AI Model Research Documentation Sheet (AIRDocS)

As the field of AI research increasingly relies on comprehensive documentation of datasets and models, documentation sheets become critical for enabling transparency, reproducibility, and accountability. Inspired by principles such as those in datasheets for (cultural heritage) datasets (Gebru et al., 2021 or Alkemade et al. 2023), the AI Model Research Documentation Sheet (AIRDocS) extends these concepts to model usage to address the need for documentation for researchers or students working with language/vision/multimodal models. The sheet offers a structured approach to documenting the context, content, and processes of AI research, ensuring that both data and model usage are communicated effectively. Similar to instruction manuals, AIRDocS provides essential metadata on model usage purpose, bias assessments, and ethical considerations. By including this documentation sheet with research papers, authors provide educators, readers and reviewers with detailed insights into their methodological choices, technical implementations, and experimental conditions. This standardized approach to documentation enables better evaluation of research findings, facilitates replication studies, and promotes responsible research practices across different AI domains.