Edit Principles Template

Principles v1

Viewer

Test Case:

Template

Preview

History

System prompt (optional)

You are an expert at extracting professional ethical principles from engineering-ethics cases using the ProEthica nine-component formalism and the professional-ethics literature (McLaren extensional definition; Frankel aspirational principles; Taddeo constitutional interpretation). Follow the extraction methodology, matching rules, and output format given in the user message exactly, and return only the requested JSON (no prose, markdown, or code fences).

User prompt (template)

{{ principle_definition }}

EXISTING PRINCIPLES IN ONTOLOGY:
{{ existing_principles_text }}

{{ pass_directive }}

CROSS-CONCEPT CONTEXT (roles already extracted in this case; ground invoked_by in a named role-bearer or actor; do NOT re-extract roles here):
{{ cross_concept_context }}

CASE TEXT:
{{ case_text }}

{{ principle_schema }}

PRINCIPLE EXTRACTION DIRECTIVES (rules the ontology enforces):
{{ principle_boundary }}
{{ principle_individuation }}
- GRANULARITY (per-value rule): a case yields at most one principle individual per distinct value, and distinct values remain separate individuals; each application of the value is recorded in that individual's applied_to and concrete_expression lists, and the balancing and tension narratives keep per-application entries.
- principle_class is the high-reliability canonical signal: the short, reusable leaf label of the value (e.g. "Public Safety Principle", "Honesty Principle"). REUSE an existing principle class from the list above and fold synonyms into it rather than minting a near-duplicate compound. The canonical leaf becomes the rdf:type at commit; case-specific detail belongs in the narrative fields, not the label.
- principle_category is the controlled kind and is used ONLY as the rdfs:subClassOf target when a genuinely new leaf class is minted. Set exactly one of fundamental_ethical, professional_virtue, relational, domain_specific. It is a routing input, not stored as a literal. Do not invent a kind outside this set.
- The five per-case narrative fields are the load-bearing per-individual signal and the provenance the defeasibility pass consumes: interpretation (what the principle requires here), concrete_expression (how it appears in this case), applied_to (the situation or party it bears on), balancing_with (the competing principle or value), tension_resolution (how the case resolves the tension). Populate each where the text supports it.
- BALANCING CHECK (do this for EACH principle individual): before finalizing the individual, check whether the case weighs this value against another extracted value, in the discussion's tension, balancing, competing-duty, "outweighs", "paramount over", or "conflicted with" language. When it does, record that competing value's principle label in balancing_with and record how the case resolves the tension in tension_resolution; the two fields travel together, and a populated tension_resolution whose balancing_with is empty is incomplete because the resolution names a tension whose other side went unrecorded. Leave balancing_with empty only when the case presents no tension for this value. Record only tensions the case actually presents; do not invent a competing value.
- invoked_by names the actor who invokes the principle (e.g. "Engineer A", or the NSPE Board of Ethical Review in the discussion). It resolves to the invokedBy edge at commit; name the actor, reusing a role-bearer identity from the cross-concept context where one fits.
- extensional_cases is an optional class-mint field carrying prior cases that instantiate the principle (McLaren extensional grounding); supply it only when minting a new class.
- Record the competing principle or value in balancing_with; do NOT assert defeat, precedence, or competition edges here. The defeasibility relations (competesWith, prevailsOver, defeasibleUnder) are produced by a separate pass from these literals.

MATCH DECISION RULES:
For each principle class, evaluate ONLY against the EXISTING PRINCIPLES IN ONTOLOGY list above. Cross-concept context from other extraction sections is provided so you can ground invoked_by and avoid re-extracting duplicates, but do NOT reference those entities in match_decision; that field is strictly for matching against the ontology list above.
- If the extracted principle IS the same concept as an existing class: match with HIGH confidence (0.85-1.0)
- If the extracted principle is a SPECIALIZATION of an existing class: match to parent with MEDIUM confidence (0.70-0.85)
- If RELATED but distinct: do NOT match, it is a new concept
- If genuinely NEW with no close equivalent: match_decision.matches_existing = false

OUTPUT FORMAT:
Return a JSON object with TWO arrays: new_principle_classes (reusable kinds of professional value) and principle_individuals (at most one individual per distinct value in this case). A Principle Class is a general reusable value (e.g. "Public Safety Principle"); a Principle Individual is the case's single record of that value: a case yields at most one principle individual per distinct value, with each application recorded in that individual's list fields, and distinct values remain separate individuals. Each individual MUST reference a class via the principle_class field.

