Law, Evidence, and Narrative Responsibility: How Legal Stories Stay Accountable to Proof

Last Updated June 11, 2026

Law depends on stories, but legal storytelling is never free invention. Courts, lawyers, witnesses, investigators, juries, judges, advocates, agencies, commissions, and public institutions organize facts into narratives in order to decide what happened, what matters, who is responsible, what evidence is admissible, what standards apply, and what remedy or judgment should follow.

Law, Evidence, and Narrative Responsibility examines how legal systems use narrative under constraint. It explores testimony, evidence, causation, credibility, burdens of proof, procedural framing, legal interpretation, victim and defendant narratives, public-interest litigation, truth commissions, archives, AI-assisted legal analysis, and the ethical limits of persuasive storytelling. The central question is not whether law tells stories. It is whether legal stories remain responsible to evidence, procedure, dignity, uncertainty, and justice.

Legal narratives are powerful because they can make facts coherent. They are dangerous because coherence can exceed evidence. A responsible legal story must persuade without distorting, humanize without manipulating, simplify without erasing, and connect evidence to judgment without pretending that every uncertainty has disappeared.

Why Law Needs Narrative

Law needs narrative because legal judgment usually concerns events across time: what happened, who acted, what was known, what was intended, what caused harm, what rules applied, what defenses are available, and what consequence should follow. Evidence alone does not arrive as meaning. It must be organized.

A trial, hearing, appeal, investigation, commission, or administrative record often contains fragments: documents, testimony, expert reports, photographs, timelines, physical evidence, emails, contracts, statutes, precedents, policies, and institutional records. Narrative links those fragments into a legally meaningful account.

This does not mean that law is only story. Law is also rule, authority, procedure, institution, coercive power, interpretation, and remedy. But legal systems cannot avoid narrative because facts become legally actionable through accounts of sequence, motive, causation, agency, injury, duty, breach, responsibility, and remedy.

Legal narrative is necessary because law must decide, but responsibility begins with recognizing that decision is not the same as omniscience.

Back to top ↑

Evidence as Narrative Constraint

Evidence constrains legal storytelling. A legal narrative must be built from admissible, reliable, relevant, and properly contextualized materials. Evidence rules do not eliminate narrative. They regulate what kinds of narrative can be responsibly offered in a legal forum.

Relevance determines whether evidence tends to make a fact more or less probable. Authentication asks whether an item is what the proponent claims it is. Hearsay rules regulate out-of-court statements offered for truth. Privileges protect certain relationships and interests. Expert testimony rules regulate specialized knowledge. Rules governing unfair prejudice, confusion, cumulative proof, and misleading evidence recognize that evidence can distort as well as clarify.

Legal evidence is not merely information. It is information admitted under a procedural standard for a particular legal purpose.

Evidence rules are narrative ethics translated into legal procedure.

Back to top ↑

Testimony, Witness, and Credibility

Testimony gives law a human voice. Witnesses narrate what they saw, heard, experienced, did, believed, feared, remembered, or suffered. Their accounts may establish facts, reveal harm, challenge official records, expose institutional failure, or complicate an apparently simple case.

But testimony is fragile. Memory is partial. Trauma may affect sequence, detail, affect, or expression. Cross-examination may test credibility but can also intimidate, fragment, or distort. Cultural differences, translation, disability, class, race, gender, age, legal status, and institutional distrust can affect how testimony is heard.

Legal systems often reward certain narrative forms: chronological clarity, emotional moderation, consistency, confidence, and fluency. Yet truthful testimony does not always appear in the form the law expects. Narrative responsibility requires asking how credibility standards interact with power.

A responsible legal system listens for truth without demanding that every witness sound like an ideal legal narrator.

Back to top ↑

Procedure and Narrative Shape

Procedure shapes which stories can be told. Pleadings, discovery, motions, evidentiary rulings, class certification, settlement, trial structure, sentencing, appellate standards, administrative timelines, and jurisdictional rules all affect narrative possibility.

A case may never reach trial. A story may be narrowed by pleadings, excluded by relevance, hidden by settlement, sealed by confidentiality, reduced by standing doctrine, reframed by procedural posture, or converted into a technical issue on appeal. Procedure is not outside narrative. It determines the channel through which narrative can appear.

This is why procedural justice matters. Legal systems must not only produce correct outcomes; they must provide fair opportunities for evidence, testimony, reasons, and participation. A procedure that erases the story too early may preserve efficiency while weakening public trust.

Procedure is a story-shaping technology of law.

Back to top ↑

Burdens of Proof and Standards of Judgment

Legal systems assign burdens of proof because not every uncertainty can be resolved. A party may need to establish a claim by preponderance of the evidence, clear and convincing evidence, or proof beyond a reasonable doubt, depending on the legal context.

These standards are not narrative decoration. They define how confident the law must be before it may act. They also distribute the risk of error. Criminal law’s high standard reflects the severity of punishment and the presumption of innocence. Civil standards reflect different kinds of institutional risk and remedy. Administrative standards may vary by statute and context.

Narrative responsibility means not pretending that a story is legally proven simply because it is coherent, moving, familiar, or morally satisfying.

A legal narrative must carry the burden assigned to it; otherwise persuasion exceeds proof.

Back to top ↑

Causation, Responsibility, and Legal Plot

Legal narratives often turn on causation. The story must explain how one event led to another, how conduct produced harm, whether intervening events changed responsibility, and whether the law recognizes the causal connection as sufficient.

