Disclosure in the era of generative artificial intelligence

Area of Science:

• Academic Publishing
• Artificial Intelligence
• Scientific Communication

Background:

• Generative artificial intelligence (AI) is increasingly used in academic writing for various tasks.
• Current disclosure practices for AI in manuscripts are inconsistent and often binary.
• Existing methods fail to differentiate between minor AI assistance and significant contributions to content.

Purpose of the Study:

• To propose a novel risk-proportional disclosure framework for generative AI in academic writing.
• To address the inconsistencies in current journal disclosure policies.
• To provide a more nuanced approach to reporting AI's role in manuscript preparation.

Main Methods:

• Analysis of current journal disclosure policies regarding AI.
• Development of a risk-based categorization for AI use in academic writing.
• Proposal of a tiered disclosure system based on the level of AI involvement.

Main Results:

• Current binary disclosure systems are inadequate for the diverse applications of AI.
• A risk-proportional approach can better reflect the extent of AI's influence on a manuscript.
• Separating disclosure by risk allows for more accurate reporting of AI's contribution.

Conclusions:

• A risk-proportional disclosure framework is essential for transparency in AI-assisted academic writing.
• Adopting this approach will enhance the integrity of scholarly communication.
• Journals should implement nuanced disclosure policies that account for the varying impact of AI tools.