Beyond principlism: Practical strategies for ethical AI use in research practices

Abstract principles alone are insufficient: The paper identifies a "Triple-Too" problem where there are too many high-level ethical initiatives, principles that lack practical relevance, and excessive focus on restrictions rather than benefits. Researchers need concrete, actionable guidance rather than abstract ethical frameworks to navigate AI use in their daily work.

Five specific, practical goals should guide AI use in research: Rather than vague principles, the paper proposes five concrete objectives: understanding model training and bias, protecting privacy and copyright, avoiding plagiarism, comparing AI benefits against existing alternatives, and ensuring transparent and reproducible use. Each goal includes specific strategies and real-world examples of misuse.

Utility comparison matters more than isolated performance: The paper argues that ethical AI application requires evaluating whether AI actually performs better than existing alternatives in a given research context, rather than simply assessing the AI system's performance metrics in isolation. This shifts focus from "can AI do this?" to "should we use AI for this task?"

Documentation and transparency are essential for responsible use: The paper proposes documentation guidelines to enhance transparency and reproducibility, suggesting that researchers should clearly record how, why, and what AI tools were used in their work—making AI use in research verifiable and accountable.

User-centered, realistic approaches work better than technological fixes alone: The paper rejects over-reliance on technological solutions and rigid rule-following, instead advocating for practical strategies grounded in how researchers actually work, with realistic case studies of both misuse and corrective measures.