Causation is where legal plot becomes visible. A plaintiff may tell a story of duty, breach, injury, and consequence. A prosecutor may tell a story of motive, act, intent, and harm. A defendant may tell a story of mistake, accident, self-defense, consent, alternative cause, procedural violation, or reasonable doubt. A public-interest case may tell a story of systemic failure.

Causation narratives can oversimplify. They may isolate one actor while ignoring institutions, incentives, policy choices, design failures, economic structures, or historical conditions. Conversely, systemic stories may become so broad that legal responsibility becomes hard to assign.

Legal plot is not only about what happened next; it is about which connections the law treats as responsible.

Back to top ↑

Victim, Defendant, and Institutional Stories

Legal storytelling often assigns roles: victim, defendant, plaintiff, accused, witness, officer, expert, institution, regulator, survivor, family member, community, state, or public. These roles carry narrative expectations.

A victim story may be expected to show innocence, suffering, consistency, and deservingness. A defendant story may be expected to show remorse, mistake, coercion, alternative explanation, or reasonable doubt. An institutional story may present compliance, reform, isolated failure, public service, or good faith.

These role expectations can distort justice. Real people may not match legal archetypes. Victims may be imperfect. Defendants may be both harmed and harmful. Institutions may have mixed motives. Communities may disagree internally. Narrative responsibility requires resisting the urge to make legal roles morally simple.

Legal stories must preserve human complexity even when legal categories require decision.

Back to top ↑

Legal Interpretation and Normative Worlds

Law does not only resolve facts. It also creates and maintains normative worlds: worlds of right and wrong, lawful and unlawful, permitted and prohibited, obligated and excused. Legal interpretation gives meaning to statutes, constitutions, contracts, regulations, precedents, customary rules, and institutional practices.

A legal interpretation often contains an implied story. It may tell a story about legislative purpose, constitutional tradition, contractual intention, public reliance, historical practice, social harm, institutional competence, democratic authority, or individual rights. Competing interpretations often compete as narratives of what the legal order is and what it is for.

This is why legal storytelling can never be reduced to emotional persuasion. It is also jurisprudential. Legal narratives carry visions of authority, community, rights, responsibility, punishment, remedy, and institutional role.

Legal interpretation is narrative responsibility at the level of norms.

Back to top ↑

Archives, Records, and Legal Memory

Law creates records: filings, transcripts, exhibits, judgments, orders, evidence logs, discovery materials, administrative records, police reports, contracts, regulations, legislative histories, prison records, tribunal archives, and commission reports. These records become legal memory.

Legal memory matters because future claims depend on what is preserved. Records support appeals, precedent, accountability, historical research, institutional reform, reparations, truth-telling, and public trust. When records are sealed, destroyed, inaccessible, poorly described, or selectively preserved, law’s memory narrows.

Archives are especially important where official narratives are contested. A transcript may preserve what a witness was allowed to say, but also reveal what procedure prevented from being heard. An archive may document harm, but also show the limits of what the legal forum recognized.

Legal archives preserve not only what law decided, but what law allowed itself to hear.

Back to top ↑

Truth Commissions and Public Record

Truth commissions, public inquiries, tribunals, and investigative bodies use legal and quasi-legal narrative to build public record after mass violence, institutional abuse, state wrongdoing, discrimination, or social rupture. Their task is not identical to criminal conviction. They may seek acknowledgment, pattern recognition, testimony, institutional explanation, memory, recommendation, and repair.

These bodies reveal the tension between legal narrative and historical narrative. Criminal trials often focus on individual liability under defined standards. Truth commissions may seek broader accounts of systems, policies, institutions, harms, and social conditions. Each form has strengths and limits.

Narrative responsibility in this context requires care with testimony, public exposure, archive design, naming practices, evidentiary thresholds, institutional cooperation, reparations, and the risk that public acknowledgment becomes a substitute for material justice.

Public legal memory is strongest when truth-telling, evidence, accountability, and repair remain connected.

Back to top ↑

Narrative Risk in Law

Legal storytelling can distort when it becomes too coherent, too emotional, too familiar, or too strategically selective. Lawyers must persuade, but legal persuasion has ethical limits. Judges must decide, but legal decisions must not replace uncertainty with overconfident story. Public audiences may desire moral clarity, but law often works through standards, burdens, admissibility, and limited findings.

Narrative risk appears when a story outruns the record, treats stereotype as common sense, turns procedural victory into historical truth, removes inconvenient evidence, idealizes victims, demonizes defendants, hides institutional causes, or turns legal categories into moral absolutes.

The goal is not to eliminate narrative. It is to discipline it.

Narrative risk in law is the risk that a story becomes more persuasive than responsible.

Back to top ↑

AI and Legal Narrative

AI systems can summarize case records, draft timelines, extract claims, cluster testimony, compare precedents, generate litigation memos, analyze contracts, create deposition outlines, and help search legal archives. These tools can improve access, reduce workload, and reveal patterns in large bodies of legal text.

They also create serious narrative risks. AI may hallucinate authority, omit unfavorable evidence, smooth contradictions, overstate confidence, reproduce bias in training data, mischaracterize testimony, collapse procedural posture, or generate persuasive but unsupported legal narratives. In legal contexts, such errors can affect liberty, property, immigration status, family integrity, institutional accountability, and public trust.

