Naming Our Values

A congregational values exercise for more thoughtful AI decisions

Facilitator Guide  ·  Draft for feedback  ·  AI Pastoral Toolkit

Purpose

This exercise helps a congregation's leadership team (clergy and lay leaders together) move from vague agreement ("we value community") to a working "filter" — values specific enough that a real decision can be run through them and get an answer. It produces two linked outputs: a general values filter, rooted in the congregation's own stories, and a second, shorter layer that translates those values into guidance for AI-related decisions specifically.

The process deliberately starts with story, not a list of words. A congregation's values are easier to recognize in a memory of the church at its best than to select from a menu — and stories generate behavior-level detail (what did people actually do?) that a word like “Integrity” on its own cannot.

Two versions are provided: a 60–90 minute single session, and a retreat-style version breakable into three or four shorter sessions over several weeks. Both are built for a congregation's leadership team, not the full congregation — though either can be adapted for broader participation.

What You Need

• A room, a whiteboard or large paper, markers, sticky notes or index cards.

• This guide, including the case bank excerpts below.

• Optional: a list of values [TO BE CREATED?], for anyone who gets stuck finding a word for what a story revealed. It is a backup vocabulary aid only — it is not where the exercise starts.

• A designated notetaker to capture values, behaviors, and gaps as they surface.

The Process, in Four Moves

Move 1 — Harvest values from stories

Ask each participant to bring, or tell live, a short story: a moment when this congregation was at its best — a time they'd hold up and say, “this is who we really are.” After each story, ask the group: what value was actually operating here? Let the group name it in their own words. Resist the urge to reach for a pre-made list; if someone is genuinely stuck, that is the moment to offer Brown's list as a vocabulary aid, not a starting menu.

Move 2 — Cluster and define

Cluster repeated or overlapping values on the board. Narrow to three to five — a filter only works if it is short enough to actually hold in mind during a real decision. For each chosen value, draw the behaviors directly out of the stories just told:

• A supporting behavior: what did people in the story actually do that lived this value out?

• A “slippery” behavior: what would look like this value from a distance, but isn't — the thing that lets someone claim the value while missing its substance?

Example: if the stories keep surfacing staying at a hospital bedside past visiting hours, the value might be named Presence, and the slippery version is sending a card and moving on.

Move 3 — Test the values against real cases

A values statement that has never been tested against a hard case is just a hope. Use the case scenarios in the appendix below — drawn from your group's own design sprint work — to pressure-test the draft filter. For each case: given our stated values, what would we actually do here — and does that answer feel right? Where it doesn't, that's data, not failure.

Move 4 — The gap check

When the group's honest answer to a case doesn't trace back cleanly to any named value, stop and ask, in order:

• Is this genuinely a new value we haven't named — or is it a behavior of a value we already have, just defined too narrowly?

• If it is new, does it hold up to Brown's original test: does this define us at our best? Is this a filter we'd actually use again, not just once?

A gap that shows up on one unusual case may just be a judgment call. A gap that keeps showing up across cases is a sign the filter is incomplete — case-testing is meant to revise the list, not just confirm it. In the 60–90 minute version, do this gap check once, at the end, after the single case you test. In the retreat version, do it after each case, so the specific case that revealed the gap doesn't get lost by the time you circle back.

Applying the Values to AI [note from MAMD: this section needs fleshing out]

Once the general values have survived non-AI case-testing, translate them into an AI-specific layer using the case studies elsewhere on the document. The idea was to use some of the case studies and see how the identified values would support a faithful response/good policy; if the values are inadequate to the task, do we need to add some?

Case Studies

(Note: this would link to the case studies elsewhere on the site. All 22? Or the 5 more fleshed out ones?)