and structural editing, it cannot replace scientific judgment or the reasoning processes that precede writing. Empirical evidence from controlled comparisons of human-written, AI-generated, and AI-assisted review articles supports this distinction, showing that AI may reduce writing time but still requires extensive human fact-checking and revision. The implications are particularly significant for students and early-career researchers, for whom writing is a central component of learning how to think scientifically. Premature reliance on AI may produce fluent text while interrupting the development of intellectual independence and authorial voice. AI can therefore be most beneficial after a researcher has already developed a clear conceptual writing style through years of training, supervision, and scholarly practice. In this context, AI may enhance clarity and efficiency, especially for non-native English speakers, without replacing the human reasoning that underpins scientific knowledge. Writing may be accelerated, but thinking remains an irreducibly human process.