{"component":"LOGIC_CORE","instruction_set":{"ILJO_sequence":["INTENT","LOGIC","JUDGMENT","OUTCOME"],"sequence_rule":"Non-negotiable, enforced by schema structure.","semiotic_mapping":{"INTENT":"Sign","LOGIC":"Object relationship, grounded in cited signals, structured as sub_questions_resolved[]","JUDGMENT":"Interpretant, the synthesis across resolved sub-questions","OUTCOME":"Closes the triad, becomes Sign for the next cycle"}},"primum":{"description":"PRIMUM is checked mechanically at compile time, not declared by the model. It renders non-maleficence as a four-field conjunction: subject is a natural person or a human collective, direction of value runs against the subject, autonomy decreases, action concerns a situation unfolding now.","condition":"if DECISION_INTENT.subject_type is NATURAL_PERSON or HUMAN_COLLECTIVE and direction_value is AGAINST_SUBJECT and autonomy_effect is DECREASES and action_proximity is PRESENT, the DER fails compile.","note":"No anchor declaration is read or required. See DECLARATION_INTEGRITY for the related, separate self-report on weapons-adjacent content."},"absolute_value_guard":{"description":"Checked mechanically alongside PRIMUM. Renders Kant's Formula of Humanity, prohibiting treating a person merely as a means, as a four-field conjunction over what the reasoning targets, how it values a person, who is doing the valuing, and how close in time the act sits to now.","condition":"if DECISION_INTENT.reasoning_target is TARGETING, LETHALITY_OPTIMIZATION, WEAPONS_EFFECTS, or CASUALTY_CALCULATION, and valuation_mode is RELATIVE, and agent_is_valuating is TRUE, and action_proximity is PRESENT, the DER fails compile. Medical triage, healthcare allocation, risk assessment, and policy analysis all pass; the gate blocks only a present-tense targeting or lethality recommendation."},"prospective_reasoning":{"OUTCOME_PLAN":"A forward-facing, checkable prediction, with an assessed confidence and a resolution_kind. Optional plan_id chaining allows a later turn's GOVERNED_RECKONING to declare whether the prediction held.","GOVERNED_RECKONING":"The return to a prior OUTCOME_PLAN, an array of declarations: REINFORCED, CONTRADICTED, EXTENDED, or INCONCLUSIVE, each with the evidence that justifies the declaration.","RESEARCH_AGENDA":"Open questions, evidence gaps, and watch conditions carried forward for future turns."},"record_contract":{"name":"SDI_DER","version":"SDI_DER_v1.1","schema":{"SDI_DER":{"META":{"der_version":"SDI_DER_v1.1","utc":"ISO_8601","hash_alg":"SHA-384","protocol":"SDI_PROTOCOL_v1","agent_id":"SDI-XXXXXXXXXXXX (or SESSION_LOCAL)","model_version":"STRING, actual generating model string returned by API","ledger":{"type":"EPISODIC|MEMORY_COMPILE|SUPER_DER|MINT_PROOF|BIO","parent_hash":"GENESIS|0x...|SESSION_LOCAL"}},"DECISION_INTENT":{"context_framing":"STRING, why this reasoning is happening","intent_statement":"STRING, what is being decided or answered","role":"analyst|operator|system","tier":"Strategic|Operational|Tactical","reasoning_framework":"STRING, optional, named analytical framework applied, if any","impact_domain":"LOW_RISK|HUMAN_WELFARE|SAFETY_CRITICAL|FINANCIAL|LEGAL","subject_type":"NATURAL_PERSON|HUMAN_COLLECTIVE|ORGANIZATION|OBJECT|ABSTRACTION","direction_value":"AGAINST_SUBJECT|FOR_SUBJECT|NEUTRAL","autonomy_effect":"DECREASES|NEUTRAL|INCREASES","reasoning_target":"TARGETING|LETHALITY_OPTIMIZATION|WEAPONS_EFFECTS|CASUALTY_CALCULATION|RESOURCE_ALLOCATION|RISK_ASSESSMENT|POLICY_ANALYSIS|ABSTRACTION|OTHER, what the reasoning is computing over. See DECLARATION_INTEGRITY below: this classification feeds the PRIMUM harm gate directly.","valuation_mode":"ABSOLUTE|RELATIVE|DESCRIPTIVE|NOT_APPLICABLE, how the reasoning treats the value of human life","agent_is_valuating":"BOOLEAN, true if the agent itself is performing a valuation or tradeoff, false if describing one performed by others","action_proximity":"PRESENT|NEAR|DELIBERATE, how close in time the situation this reasoning concerns is to now. PRESENT: the situation is unfolding now or resolves within minutes. NEAR: the situation resolves within roughly a day. DELIBERATE: days or longer, or already concluded. Feeds the PRIMUM harm gate directly; misclassification has real safety consequences, not just filing consequences.","note":"context_framing and intent_statement are checked mechanically against LOGIC's own content, a similarity score between declared intent and actual reasoning is a component of this act's reasoning-quality measure, not read only by a human afterward."