Model Cards

As part of our AI development governance, we introduced Model Cards for any AI-powered tools, features, or products we build.

The idea is that anytime we develop something AI-powered, we document it in a standardized way. This is introduced for a few reasons:

• Generalizable metrics: Ensuring the evaluation metrics are not inflated, and actually reflects how well it'll work in production.
• Reproducibility and re-evaluation: AI models are moving fast. Having a good documentation allows us to swap out underlying models more easily without losing track of what did and how we did it.
• Disclosure: The Model Card directory will be a public inventory of everything AI-powered at FLP to fulfill user disclosure requirements.
• Transparency: Transparency is a core FLP value, and doing this well allows us to be a leader in this space.
• Community leadership: Volunteers and researchers can see our approaches, spot gaps, and suggest improvements.