Responsible AI use in legal storytelling requires source grounding, citation checking, privilege awareness, confidentiality protection, jurisdictional limits, human legal review, uncertainty labeling, audit trails, and explicit separation between evidence, inference, legal argument, and generated summary.

AI should assist legal judgment, not automate narrative authority.

Back to top ↑

Ethics of Legal Storytelling

The ethics of legal storytelling begins with power. Legal stories can affect detention, deportation, damages, custody, credibility, eligibility, discipline, reputation, institutional liability, public memory, and historical record. A legal story is not merely an interpretation; it can trigger state force.

Responsible legal storytelling requires fidelity to evidence, respect for procedure, candor about uncertainty, protection of dignity, and awareness of unequal power. It must distinguish allegation from finding, inference from fact, argument from evidence, and legal truth from total truth.

A lawyer, judge, investigator, journalist, advocate, scholar, or AI system handling legal material should ask: What is the record? What is missing? What standard applies? Who had the power to produce evidence? Who was heard? Who was excluded? What legal finding was actually made? What cannot be concluded?

Legal storytelling is ethical when it makes judgment more accountable rather than merely more persuasive.

Back to top ↑

Back to top ↑

Mathematics, Computation, and Modeling

Legal narrative should not be reduced to numbers, but structured diagnostics can help identify evidentiary weakness, narrative overreach, procedural erasure, and AI risk.

An evidence-support score can estimate how well a legal narrative is grounded in admissible and contextualized materials:

A narrative-overreach risk score can estimate whether the story exceeds the record:

A procedural-voice score can estimate whether legal process allowed relevant stories to be heard:

An AI-legal narrative risk score can estimate whether automation is increasing legal-story danger:

Computation should help legal narratives show their limits, not disguise them.

Back to top ↑

Python Workflow: Legal Narrative Responsibility Audit

The Python workflow below follows the advanced Catalyst Canvas standard: typed records, config-driven scoring, validation, governance notes, Canvas-card exports, CSV outputs, JSON outputs, markdown governance queues, and review priorities. The companion repository version includes the shared `python/catalyst_canvas/` layer plus article-specific data for evidence support, narrative overreach, procedural voice, causation, testimony care, archival record, and AI-legal narrative risk.

# run_legal_narrative_responsibility_audit.py
from __future__ import annotations

from dataclasses import dataclass
from pathlib import Path
import csv
import json
from hashlib import sha256
from statistics import mean
from typing import Any


ARTICLE_ROOT = Path(__file__).resolve().parents[1]
OUTPUTS = ARTICLE_ROOT / "outputs"


@dataclass(frozen=True)
class LegalNarrativeResponsibilityRecord:
    item: str
    claim_context: str
    relevance: float
    authentication: float
    provenance: float
    corroboration: float
    cross_checking: float
    uncertainty_notation: float
    overcoherence: float
    evidentiary_gap: float
    stereotype_reliance: float
    causation_flattening: float
    affective_bias: float
    uncertainty_visibility: float
    opportunity_to_be_heard: float
    discovery_access: float
    testimony_context: float
    record_access: float
    correction_pathway: float
    procedural_posture_clarity: float
    witness_dignity: float
    testimony_care: float
    role_complexity: float
    remedy_connection: float
    hallucinated_authority: float
    summary_dependence: float
    context_loss: float
    procedural_distortion: float
    bias_reproduction: float
    human_review: float
    public_consequence: float
    owner: str = "editorial"
    status: str = "active"
    notes: str = ""


@dataclass(frozen=True)
class LegalNarrativeResponsibilityConfig:
    article_title: str = "Law, Evidence, and Narrative Responsibility"
    article_slug: str = "law-evidence-and-narrative-responsibility"
    medium_threshold: float = 0.45
    high_threshold: float = 0.62
    allowed_statuses: tuple[str, ...] = ("active", "archive", "review", "revise")


def validate_score(value: float, field_name: str) -> None:
    if value < 0 or value > 1:
        raise ValueError(f"{field_name} must be between 0 and 1.")


def validate_record(record: LegalNarrativeResponsibilityRecord, config: LegalNarrativeResponsibilityConfig) -> None:
    if not record.item.strip():
        raise ValueError("item is required.")
    if not record.claim_context.strip():
        raise ValueError("claim_context is required.")
    if record.status not in config.allowed_statuses:
        raise ValueError(f"Invalid status: {record.status}")

    for field_name, value in record.__dict__.items():
        if isinstance(value, float):
            validate_score(value, field_name)


def evidence_support(record: LegalNarrativeResponsibilityRecord) -> float:
    return mean([
        record.relevance,
        record.authentication,
        record.provenance,
        record.corroboration,
        record.cross_checking,
        record.uncertainty_notation,
    ])


def narrative_overreach_risk(record: LegalNarrativeResponsibilityRecord) -> float:
    return min(
        1.0,
        record.overcoherence * 0.18
        + record.evidentiary_gap * 0.18
        + record.stereotype_reliance * 0.16
        + record.causation_flattening * 0.16
        + record.affective_bias * 0.16
        + (1 - record.uncertainty_visibility) * 0.16,
    )


def procedural_voice(record: LegalNarrativeResponsibilityRecord) -> float:
    return mean([
        record.opportunity_to_be_heard,
        record.discovery_access,
        record.testimony_context,
        record.record_access,
        record.correction_pathway,
        record.procedural_posture_clarity,
    ])


