The Generous Question

Every question is a small offer. It says something about who you think is sitting across from you, before anyone opens their mouth.

A generous question is one that assumes the person already has a working model of the thing you are asking about. It does not ask them to prove they are worth talking to. It gives them room to show you the model they already carry, including the parts they suspect are shaky. This piece is a short field note on how to write questions in that shape, and why the shape matters.

What generosity is doing, underneath

In the active-inference frame this site works from, a mind is a generative model that predicts its world and updates on prediction error (Class E, standard formulation in Parr, Pezzulo, and Friston, 2022). A learner is not an empty container waiting to be filled. They arrive with priors, with a partial map, and with uncertainty they can usually locate if given a moment. A generous question invites the map. A stingy question tests for the label.

Our facilitator prompts and intake templates are configured to treat every answer as a snapshot of a current model rather than as a scored response (Class C, this is how the workshop protocol and the /api/intake capture logic are wired). When the framing is show me your model, an answer that is partly wrong stops being a personal indictment and becomes the specific place the next move can start from.

Three moves that make a question generous

Assume the model exists. Start the sentence as if the person has already been thinking about this. "You have probably run into a version of this before. What did the last one look like for you?" is a very different opening than "Do you know what X is?" The first says, you belong in this conversation. The second holds an entrance exam at the door.

Ask for the shape, not the label. A definition is easy to fake and hard to learn from. A description of how the person thinks a thing hangs together is neither. "How would you explain the way this fits with the piece we did last week" gives a facilitator far more to work with than "what is the term for this."

Name the fog on purpose. Add an explicit invitation to point at what is uncertain. "Where does this feel solid, and where does it feel foggy for you right now" gives the learner permission to hand you an incomplete map without apologizing for the gaps. That permission is the whole game. Without it, people default to guessing what the facilitator wants to hear, and the room's real model stays hidden.

A worked example

Stingy. "What is prediction error?"
Generous. "You have almost certainly had the experience of being sure something would go one way and then being surprised. If you were teaching a colleague what surprise is doing for a mind, what would you want them to notice about the moment right after?"

The stingy version is a checkpoint. The generous version is an invitation to sketch a model of a familiar experience, and it makes the answer legible whether the person has ever seen the phrase "prediction error" before. Nobody has to hide.

What is not yet demonstrated

Whether consistently generous framing measurably shifts learner participation and uncertainty-disclosure over the arc of a cohort is (Class U) not yet established at the scale we would need to make a general claim. We are logging the pattern through the workshop's repair-loop and question-shape records, and we will publish the falsifier alongside any positive result: if the shift does not appear in the data, this piece gets revised or retracted (Class F).

A quiet close

A room learns in the shape of what it is asked. A generous question does not lower the standard. It moves the standard from can you produce the right token to can you show me your working model, including where it bends. That second standard is harder, more honest, and, in every session we have run so far, is the one that leaves people with something they can still use on Monday.