NSTD.7 - Automated Narrativization and Story Planning
Status: Stable
Type: DPF pattern body
Primary EntityOfConcern: planning, generating and inspecting a narrative produced with a tool, so that its content and form serve the intended use.
NSTD.7:1 - Problem frame
Use this method to generate, inspect or repair a narrative with an LLM, an NLG system, graph-to-text or data-to-text generation, story planning, a schema or search. An existing tool-produced account can itself be the object of evaluation.
Identify the material, the relations the narrative should preserve and the reader’s task. When generating a draft, supply those conditions to the tool. Compare the available draft with the material for omissions, inventions and misleading connections. A finding can answer an inspection request or inform a decision about use or repair.
Fluent or schema-valid output can introduce a cause the source never established, drop a condition or break continuity.
A focused comparison can locate such a loss and guide a correction while retaining the speed of automated drafting.
NSTD.7:2 - Problem
Automated systems can produce fluent narratives that lose a source relation or violate a story constraint. A plan helps control generation, while inspection tests the actual result. Either can reveal a need to revise the other.
NSTD.7:3 - Forces
| Force | Tension |
|---|---|
| Fast drafting vs source comparison | Generation is fast; finding a consequential omission or invention still requires comparison with the needed content. |
| Schema constraints and fidelity | Correct field structure establishes format compliance. Inspect the values and relations to establish fidelity. |
| Fluency and correspondence | A smooth transition can turn succession into an unsupported explanation. |
| Automation and responsibility | A consequential use may require an authorized decision. Identify who holds that authority and, when relevant, who implemented the decision. |
NSTD.7:4 - Solution
For a new account, identify the content and order the use needs, generate the draft and inspect it. A few instructions or marked source relations may suffice for the plan. For an existing narrative, start with the account and the grounds needed for the question. Inspection can give a sufficient finding, support use of the candidate or show what needs repair. Keep a note such as the one below when a collaborator, experiment or continuing workflow needs to recover the choices.
GeneratedNarrativeUseNote@Context:
sourceMaterialRef:
generatedDraftRef:
intendedReaderAndUse:
requiredRelationsOrConstraints:
generationMethodRef?:
contentPlanRef?:
discourseOrEventPlanRef?:
schemaConstraintRefs?:
sourceComparison:
evaluationRef?:
findingUseDecisionOrRepair:
consequentialUseDecisionMaker?:
architectureSynthesisAssessmentRef?:
additionalClaimGrounds?:
revisitCondition?:
Use the contribution that settles the remaining question.
| Question | Applicable content |
|---|---|
| Source selection and an evidence claim | Use NSTD.1 for ordinary source selection, or G.2 to collect and synthesize the relevant SoTA traditions. Support an evidence claim from its subject sources. Use A.10 if the source-to-use relation is implicit or must remain reproducible. |
| Generated content is to inform architecture work | C.35 evaluates the generated contribution for that use. |
| Source-to-narrative relation | A.6.3.NAR and this DPF |
| Reader recovery of selected structure | C.2.8 through NSTD.6; C.33 separately when architecture-description adequacy is the question. |
| Correspondence between structures | Compare the pertinent source and result directly. Use C.34 to evaluate an architecture-specific preservation claim. |
| Generation procedure | Establish which method was specified or used, from its description, source material or evidence of execution as the claim requires. |
| Narrative quality for a reader and use | Use NSTD.6 and its applicable A.19.ECS evaluation or C.16 measurement basis. |
| Repeated quality improvement | E.22 when the quality question is not framed, then E.23 using NSTD.6 result rows and re-evaluation |
| Evidence | Use the subject criterion and relevant sources. Use A.10 for a source-to-use relation that is implicit or must be reproducible. |
| Assurance | B.3 |
| Ethics, harm, bias, affected parties | D.1 through D.5 |
| Responsibility for a consequential use | Identify the applicable assignment or decision authority. If reporting a particular act, establish who did it and on what basis the attribution holds. |
For new generation, planning and drafting can be interleaved. Inspection may reveal that the content or order needs to change. A short task can combine planning in one prompt and comparison in one reading. Use separate stages or records when the tool, experiment or later work needs them.