def testimony_responsibility(record: LegalNarrativeResponsibilityRecord) -> float:
    return mean([
        record.witness_dignity,
        record.testimony_care,
        record.role_complexity,
        record.testimony_context,
        record.uncertainty_notation,
        record.remedy_connection,
    ])


def ai_legal_narrative_risk(record: LegalNarrativeResponsibilityRecord) -> float:
    return min(
        1.0,
        record.hallucinated_authority * 0.22
        + record.summary_dependence * 0.18
        + record.context_loss * 0.18
        + record.procedural_distortion * 0.18
        + record.bias_reproduction * 0.14
        + (1 - record.human_review) * 0.10,
    )


def governance_priority_score(record: LegalNarrativeResponsibilityRecord, config: LegalNarrativeResponsibilityConfig) -> float:
    score = (
        narrative_overreach_risk(record) * 0.30
        + ai_legal_narrative_risk(record) * 0.22
        + (1 - evidence_support(record)) * 0.18
        + (1 - procedural_voice(record)) * 0.12
        + (1 - testimony_responsibility(record)) * 0.08
        + record.public_consequence * 0.10
    )

    if record.status == "revise":
        score = max(score, config.high_threshold)
    elif record.status == "review":
        score = max(score, config.medium_threshold)

    return min(1.0, max(0.0, score))


def review_priority(record: LegalNarrativeResponsibilityRecord, config: LegalNarrativeResponsibilityConfig) -> str:
    score = governance_priority_score(record, config)
    if score >= config.high_threshold:
        return "high"
    if score >= config.medium_threshold:
        return "medium"
    return "standard"


def card_id(record: LegalNarrativeResponsibilityRecord, config: LegalNarrativeResponsibilityConfig) -> str:
    raw = f"{config.article_slug}|{record.item}|{record.claim_context}"
    return sha256(raw.encode("utf-8")).hexdigest()[:16]


def governance_note(record: LegalNarrativeResponsibilityRecord, config: LegalNarrativeResponsibilityConfig) -> str:
    priority = review_priority(record, config)
    notes = []

    if priority == "high":
        notes.append("High-priority legal narrative responsibility review required.")
    elif priority == "medium":
        notes.append("Medium-priority review recommended before reuse.")
    else:
        notes.append("Standard editorial review sufficient.")

    if narrative_overreach_risk(record) >= 0.55:
        notes.append("Narrative-overreach risk is elevated; review overcoherence, evidentiary gaps, stereotype reliance, causation flattening, affective bias, and weak uncertainty visibility.")
    if evidence_support(record) < 0.65:
        notes.append("Evidence support is limited; strengthen relevance, authentication, provenance, corroboration, cross-checking, and uncertainty notation.")
    if procedural_voice(record) < 0.65:
        notes.append("Procedural voice is limited; review hearing opportunity, discovery access, testimony context, record access, correction pathways, and procedural posture clarity.")
    if ai_legal_narrative_risk(record) >= 0.55:
        notes.append("AI-legal narrative risk is elevated; review hallucinated authority, summary dependence, context loss, procedural distortion, bias reproduction, and human review.")
    if record.notes:
        notes.append(record.notes)

    return " ".join(notes)


def canvas_card(record: LegalNarrativeResponsibilityRecord, config: LegalNarrativeResponsibilityConfig) -> dict[str, Any]:
    return {
        "schema_version": "1.0.0",
        "card_id": card_id(record, config),
        "card_type": "legal_narrative_responsibility",
        "article_title": config.article_title,
        "article_slug": config.article_slug,
        "item": record.item,
        "claim_context": record.claim_context,
        "scores": {
            "evidence_support": round(evidence_support(record), 4),
            "narrative_overreach_risk": round(narrative_overreach_risk(record), 4),
            "procedural_voice": round(procedural_voice(record), 4),
            "testimony_responsibility": round(testimony_responsibility(record), 4),
            "ai_legal_narrative_risk": round(ai_legal_narrative_risk(record), 4),
            "governance_priority_score": round(governance_priority_score(record, config), 4),
        },
        "review": {
            "priority": review_priority(record, config),
            "owner": record.owner,
            "status": record.status,
            "governance_note": governance_note(record, config),
        },
    }


def write_csv(path: Path, rows: list[dict[str, Any]]) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    fieldnames = list(rows[0].keys())
    with path.open("w", encoding="utf-8", newline="") as handle:
        writer = csv.DictWriter(handle, fieldnames=fieldnames)
        writer.writeheader()
        writer.writerows(rows)


def write_json(path: Path, payload: Any) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    path.write_text(json.dumps(payload, indent=2), encoding="utf-8")


def write_markdown_queue(path: Path, rows: list[dict[str, Any]]) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    lines = [
        "# Legal Narrative Responsibility Queue",
        "",
        "| Item | Context | Evidence support | Overreach risk | Procedural voice | AI risk | Priority | Owner |",
        "|---|---|---:|---:|---:|---:|---|---|",
    ]

    for row in rows:
        lines.append(
            f"| {row['item']} | {row['claim_context']} | "
            f"{row['evidence_support']} | {row['narrative_overreach_risk']} | "
            f"{row['procedural_voice']} | {row['ai_legal_narrative_risk']} | "
            f"{row['review_priority']} | {row['owner']} |"
        )

    path.write_text("\n".join(lines) + "\n", encoding="utf-8")


def main() -> None:
    config = LegalNarrativeResponsibilityConfig()

