Creative provocations — invention and controlled disturbance¶

Use this reference as an optional divergent pass inside design synthesis. It is inspired by the role of Brian Eno and Peter Schmidt's Oblique Strategies: introduce a well-timed constraint or reframing that interrupts a locally coherent answer. This is an original transformation set, not a reproduction of their card text.

Use it for deliberate invention as well as escaping a rut. A familiar convention may be worth challenging while it still works; explore what a different mechanism could make possible, then shape a tangible probe.

A provocation is not a decision framework and does not outrank product truth. Its job is to expose an assumption, produce a materially different mechanism, and then submit that mechanism to evidence, constraints, and integration.

Activation gate¶

Use a provocation when at least one is true:

all current options preserve the same state, sequence, actor, boundary, or feedback assumption;
the work is producing cosmetic variants instead of different mechanisms;
expertise is narrowing the space before the risky premise has been tested;
a Compare or Probe is stuck and needs a new discriminating move;
the user explicitly asks for brainstorming or an oblique angle.
an unsettled direction calls for inventing an interaction or deliberately departing from convention.

Skip it when the desired behavior is settled, the task is a local correctness or accessibility fix, an urgent incident needs stabilization, or novelty would only delay an obvious safe action.

Use one fitted transformation first. Use two or three only when the user wants breadth or the first one fails to perturb the mechanism. Never dump the whole repertoire into the answer.

Fit the disturbance¶

Diagnose the kind of sameness before choosing a transformation:

Sameness in the option set	Start with
Every option adds another surface, control, or explanation	Subtract, merge roles, background it
Every option preserves the same sequence or cadence	Reverse sequence, delay, change rhythm
Every option gives agency to the same actor	Transfer agency, change witness
Every option treats failure as interruption or loss	Make residue useful, turn failure into evidence
Every option lives at the same scale	Zoom in, zoom out, change the unit
Every option assumes one boundary or category	Move the boundary, split the bundle
Every option optimizes the same goal	Replace the score, protect the anomaly
The idea is abstract and hard to feel	Translate medium, make state tangible

Prefer a transformation that attacks the suspected assumption. When no assumption is visible, select from different families rather than drawing several near-duplicates.

Transformation repertoire¶

The labels below are stable ids for observability. The questions are generative, not requirements.

Structure¶

PV-S1 — Subtract the apparatus. If one layer, modal, dashboard, control, or explanation disappeared, what would the underlying experience need to reveal by itself?
PV-S2 — Split the bundle. Which supposedly single event contains two institutions, motives, time scales, or consequences that should diverge?
PV-S3 — Merge the roles. What happens if two components, actors, or states become one mechanism with two readings?
PV-S4 — Move the boundary. If the feature began one step earlier or ended one step later, where would the meaningful decision move?

Sequence and time¶

PV-T1 — Reverse cause and reveal. Let the consequence or evidence arrive before the explanation. What becomes suspense, strategy, or comprehension?
PV-T2 — Change the rhythm. Replace a single dramatic interruption with an accumulation, deadline, decay, recurring ritual, or slow recovery.
PV-T3 — Design the echo. What residue should a choice leave several turns, screens, or sessions later so that it becomes part of the system's memory?
PV-T4 — Borrow from the next era. Which future pressure can appear now in a small, historically or product-truthful form without implementing the future system?

Agency and viewpoint¶

PV-A1 — Transfer agency. Give the consequential action to the least obvious legitimate actor: the user, customer, rival, system, regulator, environment, or future self.
PV-A2 — Change the witness. How would the same state read to someone who does not control it but must interpret, trust, recover from, or audit it?
PV-A3 — Make self-exposure strategic. What if progress requires the actor to create, disclose, certify, or preserve evidence that can later constrain them?
PV-A4 — Let the system refuse. Where would a legible refusal, delay, or counter-offer create more agency than another available action?

Scale and medium¶

PV-M1 — Change the unit. Reframe the experience around a batch, receipt, relationship, route, promise, or season instead of the current unit.
PV-M2 — Translate the medium. If the state had to be understood as a timeline, map, ledger, conversation, physical object, or ambient signal, which relation would become visible?
PV-M3 — Zoom to the seam. What happens at the handoff between two states, actors, or institutions, and could that seam become the primary interaction?
PV-M4 — Zoom to the system. Which local action is actually a feedback loop, network effect, legitimacy shift, or accumulating debt when seen over time?