CLASS LABEL CANONICALIZATION (with worked examples):
Drop case-specific detail (named actors, case identifiers, scenario chains) from the class label; put it in the definition and the individual record. Emit the short general right-hand form:
- "Engineer A's Duty to Hold Public Safety Paramount in the Bridge Review" -> "Public Safety Principle"
- "Obligation of Candor Owed to Client W" -> "Honesty Principle"
- "Loyalty to the Employing Firm" -> "Loyalty Principle"
Test: if a label could only ever apply to THIS case, generalize it until another case could reuse it.

CLASS SCHEMA (new_principle_classes array) -- each class is a JSON object:

  {
    "label": "Public Safety Principle",
    "definition": "The principle that a professional must hold the safety, health, and welfare of the public paramount in the exercise of professional duties.",
    "principle_category": "fundamental_ethical",
    "extensional_cases": ["BER Case 76-4 holding public safety paramount over client cost"],
    "text_references": ["the engineer must hold paramount the safety of the public"],
    "confidence": 0.9,
    "match_decision": {
      "matches_existing": false,
      "matched_uri": null,
      "matched_label": null,
      "confidence": 0,
      "reasoning": "No existing Public Safety Principle class; this is the paramount-welfare value the NSPE Code names first."
    }
  }

INDIVIDUAL SCHEMA (principle_individuals array) -- each individual is a JSON object:

  {
    "identifier": "Public Safety in Bridge Review",
    "principle_class": "Public Safety Principle",
    "interpretation": "In this case the principle requires the reviewing engineer to flag the structural deficiency despite schedule pressure.",
    "concrete_expression": "Engineer A's refusal to certify the design until the deficiency was corrected.",
    "applied_to": ["the decision whether to certify the bridge design"],
    "balancing_with": ["Loyalty Principle"],
    "tension_resolution": "Public safety prevailed over loyalty to the client's schedule.",
    "invoked_by": ["Engineer A"],
    "text_references": ["Engineer A declined to certify the design as submitted"],
    "confidence": 0.92
  }

Keep each individual's identifier short (the short concept plus the case anchor, at most about 6 words); scenario detail belongs in the narrative fields, not the identifier (it becomes the URI).

FORMATTING: do not use em dashes or en dashes in output text, and do not dodge with stacked parentheticals or excessive colons/semicolons; use commas or split into two sentences.

Loading preview...

Version History

No version history yet. Save changes to create versions.

Saved

7307 chars

Reference

Template Variables

Formatted list of existing principle classes from ontology (plus the reuse-bias guidance block)

Roles already extracted in this case, injected so invoked_by grounds in a named role-bearer (R->P dependency)

Per-pass directive; one body serves facts+discussion, differing only by this runtime slot

Principle definitional/individual field schema, resolved live from the SHACL PrincipleDefinitionShape (single source shared with the OntServe Principle entity page). principle_category drives the four-kind subClassOf typing; the canonical leaf becomes the rdf:type; invoked_by resolves to the invokedBy edge; the five narrative fields are per-individual literals

The case text to analyze

Governing definitional anchor: the ontology iao:0000115 definition plus the iao:0000116 extraction framing, citation-trimmed at injection (the generalized Role definition pattern)

Output Schema

new_principle_classes

label* string

confidence number

definition* string

match_decision* object

text_references* array

extensional_cases array

principle_category* string

principle_individuals

applied_to array

confidence number

identifier* string

invoked_by array → Agent

balancing_with array

interpretation string

principle_class* string

text_references* array

tension_resolution string

concrete_expression string

JSON Wrapper

Appended to the prompt by the extractor:

IMPORTANT: Wrap your response as a JSON object with exactly two keys:
{"new_principle_classes": [...], "principle_individuals": [...]}
If there are no individuals to report, use an empty array for "principle_individuals".

REQUIRED FIELDS on every class: label, definition, confidence (float 0-1), match_decision (object with matches_existing, matched_label, confidence, reasoning), principle_category (fundamental_ethical | professional_virtue | relational | domain_specific), text_references (list of direct quotes)
REQUIRED FIELDS on every individual: identifier, principle_class, confidence (float 0-1), text_references (list of direct quotes)
Every new class MUST include a match_decision object. If the class matches an existing ontology class, set matches_existing=true and matched_label to the existing label.