<!-- thinking-framework-skills | https://github.com/product-on-purpose/thinking-framework-skills | Apache-2.0 -->

Role-Storming

Ordinary brainstorming stalls in two ways that have nothing to do with how much the room knows: people self-edit the odd idea because it would be theirs to own, and everyone stays trapped in the same default associations. Role-storming attacks both at once by changing WHO is generating. Instead of producing ideas as yourself, you adopt a specific assumed identity - "an eccentric inventor," "a curious ten-year-old," "a street artist," "a rival founder" - and generate the ideas you imagine that persona would offer. The durable cognitive move, stripped of the slogan, is: generate divergent ideas while inhabiting a chosen non-self identity, using that identity as both an inhibition-lowering shield (the idea is the persona's, not yours - ownership distance) and a distancing lens (the persona's standpoint pulls you off your default associations - standpoint distance). The artifact is a persona-tagged idea list: candidate ideas attributed to each adopted role, so the distancing is auditable and the pile is ready for a downstream convergence step. This is a generation engine, not the fixed functional-lens evaluation the perspective family otherwise runs.

When to Use

• Idea flow is being throttled by self-consciousness rather than a shortage of knowledge - the room where two confident voices set the frame and everyone else self-edits, or where the participants' "serious" identity makes a playful or extreme idea feel unsafe to voice.
• The thinking is fixated on one frame and a deliberately foreign standpoint ("how would a regulator / a child / a street artist see this?") could break the fixation and surface candidates the default frame will not reach.
• You want more and stranger candidates at the divergent stage, with a downstream convergence step to follow - role-storming only widens the pool, it does not pick.
• A timid but possibly valuable idea needs a deniable owner so it can be said aloud at all.

When NOT to Use

• You actually need a balanced appraisal of a decision, not more ideas. Generating "as the customer" is not the same as evaluating a choice through the customer's interests. If the job is a multi-mode, convergent judgment of one option, route to think-parallel-perspectives-review (fixed functional lenses - facts, upside, risk, intuition, alternatives - over a single decision, producing a synthesis). Role-storming gives a creative pile, not a verdict.
• You need genuine challenge, not role-played opposition. A persona arguing against the plan is a costume, and role-played dissent underperforms authentic dissent and can even harden the original view. If the goal is real challenge, route to think-authentic-dissent, which engineers for genuine minority dissent rather than impersonated objection.
• The natural persona for the task is an inhibited or narrowing one. This is the method's sharpest, evidence-backed failure mode - the one controlled adjacent result shows the effect FLIPS NEGATIVE with a rigid persona (an "eccentric poet" raised originality, a "rigid librarian" lowered it below baseline). Do not role-storm "as a cautious auditor" or "a risk-averse executive" to get ideas. Mandate an uninhibited or deliberately-foreign persona; forbid inhibited ones.
• You want structured generation by other machinery. If the move is to negate a foundational premise, use think-assumption-reversal; to apply seven transformations to a seed, use think-scamper; to transfer deep structure from a distant domain, use think-far-analogy-ideation; to run silent parallel written generation, use think-brainwriting. Role-storming's engine is identity-adoption specifically; reach for it only when the lever you want is who is generating.
• The exercise is collapsing into caricature. When "the angry customer just complains" and the impersonation becomes the point, you are producing theatrical noise (and risk importing crude stereotypes), not ideas. The value is the distancing, not the performance.

Instructions

When asked to generate ideas via role-storming (or to break a stalled, self-censoring brainstorm), follow these steps:

1. State the generative question and confirm the fit. Write the one-line "how might we / what could we do" prompt the ideas should answer. Confirm the block is self-consciousness or fixation, not missing knowledge, and confirm a downstream convergence step exists. If the real job is to evaluate a decision, stop and route to think-parallel-perspectives-review; if it is to challenge a plan, route to think-authentic-dissent.
2. Choose 3-5 personas, each uninhibited or deliberately foreign. Pick assumed identities chosen for DISTANCE from the default frame - an unconventional maker, an outsider, a child, a rival, a user from a very different context. Apply the persona gate: each persona must be uninhibited or frame-breaking. Reject any inhibited or narrowing persona ("cautious auditor," "risk-averse executive," "rigid bureaucrat") - it suppresses ideas instead of freeing them. Note in one phrase why each persona is expected to free or shift thinking.
3. Generate in character, one persona at a time. Inhabiting one persona, generate that persona's candidate ideas for the question. Stay in standpoint: surface what THAT identity would notice or want, not what you would. Push past the first obvious idea per persona; the point is the ideas your own frame would not reach. Keep each idea a concrete candidate, not a complaint or a performance.
4. Tag every idea with its persona. Attribute each idea to the persona that produced it, so the distancing is auditable and a reader can see which standpoint generated what. This persona tag is what makes the artifact a role-storming output rather than an undifferentiated brainstorm.
5. Sweep for caricature and stereotype. Drop or rewrite any "idea" that is pure stereotype performance (the persona just complaining or mugging) or that imports a crude stereotype. Keep the distancing; discard the impersonation-as-point.
6. Do not converge here - hand off. Mark the list as a divergent pool for a separate convergence step. Optionally flag a few candidates that look most promising or most surprising, but do NOT rank, score, or down-select inside this artifact; that is a different move.
7. Emit the persona-tagged idea list artifact per references/TEMPLATE.md: the generative question, the chosen personas with their distancing rationale, the persona-tagged ideas, and the hand-off note. The deliverable is the tagged list, not a recommendation.

Output Format

Use the template in references/TEMPLATE.md. The deliverable is the filled persona-tagged idea list - the generative question, the personas (each justified as uninhibited or frame-breaking), and the candidate ideas each attributed to the persona that produced it, with a hand-off note to convergence. It is a divergent pool, not a ranked recommendation and not a per-lens evaluation. Never present it as a decision or a judgment.

Quality Checklist

Before finalizing, verify:

• [ ] The generative question is stated in one line, and the block really is self-consciousness or fixation, not missing knowledge.
• [ ] Every persona is uninhibited or deliberately foreign; no inhibited or narrowing persona (auditor, risk-averse executive, rigid bureaucrat) was used to generate.
• [ ] Each persona carries a one-phrase rationale for why it is expected to free or shift thinking.
• [ ] Every idea is tagged with the persona that produced it, so the distancing is auditable.
• [ ] Ideas are concrete candidates, not complaints or stereotype performances; the caricature sweep was done.
• [ ] The output stays divergent - no ranking, scoring, or down-selecting inside the artifact - and names the downstream convergence hand-off.
• [ ] The artifact is not framed as an evaluation of a decision (that is think-parallel-perspectives-review) or as genuine challenge (that is think-authentic-dissent).
• [ ] No overclaiming: this is a practitioner-grade, transferred-evidence generation aid; claim more and more-varied candidates, not a measured gain in idea quality (see evidence/dossier.md).

Evidence

Tier P (governing). Role-storming is a real, named, long-lived ideation heuristic with a clear origin (Rick Griggs, c. 1985) and a place in the creativity-facilitation canon (Arthur VanGundy), and it has a coherent mechanism story (ownership distance lowers inhibition; standpoint distance widens associations). That is the extent of what is directly supported: a respectable practitioner technique with a plausible mechanism. What the evidence does NOT support is the method's own idea-count claim - there is no controlled or comparative study that measures role-storming itself against ordinary brainstorming or brainwriting; the "lowers anxiety / more ideas" framing is Griggs's practitioner assertion, not a measured result, and any circulating effectiveness percentage traces to no primary source. The nearest real controlled evidence (Dumas and Dunbar 2016, the "creative stereotype effect," M-tier) measures a DIFFERENT operation - stereotype priming before a solo divergent-thinking test - and its central result is conditional: an uninhibited persona helps, an inhibited persona HURTS. That result is counted here only as a boundary condition (it shapes the persona gate in the When-NOT wall), never banked as evidence FOR role-storming; borrowing its effect sizes would be exactly the transferred-evidence laundering this library exists to prevent. Transfer caveat: every adjacent result is from human subjects (lab undergraduates, VR participants); none studies role-storming performed by or with an AI agent. For a language model "adopting a persona," the inhibition-release rationale (defeating social self-censorship) may not even apply, while the distancing-via-standpoint rationale might - this is untested. The skill ships as a divergence aid with a hard persona gate, never as a proven idea multiplier. Full grading, sources, and caveats: evidence/dossier.md.

Examples

See references/EXAMPLE.md for a completed persona-tagged idea list on a real decision.