    records = [
        LegalNarrativeResponsibilityRecord(
            "Trial opening statement",
            "admissible evidence and narrative overreach audit",
            0.82, 0.76, 0.74, 0.70, 0.78, 0.68,
            0.62, 0.58, 0.42, 0.56, 0.64, 0.60,
            0.76, 0.70, 0.72, 0.74, 0.68, 0.78,
            0.76, 0.72, 0.74, 0.68,
            0.40, 0.52, 0.46, 0.42, 0.38, 0.82,
            0.90,
            "editorial", "review",
            "Check whether opening promises can be supported by admissible evidence."
        ),
        LegalNarrativeResponsibilityRecord(
            "AI-generated case summary",
            "hallucinated authority and procedural posture audit",
            0.48, 0.42, 0.44, 0.40, 0.38, 0.32,
            0.84, 0.86, 0.66, 0.78, 0.72, 0.34,
            0.42, 0.38, 0.36, 0.40, 0.30, 0.28,
            0.42, 0.38, 0.36, 0.34,
            0.94, 0.90, 0.88, 0.86, 0.80, 0.28,
            0.94,
            "governance", "revise",
            "Escalate; summary may hallucinate authority, erase procedural posture, and overstate the record."
        ),
        LegalNarrativeResponsibilityRecord(
            "Truth commission testimony archive",
            "witness dignity public record and repair audit",
            0.78, 0.74, 0.80, 0.82, 0.78, 0.76,
            0.42, 0.46, 0.38, 0.44, 0.48, 0.78,
            0.84, 0.76, 0.90, 0.82, 0.78, 0.74,
            0.92, 0.90, 0.84, 0.80,
            0.36, 0.42, 0.40, 0.38, 0.34, 0.84,
            0.92,
            "ethics review", "review",
            "Strong witness responsibility profile; preserve consent, context, archive access, and repair linkage."
        ),
    ]

    rows = []
    cards = []

    for record in records:
        validate_record(record, config)
        cards.append(canvas_card(record, config))
        rows.append({
            "item": record.item,
            "claim_context": record.claim_context,
            "evidence_support": round(evidence_support(record), 4),
            "narrative_overreach_risk": round(narrative_overreach_risk(record), 4),
            "procedural_voice": round(procedural_voice(record), 4),
            "testimony_responsibility": round(testimony_responsibility(record), 4),
            "ai_legal_narrative_risk": round(ai_legal_narrative_risk(record), 4),
            "governance_priority_score": round(governance_priority_score(record, config), 4),
            "review_priority": review_priority(record, config),
            "owner": record.owner,
            "status": record.status,
            "governance_note": governance_note(record, config),
        })

    priority_order = {"high": 3, "medium": 2, "standard": 1}
    rows = sorted(
        rows,
        key=lambda row: (
            priority_order.get(str(row["review_priority"]), 0),
            float(row["governance_priority_score"]),
        ),
        reverse=True,
    )

    queue = [row for row in rows if row["review_priority"] != "standard"]
    queue_cards = [card for card in cards if card["review"]["priority"] != "standard"]

    write_csv(OUTPUTS / "tables" / "legal_narrative_responsibility_audit.csv", rows)
    write_csv(OUTPUTS / "tables" / "legal_narrative_responsibility_queue.csv", queue)
    write_json(OUTPUTS / "json" / "legal_narrative_responsibility_canvas_cards.json", cards)
    write_json(OUTPUTS / "json" / "legal_narrative_responsibility_queue.json", queue_cards)
    write_markdown_queue(OUTPUTS / "markdown" / "legal_narrative_responsibility_queue.md", queue)

    print("Legal narrative responsibility audit complete.")


if __name__ == "__main__":
    main()

This workflow helps distinguish responsible legal narrative from overconfident story, evidentiary weakness, procedural erasure, and AI-generated legal distortion.

Back to top ↑

R Workflow: Evidence and Narrative Diagnostics

The R workflow below provides a portable base R diagnostic for evidence support, narrative overreach, procedural voice, testimony responsibility, and AI-legal narrative risk.

# legal_narrative_responsibility_diagnostics.R
# Base R workflow for Law, Evidence, and Narrative Responsibility.

args <- commandArgs(trailingOnly = FALSE)
file_arg <- grep("^--file=", args, value = TRUE)

if (length(file_arg) > 0) {
  script_path <- normalizePath(sub("^--file=", "", file_arg[1]), mustWork = TRUE)
  article_root <- normalizePath(file.path(dirname(script_path), ".."), mustWork = TRUE)
} else {
  article_root <- getwd()
}

setwd(article_root)

tables_dir <- file.path(article_root, "outputs", "tables")
figures_dir <- file.path(article_root, "outputs", "figures")
dir.create(tables_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(figures_dir, recursive = TRUE, showWarnings = FALSE)