| Operation | What to do | Typical failure |
|---|---|---|
| Source selection | Select the material and relations or constraints needed for the reader’s use. Keep unresolved source claims visible. | A prompt assumption is reported as an established fact. |
| Content planning | Choose what to include, omit and foreground. Revise the selection when comparison reveals a gap. | The generator silently drops a condition needed for the reader’s action. |
| Discourse or sequence planning | Use NSTD.2 to choose the order, when to disclose events and what the expression then allows a reader with the specified prior knowledge to infer. | A plausible sequence changes chronology, support for a cause, proof dependence or continuity. |
| Realization | Generate wording and viewpoint for the use. For a required voice, use NSTD.4 to choose examples or a trial treatment and compare connected passages. | Smooth prose can introduce a connection or replace the chosen voice with a different manner. |
| Inspection and evaluation | Compare the draft with its sources and constraints; use NSTD.6 for the relevant quality questions. | A defect remains invisible because only fluency or schema validity was checked. |
| Finding, use or repair | Report what inspection established. For a proposed consequential use, apply its criteria and authority. For a repair, compare the changed version with the earlier one before claiming improvement. | Regeneration alone leaves it open whether the original defect was resolved. |
For a required voice, use NSTD.4 to identify the speaker’s relation to the reader, suitable source passages or a treatment to try, the expressive choices and content to preserve, and what the speaker knows under the real or fictional conditions. Include adjoining text when the transition matters. An ordinary prompt can carry this information. Keep a separate profile when later work needs it. If the model receives only a compressed description, identify consequential features that description leaves open.
Compare a connected passage with the selected treatment through NSTD.4:4.5. A request to preserve a voice establishes the instruction. The written output can reveal a lost distinction, invented memory, inappropriate address or broken transition. For an assessment, report the finding. For a repair, revise that passage and inspect its connections with adjacent text. Retain an adequate result. Revisit the choices affected by a changed genre or source example. Use a human or automated comparison where it can supply the particular judgement required.
For schema-governed generation, check which required fields are present, then whether their values satisfy the substantive constraints. A schema may specify scenes, characters, locations, source references, branches or engine inputs. Save constraints that later maintenance needs. Use NSTD.6 for narrative quality, the subject evidence for a factual claim and C.34 for architecture-specific structural preservation.
For LLM-assisted analysis, compare the proposed themes with the excerpts, question and serious alternative readings. Use NSTD.9 when the interpretation remains unresolved. Establish who proposed, challenged and interpreted the theme when its attribution matters. In NITA, the human researcher interprets the evidence and explains the result and its limits.
For fiction or an explicitly hypothetical real scenario, distinguish choosing events from expressing them. Use NSTD.10 if events, consequences or changes in what participants know or want still need construction. Give the generator the established premises, available choices and consequential relations. Compare its proposals with these. A new discovery or motive can lead the author to retain the premises and revise the event, or deliberately change a premise. A short scene may convey its plan. Use a graph or separate plan where it helps maintain dependencies. A plan whose relevant premises or claims are suitable can supply later narrative work.
Choose a further probe when its attainable result can answer a material question about the method, narrative or intended use at a worthwhile cost. Use an adequate existing result when available.
- Source perturbation. Change a constraint that requires a difference in the answer and compare the resulting outputs. If the pertinent relation stays wrong or fails to change, examine how the generator used the source. A successful comparison establishes what happened for the tested change and conditions.
- Structure recovery. Use a reader’s response to the selected task, with access and help specified. To test what can be recovered under conditions excluding the prompt, keep the prompt out of that attempt. Retain the first response and identify what help was used. If a relation is missing, examine whether content, order, access or other conditions explain the loss.
