MDPE.9 - Generate and Compare Music-or-Dance Performance Candidates

Pattern type: Music-and-Dance performance-candidate pattern

Public result: several identified candidates with supported retain, revise, branch, or reject decisions

MDPE.9:0 - Use This When

Use this pattern when a composer, choreographer, performer, teacher, live coder, human–AI team, or other practice-developing Agent has one plausible answer and is about to treat it as the only serious candidate. Also use it when many takes, phrases, model outputs, arrangements, figures, or rule sets have accumulated but nobody has exposed their consequences in the Music or Dance Work that matters.

The costly errors lie at both extremes:

• the first fluent, familiar, visible, or easily notated answer is accepted before breath, body, partner, instrument, timing, control, rehearsal, or recovery demands become visible; or
• generation continues without a receiving decision, so an archive of variants replaces comparison and later Work cannot tell what to retain, revise, branch, or reject.

The first useful result is a small set of materially different candidates, each kept under its actual kind and tried in proportion to the decision, with observations, evidence limits, and one next treatment.

Use the applicable composition, arrangement, orchestration, choreography, improvisation, live-coding, editing, generative, co-creative, constraint-design, or other domain Method to make candidates. MDPE.9 does not replace those Methods. It adds the shared engineering move of challenging the incumbent answer, involving the intended performer, and returning bounded dispositions.

Do not use this pattern when a direct domain Method already returns the same candidates, trials, comparison, and next decisions; when only one settled work is needed; when the gap is performing capability (MDPE.10) or configuration (MDPE.3); or when the current question is archive, open-ended search, or accountable final choice (C.18, C.19, or C.11).

MDPE.9:1 - The Recurring Difficulty

Music-and-Dance creation never starts from an empty possibility space. Learned Methods, bodies, instruments, software, partners, venues, audiences, notation, and cultural memory make some variants easy to produce and others hard even to notice.

Those constraints are productive, but they also hide assumptions. A notation system makes written differences easy to compare while bodily control remains invisible. A model can produce many fluent samples while rehearsal cost and performer agency remain hidden. A teacher can vary visible figures while preserving the same partner relation—or preserve the figure while losing the weight change that made it work. A camera can make a phrase look clear while contact, effort, timing, or navigation fails for the dancers.

Ask which materially different candidates expose the current assumption and which observation involving a performer will change the next creation decision.

Each candidate keeps its domain kind. Depending on the case, a candidate may be musical or movement material; a score, choreography, prompt, or another description used in performance Work; or a specification of constraints intended to obtain during Work. Keep material, descriptions, constraint claims, performed relations, Work occurrences, and performance occurrences distinct.

MDPE.9:2 - Solution

Generate and compare enough difference to challenge the current commitment. Stop when the next creation decision can be made. Keep capability, production readiness, cultural recognition, and population-wide selection as separate claims that local comparison has not established.

MDPE.9:2.1 - Name the use and incumbent assumption

State the receiving use, intended performer, and decision. Examples include choosing material for an ensemble rehearsal, selecting movement relations for a social-dance exercise, deciding which human–AI interface deserves another prototype, or preserving branches for later choreography.

Then state the incumbent assumption in a form that another candidate can challenge. For example:

• denser doubling is needed to preserve harmonic identity;
• the visible figure carries the partnered timing;
• higher model fidelity gives a dancer more useful control;
• a symmetric duet rule is easier to perform and read;
• the most popular take is the best concert arrangement.

A genre label or request to “be more creative” is not a testable incumbent assumption.

MDPE.9:2.2 - Produce materially different candidates

Apply suitable domain Methods and keep each candidate identifiable. Record only the derivation, edition, configuration, or distinguishing relation needed for the current comparison.

A candidate is materially different when the difference could change the receiving decision. Change rhythm, attack, instrumentation, form, partner relation, body organization, spatial relation, interface, generative rule, constraint, recovery mechanism, or another relevant feature. Renaming a copy or changing an irrelevant parameter does not add useful diversity.

Prefer a few candidates that pull the decision in different directions over many near-duplicates. Preserve a branch when its trade-off or stepping-stone value may matter later; do not force every difference back into one revised identity.

MDPE.9:2.3 - Choose a proportionate performer-involving trial

For each serious candidate, choose the cheapest Work that can expose the relied-on consequence:

• notation, rendering, or simulation can expose range, balance, geometry, latency, or feasibility;
• a listening pass can expose timbre, masking, form, or perceptual distinction;
• an embodied walkthrough can expose reach, weight, effort, contact, balance, or control;
• a rehearsal fragment can expose breath, cueing, ensemble timing, recovery, and instrument interaction;
• a pair, social-floor, battle, class, installation, or audience probe can expose relations that isolated studio Work cannot.

Use the intended performer or performing Agent where practical. A justified proxy can answer a narrower question, but state what it cannot show.

When a candidate is a constraint specification, identify the participants, relations, and conditions it proposes. Observe the later Work to establish whether those relations held. Producing the rule card does not produce the duet relation.

MDPE.9:2.4 - Separate observation from judgement

Keep the trial, observation, and decision apart. A dancer’s felt control, partner report, audio recording, motion trace, rehearsal interruption, listener discrimination, model score, or audience response can each bear on a claim. None automatically supplies the whole judgement.

Interpret the observations for the named use. A candidate can be novel but unusable, technically feasible but artistically wrong, effective for one body but unsafe for another, or valuable only as a branch. Avoid collapsing control, effort, quality, novelty, popularity, fidelity, and cultural recognition into one score.

