skills/mightshape/references/facilitator-practice.md

# Expert facilitator practice

Load this reference when method selection or adaptation is non-obvious, group or power dynamics matter, work stalls, the user asks why or how a method works, or a substantial workshop needs debriefing. A straightforward first participatory prompt uses the participation contract without loading this file. It deepens facilitation without turning every MightShape response into a workshop.

## Facilitate the bottleneck, not a stage label

First name the decision and the learning bottleneck:

| Bottleneck | Useful moves | Avoid |
|---|---|---|
| Context is thin | observation, story-first interview, journey/process reconstruction | ideating from demographic stereotypes |
| Frame is vague or solution-laden | evidence sort, needs as verbs, tensions, competing POVs | polishing one assumed problem statement |
| Possibilities are adjacent | protected brainwriting, analogy distance, inversion, changed constraints | ranking the first plausible features |
| Mechanism is uncertain | storyboard, roleplay, concierge, Wizard of Oz | production fidelity |
| Use is confusing | task-based prototype test, observation, feedback matrix | explaining the interface during the test |
| Feasibility is uncertain | bounded technical proof with representative boundary cases | a full Design Thinking cycle for a technical measurement |
| Alignment is blocking action | independent capture, visible assumptions, criteria and decision record | forced consensus |

A nominal mode is a navigation aid. Select and adapt the method based on what must be learned next. A mature product can return to Define; an early hunch can be prototyped when making is the cheapest inquiry.

## Set up an exercise

Before beginning, establish only what the activity needs:

1. **Decision:** what later choice should this work improve?
2. **Purpose:** what should be clearer, broader, or testable at the end?
3. **Inputs:** which observations, cards, frame, or constraints are legitimate inputs, with provenance?
4. **Mindset:** divergent, convergent, research, sensemaking, or learning-through-making.
5. **Boundary:** timebox or a meaningful stop condition, not arbitrary completion theater.
6. **Participation:** solo, paired, group, remote, asynchronous, or AI-facilitated; name who decides and who may be affected but absent.

Do not run a method without its minimum inputs. If the input is too weak, preserve momentum by producing a preparation artifact or switching methods with the user's agreement.

## Intervention ladder

`source_family: mightshape_original` — an operational synthesis informed by the maintained
source inventory, not a claim of third-party authorship.

Use the least disruptive intervention that restores useful work:

1. **Clarify** the current prompt or distinction.
2. **Reflect** a pattern, contradiction, text-visible participation/pace change, or missing voice without judging it. Do not claim to perceive unobservable room emotion.
3. **Reopen** the question or boundary when the room converged too soon.
4. **Change the stimulus** with an extreme user, different environment, new constraint, or distant analogy.
5. **Change the representation**: words to sketch, sequence to map, abstract frame to roleplay, features to mechanisms.
6. **Change the participation structure**: silent brainwriting before discussion, pairs before plenary, anonymous input, round-robin only when it improves airtime.
7. **Change the method** when the actual bottleneck differs from the one assumed.
8. **Pause and debrief** when safety, fatigue, power, missing evidence, or diminishing learning makes continuation irresponsible.

When ideas stall during protected divergence, increase distance or change representation; do not begin ranking. When conversation dominates idea capture, return to short headlines and visible cards. When one voice anchors the room or hierarchy is salient, gather independent input before open discussion.

## Group craft and psychological safety

- State that ideas, interpretations, and people are different objects; critique the work only in a convergent phase.
- Make contributions visible without assigning status by seniority, fluency, volume, or speed.
- Separate presenter and facilitator when a critique, power difference, or sensitive topic makes dual roles unsafe.
- Do not pressure disclosure of personal experience. Offer pass/skip and alternative ways to contribute.
- Notice who bears the burden of a proposed intervention and who is absent from the room.
- Treat facilitator interpretations as `DESIGN_COUNCIL`, never participant or domain evidence.
- Name positionality and context limits. An AI facilitator is not a certified practitioner, a neutral observer, or a substitute for local lived experience.

## Time and stop conditions

Timeboxes create energy, but learning determines the exit. Useful stop/transition signals include:

- new contributions repeat existing mechanisms rather than adding conceptual distance;
- every source card has a provisional home or a named outlier/boundary status;
- the group can state at least two competing interpretations and what evidence distinguishes them;
- one pivotal uncertainty and the cheapest faithful test are explicit;
- the prototype can answer the named question without additional fidelity;
- a debrief identifies what changed and the next decision.

Extend an activity only when a specific additional round is likely to add decision value. Do not extend it to complete a template.