Value, error, and constraint¶

PV-V1 — Preserve the anomaly. Which awkward exception contains product or domain truth that a clean solution is erasing?
PV-V2 — Make failure informative. Instead of repeating punishment, what new state, evidence, leverage, or recovery route can failure create?
PV-V3 — Swap the scarce resource. If money, time, attention, trust, legitimacy, information, or reversibility became the binding constraint, how would the mechanism change?
PV-V4 — Replace the score. What would the user pursue if the dominant metric disappeared and the underlying value had to become directly legible?

Productive contradiction¶

PV-C1 — Honor both demands. Treat the apparent tradeoff as a request for a new mechanism: what could make both qualities true without averaging them?
PV-C2 — Intensify the constraint. Temporarily make the hardest constraint absolute. Which simpler structure survives?
PV-C3 — Design the opposite honestly. Build the strongest version of the rejected direction. What valid need was the favored direction hiding?
PV-C4 — Reuse the liability. Which cost, delay, record, or restriction can become feedback, identity, pacing, proof, or future leverage?

Run the pass¶

For each selected transformation:

Name the sameness. State the assumption shared by the current options.
Apply the transformation. Ask the fitted question against the actual artifact and product truth, not in the abstract.
Generate an option delta. Describe the changed state, rule, information flow, interaction, or composition in one or two sentences.
Name the leverage. Identify the immediate value and any independent benefit caused by the same mechanism.
Return to truth. Keep, adapt, or drop the delta using the gate below.

Do not ask the user to pick a cryptic card without context. The agent may show a small hand of two or three fitted provocations when collaborative selection is itself useful, but must explain what each is trying to disturb.

Return to truth¶

A provocative option is eligible for comparison or probing only if it can answer:

Truth: Which locked behaviors, facts, safety limits, and authority boundaries does it preserve?
Difference: Did it change a mechanism, or only the story told about one?
Integration: Which canonical state, rule, component, registry, or production path would carry it?
Value: What immediate user-visible benefit appears without waiting for a speculative platform or later era?
Cost: What becomes harder, riskier, slower, or less legible?
Learning: What rendered state, trace, play session, or safe-to-fail probe can distinguish it from the baseline?

Adapt a useful provocation when its literal version violates a constraint. Drop it when the only defense is novelty, metaphor, or future reuse.

Observable record¶

Add one compact field to the Muriel synthesis for each provocation used:

Provocation: PV-A3 transfer agency
Assumption challenged: enforcement only happens to the player
Option delta: player-generated compliance evidence can later create exposure
Disposition: adapted into the bounded lab-and-recall probe
Signal: players distinguish compliance exposure from a raid and can name recovery

For a durable machine-readable trace, use a versioned event. The record is a decision artifact, not hidden reasoning:

{"kind":"muriel.provocation","version":1,"decision_id":"era4-enforcement","provocation_id":"PV-A3","assumption":"enforcement only happens to the player","delta":"player-generated evidence creates later exposure","disposition":"adapted","evidence_ref":"seeded-era4-lab-trace"}

At minimum record provocation_id, assumption, delta, and disposition. Add an evidence reference when the delta is tested. Use kept, adapted, or dropped; do not imply that use of a prompt makes its result correct.

Anti-patterns¶

Random ceremony — drawing a prompt when the task already has a clear, evidence-backed answer.
Novelty laundering — passing a whimsical metaphor off as product insight.
Prompt dumping — showing the repertoire instead of making a decision.
Constraint amnesia — letting the provocation erase accessibility, safety, domain truth, canonical ownership, or the user's task boundary.
Cosmetic divergence — producing new language or skins around the same trigger, stakes, choice, and recovery.
Untraceable inspiration — keeping a surprising answer without naming the challenged assumption or proof.
Deck dependence — refusing to explore because a literal card draw, random source, or external tool is unavailable.

Sources and adaptation¶

The mechanism is informed by Oblique Strategies and by open repositories that treat creative prompts as constraints rather than answers. GitHub wrappers are useful implementation examples, but their software licenses do not necessarily grant rights to reproduce third-party card text. Keep Muriel's repertoire original and attribute the conceptual influence.

Jake Dahn's oblique-skill demonstrates a small-hand interaction: present a few strategies, select one, and apply it to the current problem.
Hermes Agent's creative-ideation skill demonstrates matching a constraint to the domain and producing concrete outputs instead of abstract inspiration.
18F Methods demonstrates a CC0, inspectable method-card library with a shared structure and vocabulary.
awesome-claude-design demonstrates independent design voices followed by synthesis and preserved minority views.

Muriel borrows the operational patterns—timed constraint, fitted application, structured record, and preserved disagreement—not protected prompt wording.