ABOUT ML DRAFT IS CURRENTLY CLOSED TO PUBLIC COMMENT
The V0 Draft is now closed to public comment to allow the Steering Committee to integrate comments and for input from the Diverse Voices process. V1 will be released for public comment in early 2020.
Read the v0 draft
This Version 0 draft is presented in 5 chapters, 1 supplemental ML primer, and 1 appendix of documentation questions from a selection of research papers. For audience members who do not have a strong background in machine learning, we recommend starting with the Supplement: Primer on ML Lifecycles document for background knowledge and definitions of terms used throughout the version 0 draft.
Chapter 1-5 are the documents for public comment. We welcome all comments, anecdotes about experiences with documentation practices, suggestions for pilots, and beyond. Future drafts will incorporate feedback from the public and the Steering Committee and may include deep dives on other enablers of ML transparency, such as team and institutional setting, ML system-level considerations, test suites, feedback loops.
Supplement: Primer on ML Lifecycles
High level explanation of portions of the ML lifecycle, including defining machine learning system, describing types of data sets and defining ML model
Ml PrimerChapter 1: Project Overview
Provides a statement of importance for the ABOUT ML project as well as more details on the process and timeline
Chapter 1Chapter 2: Current Recommendations on Documentation for Transparency in the ML Lifecycle
Currently the bulk of ABOUT ML recommendations. A summary of existing research in the documentation for transparency space, context on why documentation is a process as well as an artifact, and deep dives into documentation questions and considerations for different portions of the ML lifecycle
Chapter 2Chapter 3: Current Challenges of Implementing Documentation
Brief summary of existing themes in challenges for implementation. We invite comments here that share any and all challenges experienced in the process of experimenting with documentation for transparency
Chapter 3Chapter 4: Evaluating Best Practices
[Currently empty, will be filled in future releases] We invite comments here to share any solutions or practices that have worked for increasing transparency via documentation. Anecdotes, code, process documents or any other formats all welcome.
Chapter 4Chapter 5: Conclusion
A brief summary of Chapter 1-4. A new conclusion section will be appended for each ABOUT ML release
Chapter 5