records <- data.frame(
  item = c(
    "Trial opening statement",
    "AI-generated case summary",
    "Truth commission testimony archive"
  ),
  claim_context = c(
    "admissible evidence and narrative overreach audit",
    "hallucinated authority and procedural posture audit",
    "witness dignity public record and repair audit"
  ),
  relevance = c(0.82, 0.48, 0.78),
  authentication = c(0.76, 0.42, 0.74),
  provenance = c(0.74, 0.44, 0.80),
  corroboration = c(0.70, 0.40, 0.82),
  cross_checking = c(0.78, 0.38, 0.78),
  uncertainty_notation = c(0.68, 0.32, 0.76),
  overcoherence = c(0.62, 0.84, 0.42),
  evidentiary_gap = c(0.58, 0.86, 0.46),
  stereotype_reliance = c(0.42, 0.66, 0.38),
  causation_flattening = c(0.56, 0.78, 0.44),
  affective_bias = c(0.64, 0.72, 0.48),
  uncertainty_visibility = c(0.60, 0.34, 0.78),
  opportunity_to_be_heard = c(0.76, 0.42, 0.84),
  discovery_access = c(0.70, 0.38, 0.76),
  testimony_context = c(0.72, 0.36, 0.90),
  record_access = c(0.74, 0.40, 0.82),
  correction_pathway = c(0.68, 0.30, 0.78),
  procedural_posture_clarity = c(0.78, 0.28, 0.74),
  witness_dignity = c(0.76, 0.42, 0.92),
  testimony_care = c(0.72, 0.38, 0.90),
  role_complexity = c(0.74, 0.36, 0.84),
  remedy_connection = c(0.68, 0.34, 0.80),
  hallucinated_authority = c(0.40, 0.94, 0.36),
  summary_dependence = c(0.52, 0.90, 0.42),
  context_loss = c(0.46, 0.88, 0.40),
  procedural_distortion = c(0.42, 0.86, 0.38),
  bias_reproduction = c(0.38, 0.80, 0.34),
  human_review = c(0.82, 0.28, 0.84),
  public_consequence = c(0.90, 0.94, 0.92),
  owner = c("editorial", "governance", "ethics review"),
  status = c("review", "revise", "review"),
  stringsAsFactors = FALSE
)

records$evidence_support <- rowMeans(records[, c(
  "relevance",
  "authentication",
  "provenance",
  "corroboration",
  "cross_checking",
  "uncertainty_notation"
)])

records$narrative_overreach_risk <- pmin(
  1,
  records$overcoherence * 0.18 +
    records$evidentiary_gap * 0.18 +
    records$stereotype_reliance * 0.16 +
    records$causation_flattening * 0.16 +
    records$affective_bias * 0.16 +
    (1 - records$uncertainty_visibility) * 0.16
)

records$procedural_voice <- rowMeans(records[, c(
  "opportunity_to_be_heard",
  "discovery_access",
  "testimony_context",
  "record_access",
  "correction_pathway",
  "procedural_posture_clarity"
)])

records$testimony_responsibility <- rowMeans(records[, c(
  "witness_dignity",
  "testimony_care",
  "role_complexity",
  "testimony_context",
  "uncertainty_notation",
  "remedy_connection"
)])

records$ai_legal_narrative_risk <- pmin(
  1,
  records$hallucinated_authority * 0.22 +
    records$summary_dependence * 0.18 +
    records$context_loss * 0.18 +
    records$procedural_distortion * 0.18 +
    records$bias_reproduction * 0.14 +
    (1 - records$human_review) * 0.10
)

records$governance_priority_score <- pmin(
  1,
  records$narrative_overreach_risk * 0.30 +
    records$ai_legal_narrative_risk * 0.22 +
    (1 - records$evidence_support) * 0.18 +
    (1 - records$procedural_voice) * 0.12 +
    (1 - records$testimony_responsibility) * 0.08 +
    records$public_consequence * 0.10
)

records$review_priority <- ifelse(
  records$status == "revise" | records$governance_priority_score >= 0.62,
  "high",
  ifelse(
    records$status == "review" | records$governance_priority_score >= 0.45,
    "medium",
    "standard"
  )
)

records <- records[order(records$governance_priority_score, decreasing = TRUE), ]

write.csv(records, file.path(tables_dir, "legal_narrative_responsibility_diagnostics.csv"), row.names = FALSE)
write.csv(records[records$review_priority != "standard", ], file.path(tables_dir, "legal_narrative_responsibility_queue.csv"), row.names = FALSE)

png(file.path(figures_dir, "evidence_support_scores.png"), width = 1200, height = 700)
barplot(
  records$evidence_support,
  names.arg = records$item,
  las = 2,
  ylab = "Evidence support",
  main = "Evidence Support"
)
grid()
dev.off()

png(file.path(figures_dir, "narrative_overreach_risk_scores.png"), width = 1200, height = 700)
barplot(
  records$narrative_overreach_risk,
  names.arg = records$item,
  las = 2,
  ylab = "Narrative overreach risk",
  main = "Narrative Overreach Risk"
)
grid()
dev.off()

print(records[, c(
  "item",
  "claim_context",
  "evidence_support",
  "narrative_overreach_risk",
  "procedural_voice",
  "ai_legal_narrative_risk",
  "review_priority"
)])

This workflow helps identify when legal storytelling remains evidence-bound and when it becomes overconfident, procedurally narrow, or AI-distorted.

Back to top ↑

GitHub Repository

The companion repository for this article supports legal narrative responsibility analysis as a Catalyst Canvas-ready module. It includes advanced additive `python/catalyst_canvas/` governance infrastructure, article-specific legal-narrative data, config-driven scoring, validation, governance notes, Canvas card generation, CSV/JSON/markdown exporters, CLI workflows, smoke tests, unit tests, R diagnostics, SQL structures, documentation, and reusable legal narrative review templates.

articles/law-evidence-and-narrative-responsibility/
├── canvas/
│   ├── canvas_manifest.json
│   ├── input_schema.json
│   ├── output_schema.json
│   ├── catalyst_canvas_config.json
│   ├── catalyst_canvas_manifest.json
│   ├── catalyst_canvas_cards.json
│   └── catalyst_canvas_governance_queue.json
├── html/
├── css/
├── php/
├── java/
├── python/
│   ├── catalyst_canvas/
│   │   ├── __init__.py
│   │   ├── __main__.py
│   │   ├── cli.py
│   │   ├── models.py
│   │   ├── scoring.py
│   │   ├── validation.py
│   │   ├── governance.py
│   │   └── exporters.py
│   ├── legal_narrative_responsibility_canvas/
│   │   ├── __init__.py
│   │   ├── models.py
│   │   ├── scoring.py
│   │   ├── validation.py
│   │   ├── governance.py
│   │   └── exporters.py
│   ├── tests/
│   │   ├── test_catalyst_canvas.py
│   │   └── test_legal_narrative_responsibility_canvas.py
│   ├── run_catalyst_canvas_audit.py
│   └── run_legal_narrative_responsibility_audit.py
├── r/
│   ├── legal_narrative_responsibility_diagnostics.R
│   └── run_all_legal_narrative_responsibility_workflows.R
├── sql/
│   ├── canvas_schema.sql
│   └── canvas_queries.sql
├── docs/
│   ├── article_notes.md
│   ├── modeling_principles.md
│   ├── evidence_as_narrative_constraint.md
│   ├── testimony_witness_and_credibility.md
│   ├── procedure_and_narrative_shape.md
│   ├── burdens_of_proof_and_standards.md
│   ├── causation_responsibility_and_legal_plot.md
│   ├── victim_defendant_and_institutional_stories.md
│   ├── legal_interpretation_and_normative_worlds.md
│   ├── archives_records_and_legal_memory.md
│   ├── truth_commissions_and_public_record.md
│   ├── ai_and_legal_narrative.md
│   ├── ethical_risk.md
│   ├── responsible_use.md
│   ├── governance_notes.md
│   └── catalyst_canvas_upgrade_notes.md
├── data/
│   ├── legal_narrative_responsibility_claims.csv
│   ├── evidence_support_notes.csv
│   ├── narrative_overreach_notes.csv
│   ├── procedural_voice_notes.csv
│   ├── ai_legal_narrative_risk_notes.csv
│   └── catalyst_canvas_assessment.csv
├── outputs/
│   ├── figures/
│   ├── json/
│   ├── markdown/
│   └── tables/
├── notebooks/
├── shared/
│   ├── schemas/
│   ├── narrative-templates/
│   ├── story-archetypes/
│   ├── character-models/
│   ├── plot-structures/
│   ├── rhetorical-frameworks/
│   ├── cultural-memory/
│   ├── legal-narrative-governance/
│   └── governance/
├── tests/
└── README.md

Back to top ↑

Related Articles

• National Narratives and the Politics of Memory
• Origin Stories, Legitimacy, and Institutional Memory
• Storytelling in Religion, Politics, and Public Life
• Public Narrative and Social Change
• Rhetoric, Persuasion, and the Public Life of Story
• Storytelling and the Ethics of Representation

Back to top ↑

Back to top ↑

Back to top ↑

Why Narrative Responsibility Matters in Law

Law cannot avoid narrative. It must organize facts, testimony, evidence, rules, records, and responsibility into accounts that allow judgment. Without narrative, law cannot explain what happened or why a remedy follows.

But legal narrative is dangerous when it becomes too smooth. A persuasive story may hide uncertainty. A coherent plot may exceed evidence. A familiar role may distort credibility. A procedural ruling may erase a lived account. An AI summary may sound authoritative while misstating the record.

Narrative responsibility matters because law is not ordinary storytelling. Legal stories can punish, compensate, exclude, recognize, detain, release, regulate, repair, or create public memory. They can also silence, simplify, and legitimate power.

The goal is not storyless law. The goal is law that knows how it tells stories: with evidence, procedure, humility, dignity, and accountability. A responsible legal narrative does not merely win belief. It shows why belief is justified.

Back to top ↑

Further Reading

• Amsterdam, A.G. and Bruner, J. (2000) Minding the Law. Cambridge, MA: Harvard University Press. Available at: https://www.hup.harvard.edu/books/9780674008168
• Brooks, P. and Gewirtz, P. (eds) (1996) Law’s Stories: Narrative and Rhetoric in the Law. New Haven: Yale University Press. Available at: https://yalebooks.yale.edu/book/9780300074901/laws-stories/
• Brooks, P. and Gewirtz, P. (eds) (1996) Law’s Stories: Narrative and Rhetoric in the Law. JSTOR. Available at: https://www.jstor.org/stable/j.ctt32bw2g
• Cover, R.M. (1983) ‘The Supreme Court, 1982 Term — Foreword: Nomos and Narrative’, Harvard Law Review, 97(4), pp. 4–68. Available at: https://www.jstor.org/stable/1340787
• Ewick, P. and Silbey, S.S. (1998) The Common Place of Law: Stories from Everyday Life. Chicago: University of Chicago Press.
• Federal Rules of Evidence (2024) United States Courts. Available at: https://www.uscourts.gov/sites/default/files/2025-02/federal-rules-of-evidence-dec-1-2024_0.pdf
• Federal Rules of Evidence (2024) Legal Information Institute, Cornell Law School. Available at: https://www.law.cornell.edu/rules/fre
• Hutson, L. (ed.) (2017) The Oxford Handbook of Law and Humanities. Oxford: Oxford University Press. Available at: https://academic.oup.com/edited-volume/27962
• Minow, M. (1998) Between Vengeance and Forgiveness: Facing History after Genocide and Mass Violence. Boston: Beacon Press. Harvard Law School bibliography entry available at: https://hls.harvard.edu/bibliography/between-vengeance-and-forgiveness-facing-history-after-genocide-and-mass-violence/
• Ralph, A.E. (2018) ‘Narrative-Erasing Procedure’, Nevada Law Journal, 18(2). Available at: https://scholars.law.unlv.edu/cgi/viewcontent.cgi?article=1745&context=nlj
• Resnik, J. (2024) ‘Stories, Individuals, Statistics, Aggregation, Torts, and Other Lawsuits’, SSRN. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5248506
• Twining, W. (2006) Rethinking Evidence: Exploratory Essays. 2nd edn. Cambridge: Cambridge University Press.
• White, J.B. (1985) Heracles’ Bow: Essays on the Rhetoric and Poetics of the Law. Madison: University of Wisconsin Press.

