Key Idea: What are FactSheets?
FactSheets are disclosure mediums developed by IBM which docucment a collection of relevant information (facts) about the creation and deployment of an ML model to enable better governance and increase stakeholder understanding.
Watch: IBM Watson AI FactSheets
Watch this video from IBM News, which provides a high-level overview of FactSheets and their proposed features upon organizational adoption.
A few proposed features of FactSheets use highlighted by the video include:
Policy creation via the FactSheets template which defines what data are collected about a model (how a model was created, tested, trained, deployed, and evaluated, what data were used, what regulations or company policies an organization must account for, who can use the model and for what purpose, how it should operate, etc.)
Automated data capture via "model facts" which help to continuously and automatically capture the model details.
Automated reporting via the FactSheet itself which is compiled in a sharable location.
Optional Reading: The Concept Paper
Read the FactSheets concept paper written by Arnold et al.
Consider the similarities and differences between FactSheets and model cards.
Cite as: Arnold M, Bellamy RKE, Hind M, et al. FactSheets: Increasing Trust in AI Services through Supplier’s Declarations of Conformity. Published online 2018. doi:10.48550/ARXIV.1808.07261
Explore: AI FactSheets 360
Explore: FactSheet Examples
Explore examples of IBM FactSheets below: