A high-dimensional framework for readying data for AI

Byna Suren – The Ohio State University

Readying data for AI may mean different activities for different use cases of AI. Widely agreed standards, definitions, and methods for “readying” data for AI are still evolving. Toward that effort, we proposed a high-dimensional framework for defining AI readiness. These dimensions of readiness that are general to various datasets include quality, understandability, data structures, impact or value, fairness, and governance. Additionally, we include a dimension that specifies AI use case-specific requirements for defining readiness. In this talk, we will describe the foundations of this framework with a tool called AIDRIN (AI Data Readiness Inspector) that evaluates and applies data preparation steps automatically or semi-automatically with human-in-the-loop approaches.