Expectations and Practices around AI Disclosure in CS Research

Research indicates that current AI disclosure policies across top computer science venues are prevalent but largely underspecified. A survey of computer science researchers revealed that disclosure is considered most necessary for tasks involving research design, especially when AI tools are used, with necessity varying based on the level of human involvement.

Why it matters

The ambiguity in AI disclosure policies in research venues poses a significant challenge for maintaining transparency and ethical standards in AI-assisted scientific work. Addressing these policy gaps is crucial for ensuring research integrity and fostering trust in results generated or supported by generative AI tools, impacting future research directions and public perception of AI's role in knowledge creation.

Key insights

• AI disclosure policies are widespread in leading computer science venues.
• These existing policies are frequently underspecified, lacking clear guidelines.
• Researchers perceive a higher necessity for AI disclosure in tasks related to research design.
• The perceived necessity of disclosure is influenced by the degree of human involvement in AI-assisted tasks.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.23271