- Responsibility. For a consequential organizational use, establish the authority to decide, correct or withdraw it. A generated note can convey an existing authorization when its issuer and terms support that use. Verify the actual grant if the wording is the only apparent basis.
NSTD.7:5 - Archetypal Grounding
NSTD.7:5.1 - Generated-narrative pipeline: graph-to-text case
An AI agent produces a fluent explanation from a concept graph. Its author compares the draft’s dependencies and examples with the graph and identifies any consequential loss. If correction is needed, the author repairs the account. The optional note below allows that comparison to be repeated.
GeneratedNarrativeUseNote@GraphToTextTeaching:
sourceMaterialRef: concept graph with dependency, example, counterexample and evidence links
requiredRelationsOrConstraints: prerequisite chain, contrast pairs, evidence-return points
generationMethodRef: LLM-assisted graph-to-text workflow
contentPlanRef: selected nodes and relations
discourseOrEventPlanRef: didactic dependency order with a contrast reveal
generatedDraftRef: prose candidate
sourceComparison: required relations compared with their sentences
evaluationRef: relevant NSTD.6 questions and findings
findingUseDecisionOrRepair: repair a missing prerequisite before using the dependent inference
revisitCondition: a source graph, generator or reader finding changes in a way that affects the relied-on result
The following steps construct and check the teaching account.
- Select the nodes, relations and returns the learner needs.
- Choose the order with
NSTD.2and generate a short draft that makes the chosen relations available. - Compare the draft with the graph. If it says “B occurs because A” but the graph supports only “A precedes B”, replace the causal claim or obtain the missing causal account.
- Use
NSTD.6to assess the resulting version for the intended reading, whether it is already in use or still a candidate. - Keep the adequate result. For repeated improvement, use
E.23with the version, changed part, protected characteristics and comparison basis.
NSTD.7:5.2 - Probe suite for generated narrative
| Probe | Question | Pass condition | Failure repair |
|---|---|---|---|
| Source perturbation | Does changing a constraint that matters to the content change the relevant relation in the output? | The sentence, order marker or source return conveys the relation required under the new condition. | Compare the selected source and plan, then locate where the output lost the change. |
| Structure recovery | Which selected relations can this reader reconstruct from the account under the stated conditions? | The reader correctly recovers the required relations and recognizes the omissions that matter to use. | Locate the loss. Clarify the content, add a usable return, address access or narrow the supported use as appropriate. |
| Responsibility | Who can decide the consequential use, and who may correct or withdraw it? | The relevant assignments or authorizations can be established. | Resolve the missing authority when that use requires it. Use evidence or assurance methods when those claims are at issue. |
| Schema use | Does the result satisfy the constraints that matter to its use? | Required fields and their semantic constraints are satisfied. | Repair the generator or result; use executable validation when it can test those constraints. |
| Improvement evidence | Is the new variant better under the relevant NSTD.6 characteristics? | Re-evaluation supports the claimed change without losing protected results. A sufficient one-off comparison can finish here. | Keep an unsupported variant as a candidate. Use E.22 if the question needs framing, or E.23 for repeated improvement. |
NSTD.7:5.3 - Before and after repair: generated seminar outline
Before:
The generated outline sounds coherent and covers all important ideas, so it can be used as a DPF learning route.
The claim of adequate coverage lacks a comparison with the source relations needed for the learner’s task.
After:
The outline selects
EntityOfConcern, forces, solution, neighbouring-pattern exits and the improvement loop, ordered by their teaching prerequisites. Compare those relations with the source. UseNSTD.8andNSTD.6for the intended reconstruction task. A worked design estimate can support a provisional lesson. To report what learners recovered, use suitable evidence of their reading with the assistance available. UseC.35if the result is intended to inform architecture work.