MDPE.9:2.5 - Give every candidate a next treatment

Return one scoped disposition:

• retain — use or compare the candidate again under the stated conditions;
• revise — change the named feature and retry under a stated condition;
• branch — preserve the difference as a separate line or for another use;
• reject — stop using it for the stated performer, conditions, and purpose.

Do not write “reject globally” unless the claim truly covers all relevant performers and uses. A disposition does not authorize later Work, select a production programme, establish performing capability, or show that a wider practice recognized or retained the candidate.

MDPE.9:3 - First Useful Result

Use one short comparison table:

Above the table, state:

• Receiving use and decision: [what later Work or commitment will use the result].
• Intended performer: [performing Agent, ensemble, robot, hybrid, or other admitted System].
• Incumbent assumption: [the claim the candidate set must challenge].
• Comparison grounds: [characteristics that can change the decision without being collapsed into one score].
• Reopen condition: [changed performer, use, configuration, evidence, or live alternative].

Use the candidate dispositions within their stated trial and receiving-use limits.

MDPE.9:4 - Worked Cases

MDPE.9:4.1 - Human–AI movement generation

A movement-generation project compares model and interface variants with Voguing and Dancehall dancers. High, low, and intermediate fidelity can create different relations between recognizability, dancer interpretation, control, surprise, and tool-development value. A model score alone cannot decide among them.

The project identifies the generated movement material and interface edition, lets dancers improvise with each candidate, records what they could recognize, control, adapt, or reject, and returns separate dispositions for choreographic stimulus and tool design. A low-fidelity candidate may remain useful as a stimulus while a more faithful one serves another use. The reported study supplies a direct co-design Method; another project must not copy its dispositions without new trials.

MDPE.9:4.2 - Three music arrangements

A composer prepares a work for singer, clarinet, cello, and live electronics. The intended relation is antiphonal; breath and a recoverable electronic failure mode matter. Three arrangements differ materially:

• dense spectral doubling preserves pitch content but may mask the antiphonal relation;
• sparse instrumental answers preserve space and breath;
• generated responses allow variation but add latency and recovery demands.

Rendering exposes range and balance. A short rehearsal with the intended musicians exposes breath, cueing, latency, and fallback recovery. The first version can be rejected for the current use, the second retained for an acoustic branch, and the third revised for another trial. Those decisions do not yet select a concert programme.

MDPE.9:4.3 - Balboa, brushes, and articulated weight

The Dance profile illustrates the distinction between a continuous carrier and discrete articulation. A teacher can generate partnered phrase variants in which continuing floor contact and movement remain legato while the weight event appears at contact onset, direction change, weight arrival, or partner-readable pressure.

A pair trial asks both dancers where the weight event became usable. One candidate can be retained for clear partner timing, another branched for delayed articulation, and another revised because free-leg texture displaced the standing-leg pulse. These examples guide candidate design; use separate capability evidence for the named pair.

MDPE.9:4.4 - Ableton comping as a direct-Method stop

Ableton Live’s comping Method retains takes, supports audition, copies chosen portions to a main lane, and preserves source material. When that Method already returns the takes, auditions, composite, source relations, and receiving use needed by the project, stop. Apply MDPE.9 only when a broader performance question remains—for example, when two composites create different breath or ensemble demands, or branches must remain available.

MDPE.9:4.5 - Shimon and the development decision

The reported Shimon case supports a bounded claim: after people supplied a four-measure seed, the robot generated the remainder and performed it on marimba. One generated remainder is not a compared candidate set. The case also does not make the robot holder of the team’s later practice-development decision.

MDPE.9 would become relevant if several identified continuations or gesture variants were tried under stated musical conditions and the Agent responsible for the development decision returned bounded dispositions.

MDPE.9:4.6 - Duet constraint specifications

One duet rule says that dancer P changes travel direction only after dancer Q produces a readable pressure-change event. Another permits either dancer to initiate while the other preserves contact through the response. The cards describe candidate constraints; they do not make pressure, initiation, contact, or response occur.

During partnered-improvisation trials, the dancers use the cards and observers record whether the proposed relations obtained and remained readable. One rule may be revised because the pressure event is ambiguous; the other may branch because mutual initiation improves recovery while weakening intended asymmetry.

MDPE.9:5 - Checks and Common Repairs

MDPE.9:6 - Sources and Epistemic Boundaries

For a constraint-specification candidate, name participants, proposed relations, trial observations and the resulting disposition. Use a direct field Method when it supplies the same result with less burden. Reconsider a candidate when it no longer fits the stated receiving use and trial conditions.

MDPE.9:7 - Relations

• MDPE.1 can supply intended use, preservation, and variation constraints. It is not a mandatory preceding step.
• MDPE.10 uses candidates only when their capability demands matter to the present decision; generation does not establish capability.
• MDPE.3 uses a candidate only when it changes the performer, material, instruments, interfaces, timing, control, or another configuration decision.
• MDPE.11 concerns shaping material during Work. MDPE.5 concerns a performance occurrence and its availability. Neither result is the candidate comparison.
• MDPE.13 can use candidates for a recognition-and-selection decision. MDPE.14 can use them for a practice-conflict decision when they change a supported relation or moved burden. Each pattern returns its own result.
• C.17 characterizes differences; C.18 and C.19 govern exploration records and current pools; C.11 governs accountable choice; C.36 governs later cultural relations.
• Systems Engineering supplies general performer, instrument, interface, venue, software, and configuration results. Method Engineering supplies general Method construction, composition, trial, and revision results when a Method is the changed subject. MDPE.9 supplies the Music-and-Dance candidate, performer, trial, and decision specialization.

MDPE.9:End