Facilitators
The definition of good facilitation has to come from people who facilitate. OFL is built so the definition they produce, the evaluations and the knowledge, stays a shared resource they own and govern, rather than data a platform extracts.
Contributions
Judge real facilitator turns, good or weak, in plain language. Those judgments become the evaluations every AI facilitator is measured against. It is how the definition of good facilitation gets written, case by case. The calibration tool is in development; subscribe for updates to hear when the first round opens.
Maintain the patterns, research, and definitions in the knowledge base, so the field's shared reference stays accurate and current. Open the wiki.
The vision
Today, when a facilitator helps train an AI, their expertise disappears into a black box they will never own. OFL is built on the opposite model, closer to a data cooperative like MIDATA in health care, where members own and govern what they produce. The evaluations facilitators write here become the standard the field's AI facilitators are measured against, and that standard stays one facilitators steward, own, and share in the value of, rather than expert judgment sold on by a marketplace they do not control.
We are early, and the governance is still being built. But the direction is set: the people who define good facilitation should own that standard, not rent it back from a platform.
Take part
Help calibrate the evals by judging real facilitator turns. The tool for this is in development; subscribe for updates and we will announce when the first round opens.
Edit and extend the knowledge base. Open the wiki.
Research notes and project updates. Read the Substack.