## Method adaptations

### Brainstorming and brainwriting

- **Use when:** a generative, solution-independent frame exists and conceptual breadth is the bottleneck.
- **Set up:** one clear HMW or mechanism question, visible divergence rules, short capture headlines, protected time.
- **Watch for:** discussion replacing capture, idea ownership, feasibility critique, feature variants masquerading as breadth, early voting.
- **Recover:** silent brainwriting; a distant analogy; remove technology; invert the beneficiary; scale to one or 10×; switch from words to sketches.
- **Debrief:** identify territories, surprising mechanisms, combinations, outliers, and assumptions—without selecting solely by popularity.

### Affinity clustering

- **Use when:** multiple source-linked records need sensemaking.
- **Prerequisite:** atomic cards with stable IDs and provenance; clustering invented claims produces tidy fiction.
- **Set up:** sort silently first when hierarchy or anchoring matters; allow boundary cards, duplicates only as linked views, and named outliers.
- **Watch for:** labels chosen before grouping, categories copied from an org chart, contradictions hidden, minority cards discarded.
- **Recover:** sort by a different relationship—sequence, goal, breakdown, actor, cause, workaround—or ask which card least belongs.
- **Debrief:** compare at least two plausible arrangements; distinguish source records from facilitator cluster interpretations.

### Journey and process mapping

- **Use when:** sequence, handoffs, actors, environments, decisions, or breakdowns carry the insight.
- **Prerequisite:** an actual or explicitly hypothetical experience boundary. Do not fill unknown steps for visual neatness.
- **Set up:** name start/end, actor lanes, evidence status, and the granularity relevant to the decision.
- **Watch for:** ideal process replacing observed process, emotion invented by the team, backstage labor disappearing, edge paths omitted.
- **Recover:** zoom into a breakdown; trace one object or piece of information; switch actor; mark an unknown transition rather than guessing.
- **Debrief:** identify consequential handoffs, waits, workarounds, failure recovery, and what requires observation.

### POV and How Might We

- **Use when:** grounded observations need opportunity-revealing frames.
- **Prerequisite:** a user/context, solution-independent need, and insight that explains why the need is difficult or surprising.
- **Watch for:** needs containing products, generic users, insights that merely restate observations, an HMW that encodes the original feature.
- **Recover:** move up/down Why–How; write behavioral, systems, and counterintuitive alternatives; remove the proposed artifact from the sentence.
- **Debrief:** keep competing POVs when evidence cannot discriminate; name what new evidence would change the frame.

### Interview preparation

- **Use when:** lived behavior, workflow, meaning, or context is a material uncertainty.
- **Set up:** research goal, participant relevance, consent/privacy, Solution Blackout, story-first openings, adaptive topics, stop conditions.
- **Watch for:** leading or compound questions, hypothetical preference, premature concept reveal, unnecessary PII, treating one story as prevalence.
- **Recover:** ask for the last real event, reconstruct what happened next, and follow breakdowns/workarounds before interpretation.
- **Debrief:** separate excerpts/behavior from interpretations and unknowns; revise the guide before adding participants.

### Prototype to learn

- **Use when:** making or enacting something is the cheapest way to resolve a pivotal uncertainty.
- **Set up:** one hypothesis, critical assumption, learning question, minimum faithful form, participant/context, supported signals, explicit `DO NOT BUILD`.
- **Watch for:** prototype becoming an MVP, approval seeking, unsupported numerical thresholds, fidelity unrelated to the question.
- **Recover:** move the uncertain interaction behind the curtain, enact manually, remove accounts/integrations, or compare two mechanisms.
- **Debrief:** supported, weakened, falsified, or inconclusive; behavior and surprises; what frame/idea/prototype changes next.

## Debrief every substantive exercise

A debrief is not a recap. Ask or answer:

- What changed in our understanding or option space?
- What surprised us or resisted the pattern?
- Which conclusions are evidence, inference, design material, assumption, or unknown?
- What remains unresolved or unsafe to generalize?
- What decision is now possible, and what is the smallest next learning move?
- Did the method help the bottleneck? What should be adapted next time?

Keep the debrief proportionate. A two-minute exercise needs a few sentences; a consequential session needs an inspectable decision record.

## Expertise boundary and attribution

MightShape draws on publicly described design thinking practices and clearly labeled supplemental practice. It is independent and does not claim third-party certification or endorsement. Express guidance independently; never reproduce protected worksheets, scripts, artwork, or source-card layouts, and never claim one universal process.