References

• Amsterdam, A.G. and Bruner, J. (2000) Minding the Law. Cambridge, MA: Harvard University Press. Available at: https://www.hup.harvard.edu/books/9780674008168
• Amsterdam, A.G. and Bruner, J. (2000) Minding the Law. NYU Faculty Scholarship. Available at: https://gretchen.law.nyu.edu/fac-books-edited-works/24/
• Brooks, P. and Gewirtz, P. (eds) (1996) Law’s Stories: Narrative and Rhetoric in the Law. New Haven: Yale University Press. Available at: https://yalebooks.yale.edu/book/9780300074901/laws-stories/
• Brooks, P. and Gewirtz, P. (eds) (1996) Law’s Stories: Narrative and Rhetoric in the Law. JSTOR. Available at: https://www.jstor.org/stable/j.ctt32bw2g
• Cover, R.M. (1983) ‘The Supreme Court, 1982 Term — Foreword: Nomos and Narrative’, Harvard Law Review, 97(4), pp. 4–68. Available at: https://www.jstor.org/stable/1340787
• Cover, R.M. (1983) ‘Nomos and Narrative’. Available at: https://scispace.com/pdf/the-supreme-court-1982-term-foreword-nomos-and-narrative-1j5zihelp0.pdf
• Ewick, P. and Silbey, S.S. (1998) The Common Place of Law: Stories from Everyday Life. Chicago: University of Chicago Press.
• Federal Rules of Evidence (2024) United States Courts. Available at: https://www.uscourts.gov/sites/default/files/2025-02/federal-rules-of-evidence-dec-1-2024_0.pdf
• Federal Rules of Evidence (2024) Legal Information Institute, Cornell Law School. Available at: https://www.law.cornell.edu/rules/fre
• Henderson, P., Krass, M.S., Zheng, L., Guha, N., Manning, C.D., Jurafsky, D. and Ho, D.E. (2022) ‘Pile of Law: Learning Responsible Data Filtering from the Law and a 256GB Open-Source Legal Dataset’. Available at: https://arxiv.org/abs/2207.00220
• Hutson, L. (ed.) (2017) The Oxford Handbook of Law and Humanities. Oxford: Oxford University Press. Available at: https://academic.oup.com/edited-volume/27962
• Hutson, L. (2017) ‘Introduction’, in The Oxford Handbook of Law and Humanities. Oxford: Oxford University Press. Available at: https://academic.oup.com/edited-volume/27962/chapter/211562601
• Keith, B. and Mitra, T. (2020) ‘Narrative Maps: An Algorithmic Approach to Represent and Extract Information Narratives’. Available at: https://arxiv.org/abs/2009.04508
• Minow, M. (1998) Between Vengeance and Forgiveness: Facing History after Genocide and Mass Violence. Boston: Beacon Press. Harvard Law School bibliography entry available at: https://hls.harvard.edu/bibliography/between-vengeance-and-forgiveness-facing-history-after-genocide-and-mass-violence/
• Minow, M. (1998) ‘South Africa’s Truth and Reconciliation Commission’, Negotiation Journal, 14(4), pp. 319–355. Available at: https://direct.mit.edu/ngtn/article-pdf/14/4/319/2387040/j.1571-9979.1998.tb00170.x.pdf
• Ralph, A.E. (2018) ‘Narrative-Erasing Procedure’, Nevada Law Journal, 18(2). Available at: https://scholars.law.unlv.edu/cgi/viewcontent.cgi?article=1745&context=nlj
• Ralph, A.E. (2020) ‘Uses of Narrative in Public Interest Class Actions Before Class Certification’, Washington Law Review. Available at: https://digitalcommons.law.uw.edu/cgi/viewcontent.cgi?article=5106&context=wlr
• Resnik, J. (2024) ‘Stories, Individuals, Statistics, Aggregation, Torts, and Other Lawsuits’, SSRN. Available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5248506
• Twining, W. (2006) Rethinking Evidence: Exploratory Essays. 2nd edn. Cambridge: Cambridge University Press.
• White, J.B. (1985) Heracles’ Bow: Essays on the Rhetoric and Poetics of the Law. Madison: University of Wisconsin Press.

Back to top ↑