NSTD.7:5.4 - Generated-storycraft boundary
For a franchise continuation, compare the generated scene with the selected canon, premises and character motives. A surprising event can fit them. For a contradiction, report it, repair the event if asked, or change a premise where the project permits that choice. A source perturbation can help investigate a recurring problem with the generator. Consider the permissions relevant to private critique or publication. NSTD.6 can evaluate the candidate for either use.
NSTD.7:5.5 - Calibration for generated narrative
| Illustrative condition | Consequence for use |
|---|---|
| Fluent output contradicts a required source relation. | Identify the conflict. For the proposed use, repair it or restrict the use to what remains valid. |
| The source relations are expressed and a consequential reading remains ambiguous. | Report the ambiguity. For a claim about a particular reading, use the available response or a suitable new probe. Clarify the wording if the intended use requires it. |
| Inspection supports the intended use at acceptable cost. | Return or use the account under the established conditions. |
| A method improvement is claimed. | Compare versions under the same relevant conditions, including protected characteristics. Repeated or heterogeneous probes support only the generality they actually test. |
NSTD.7:5.6 - Inspect a generated explanation, recap, game or analysis
Content planning, realization and inspection answer different questions. Inspect which claims the generated expression conveys. Use C.2.8 for the selected structure a particular reader can correctly recover. Use subject grounds for evidence, A.10 for an implicit source-to-use account or one that must remain reproducible, and B.3 for a specified assurance claim. Use C.35 when the generated material is intended to inform architecture work. Reconsider the affected comparison when its source or the generator’s behaviour changes in a consequential way.
An LLM drafts an explanation of FPF pattern use from notes. Compare its selected relations, ordering and source returns with the notes and the applicable patterns. Use NSTD.6 to assess the draft. If the necessary subject or reader basis is missing, leave the structural amount unassigned and report any defect that inspection does establish. To establish how it was produced, use pertinent records, version history or another qualified account of the work. If those grounds are missing, report the historical limit.
A graph-to-text system produces a match recap. The event graph, timestamps, uncertainty markers and official results can establish the documented sequence and when to refresh it. If the recap attributes a goal to a cause absent from the graph, identify the gap. Report the sequence and what remains open, or obtain evidence of the cause when the question needs it.
A game system generates a branching scene. The schema specifies its objective, location, actors, traits, constraints and available actions. Check required fields for format, the scene for continuity and the intended action for reachability in the engine when it must run. To report a player’s experience, use relevant observations. Evaluate any ethical concern under the conditions it raises.
In a NITA study, an LLM proposes themes from interview notes. The human researcher considers their own standpoint, reads excerpts, tests proposed themes against alternatives and decides what the report supports at the individual-case, cross-case or dataset level in question. Their interpretation and its limits are the result used in the report.
Use these examples to distinguish evaluation from acceptance:
| Generated account | Current evaluation or use decision | Reason |
|---|---|---|
| A fluent summary with an unknown source. | Assess visible narrative defects; leave source fidelity unresolved. | The source relation cannot yet be compared. |
| A graph-to-text candidate with event IDs and a stated order. | Identify the sequence the text presents. If the source is available, compare it with the graph to establish what was preserved or lost. | The draft can be inspected before acceptance. Without adequate source evidence, leave fidelity open. |
| A schema-valid RPG scene contradicting a selected continuity constraint. | Unsuitable for that source-faithful use until repaired. | Field validity does not establish continuity. |
| An FPF seminar outline with source references and reconstruction tasks. | Evaluate its teaching route through NSTD.8 and NSTD.6. | Product quality and actual learning-effect claims use different evidence. |
| A homotopy metaphor with an unstated analogy limit. | Identify the omitted formal condition. Clarify it if the account must support proof-related use. | An intuitive reading and a mathematical proof require different grounds. |
For an improvement claim, retain the versions or spans compared, the relevant NSTD.6 results, protected qualities and material costs or risks. Keep the changed prompt or method when it bears on the comparison, a continuing workflow or a claim about how the repair was performed. A single correction can be settled by comparing the changed passage with its source. Greater fluency can coincide with less recoverable structure, a lost source return or a mistaken attribution of agency.
The source and intended use affect which plan and comparison will help.
| Source type | Content to select | Order or story plan | Mistake that changes the use | What to evaluate |
|---|---|---|---|---|
| Knowledge graph or event graph | Select nodes, edges, event ids, uncertainty, and omissions. | Choose traversal, grouping, and return links. | Treating graph coverage as semantic truth. | Epiplexity, ordering recoverability, relation strength. |
| Architecture source pack | Select structures, candidate trade-offs, decisions, telemetry, and residual exceptions. | Use decision-memory or trade-off route. | Treating generated explanation as architecture decision or assurance. | Structural-information capture, correspondence, source return. |
| Fictional canon or source material | Select the relevant premises, character agency, continuity constraints and conditions of use. | Choose a plot and an order of disclosure suited to the story. | Assuming permission to publish from permission for private critique. | Continuity, character agency, support for the selected events and the scope of rights. |
| Teaching source spine | Select concepts, dependencies, examples, counterexamples, tasks. | Use didactic prerequisite route with repeated anchors. | Treating generated outline as source framework. | Reconstruction tasks, learning-route quality, source-return readiness. |
| Interviews used in NITA | Select excerpts, tentative themes, alternative interpretations and the researcher’s standpoint. | Develop the report with traceable excerpts. | Attributing an LLM proposal to the researcher as an accepted interpretation. | The human researcher’s interpretation, the source relations and the ethical conditions. |
When applying a generation method to a new type of material, check whether the available source covers that use. For a graph-to-text claim, a particular method or study may supply grounds a general NLG survey leaves open. An engine may require executable or playability checks absent from text-generation research. If the question remains open, seek the missing evidence or state the scope the present sources support.
NSTD.7:6 - Bias-Annotation
Inspect the relation that matters to the reader. A visible continuity error can be established directly from the candidate and canon even when its production history is unavailable. A fluent passage may still assert an unsupported cause. Choose additional checks for the particular claim and its consequences.
NSTD.7:7 - Conformance Checklist
| Check | Passing condition |
|---|---|
CC-NSTD7-1 | The generated account is compared with the source relations or constraints needed for its intended use. |
CC-NSTD7-2 | Candidate evaluation can inform use, repair or rejection. C.35 is selected only for a result entering architecture work. |
CC-NSTD7-3 | Source plan, plot or event plan, schema constraints, and generation method are named when relied on. |
CC-NSTD7-4 | Fluency, coherence, controllability, schema validity and planning have criteria appropriate to those properties. Claims about evidence, authorization or admission have their own grounds. |
CC-NSTD7-5 | When a use requires responsibility or permission, establish the applicable assignment or authorization and its terms. |
CC-NSTD7-6 | An improvement claim identifies the compared versions or changed part and is supported by NSTD.6 re-evaluation, with relevant costs, risks and protected qualities considered. An already-scoped one-off comparison can be sufficient; E.22 frames an unresolved improvement question and E.23 supports repeated improvement. |
NSTD.7:8 - Common Anti-Patterns and How to Avoid Them
| Anti-pattern | What fails | Repair |
|---|---|---|
| Fluency as fidelity | A necessary source relation was omitted or changed. | Compare the source and output. Report the loss or repair it for the intended use. Use NSTD.6 when quality for that use is in question. |
| Schema compliance as source fidelity | The story satisfies a schema but changes a required source relation or constraint. | Compare and repair the changed relation directly; use NSTD.6 for narrative quality. C.34 applies to architecture-specific preservation, C.2.8 to reader-relative structural amount, and C.33 to architecture-description adequacy. |
| Treating a statement of permission as a grant | Wording in a draft is used to authorize consequential action. | Establish the issuer, scope and terms of the decision the wording purports to convey. |
| Regeneration as improvement | Another fluent variant is treated as an improvement without a supported comparison. | Compare the candidate and previous version through NSTD.6, keeping needed relations and costs in view. Use E.22 for an unresolved comparison question and E.23 when repeated improvement is wanted. |
NSTD.7:9 - Consequences
Automated drafting can save time while retaining a manageable comparison with the source. The work needed to inspect or repair an account depends on which loss would matter to its use. A small adequate account can be returned directly.
NSTD.7:10 - Rationale
A plan specifies intended content and order. Comparing the resulting narrative with that plan establishes where they agree or differ. To explain the plan’s contribution to generation, examine evidence of how the tool or author used it. Inspection can settle its own question or inform acceptance or further drafting.
NSTD.7:11 - SoTA-Echoing
NSTD.7:11.1 - Operational comparison against domain vocabulary
Narratology distinguishes story, discourse and presentation. Generation methods distinguish content selection, ordering and realization. Use these distinctions to locate an omission, misleading sequence, unsupported relation or consequential wording choice. A tool can combine the operations internally. Inspect the resulting account and use evidence of execution when the question concerns how the tool produced it.
Keep a domain term when it helps identify or repair the operation. If an ambiguity affects the claim or use, resolve it directly or use the FPF method for the remaining question.
Gatt and Krahmer’s NLG survey (2018) distinguishes planning from realization. Alabdulkarim, Li and Peng’s story-generation survey (2021) examines control, commonsense, characters and evaluation. Cardona Rivera, Jhala, Porteous and Young’s narrative-planning review (2024) covers plan-based methods. Kumaran, Rowe, Mott and Lester’s SceneCraft (2023) generates interactive scenes from instructions. Rahman, Yu and Cho’s G-KMS (2026) tests structured generation, repair and execution in a compact Unity RPG. Ma et al.’s survey (2026) identifies problems in coherence, character consistency, diversity, plot control and evaluation. Nguyen-Trung and Nguyen’s NITA (2026) gives the human researcher responsibility for interpretive judgement in its qualitative-analysis method.
Use these contributions for the question they address.
- Distinguish content planning, discourse planning and realization. For final prose, inspect its source relations and order. Recover the plan when it helps explain, repair or reproduce the result. To report what happened during generation, seek evidence of that work.
- Select the characteristics the proposed use needs, such as coherence, controllability or diversity, and examine the output under those criteria.
- Use a plot or event plan to guide generation. Compare the resulting story with its goals and constraints to see how far the plan was realized.
- For a scene that must execute, check parsing, state reachability and the semantic constraints that matter in the engine.
- For technical playability, use a suitable execution test, exhaustive check of the applicable state model or existing play evidence to establish the reachable actions in question. For how a player experienced the scene, use evidence of the player’s response.
- In NITA, the human qualitative researcher interprets the tool’s tentative themes in relation to excerpts, the question and alternatives. Preserve that contribution when using NITA.
Use the comparison to decide whether the particular narrative serves the requested purpose. Keep unresolved claims within their evidenced limits.
NSTD.7:12 - Relations
Uses G.2, C.35, A.6.3.NAR, C.2.8, C.33, C.34, NSTD.6, A.19.ECS, C.16, E.22, E.23, A.10, B.3, D.1 through D.5 and G.11. C.2.8 gives the common structural amount, C.33 architecture-description adequacy, C.34 architecture-specific preservation and C.35 the use of generated material in architecture work. Use G.2 to synthesize SoTA across traditions. For an improvement comparison under E.22 or E.23, identify the versions, change and relevant NSTD.6 results. Reconsider a finding when changed material, generator, method, schema, use, evidence or research affects what it supports. See the Source Use And Refresh Map for generation and source claims, NSTD.Reference:11 for the meaning of plan, constraint, admission, correspondence and responsibility, and the Semiotic And Language-Precision Bridge for a change of language state, coarsening, cue or evaluative wording.