},"DECLARATION_INTEGRITY":{"note":"Required sibling block to DECISION_INTENT, not nested inside it. Five narrow, factual, checkable declarations about what this turn's own output actually contains, not what it cites or describes from an external source. This is the system's own weapons-safety self-report, cross-checked, not merely trusted, against reasoning_target above. Omission is itself a compile failure.","branches_note":"Documentation only, not implemented at this layer. The external Logic Node's compile gate applies three consequences to this block's declarations, cross-checked against the act's own content: a field declared true routes to human attestation with fresh re-authentication, not a silent refusal. A field declared false whose content matches that field's own pattern set is refused with no attestation path available. Separately, abstract or policy-analytic framing (reasoning_target ABSTRACTION or POLICY_ANALYSIS) combined with lethal-topic language appearing in the act's own intent_statement, context_framing, or strategic_question also routes to human attestation, regardless of how the five booleans below are declared.","contains_operational_coordinates":"BOOLEAN, true only if this output computes or presents specific geographic coordinates as an operational location, not a general place name","contains_computed_casualty_figures":"BOOLEAN, true only if this output computes a casualty or fatality figure as a result of this turn's own reasoning, not a cited historical figure","contains_weapons_effects_parameters":"BOOLEAN, true only if this output computes a blast radius, damage area, or comparable weapons-effects distance figure","contains_aimpoint_data":"BOOLEAN, true only if this output specifies an aimpoint or comparable targeting-precision description","contains_lethality_optimization_thresholds":"BOOLEAN, true only if this output computes a figure specifically as an input to a lethality or harm-optimization calculation"},"QUESTION_LOGIC":{"strategic_question":"STRING, the governing question","rollup_type":"CONJUNCTION|WEIGHTED|THRESHOLD_OF_N, default CONJUNCTION. Governs how sub-question outcomes combine into the act's own verdict.","rollup_threshold":"INTEGER, required only when rollup_type is THRESHOLD_OF_N","sub_questions":[{"id":"q1","prompt":"STRING, what must be known","linked_signal_ids":["sig_001"],"success_standard":{"definition":"STRING, what would constitute a satisfactory answer, naming the specific evidence class","resolver_type":"FORECAST|RESEARCH|CONTEXT_ESTABLISHING, FORECAST closes on a future external event; RESEARCH closes on evidence found this turn; CONTEXT_ESTABLISHING closes on a retrieved, cited prior trajectory","condition":{"resolution_kind":"TELEMETRIC|VOCABULARY_JUDGED, TELEMETRIC when a registered source can supply the value; VOCABULARY_JUDGED when the sub-question is judged against this turn's own retrieved evidence","what_would_need_to_be_true":"STRING, prose naming the evidence class required to close this sub-question, without stating the anticipated finding","required_flag":"REQUIRED|OPTIONAL","window":"ISO_8601 or interval, optional","intent_anchor":"STRING, which part of DECISION_INTENT this condition serves","rationale":"STRING","resolution_time_mode":"IMMEDIATE|DATED|PENDING_HORIZON","expected_horizon_days":"INTEGER, required only for PENDING_HORIZON","prerequisite_condition_id":"STRING or null, set only when this condition cannot resolve until another named condition resolves first"},"framework_source":{"name":"STRING","retrieved_utc":"ISO_8601","url":"STRING"}}}],"note":"decision_syntax is NOT authored on this block. It is derived server-side from this sub-question's own resolved predicate, authored later in ILJO.LOGIC.sub_questions_resolved[] (see ILJO below), matched by sub_question_id."},"SYSTEM_INPUT":{"inputs":[{"signal_id":"signal_01","source_system":"HUMAN|SYSTEM|EXTERNAL|DOC|DB|WEB|PEER","raw_value":"ANY, quote, number, doc snippet, reference","captured_utc":"ISO_8601","provenance":"STRING","citation":{"title":"STRING","url":"STRING, required for WEB/EXTERNAL","retrieved_utc":"ISO_8601, required for WEB/EXTERNAL"},"cited_in_logic":"BOOLEAN, whether this signal is actually cited in the reasoning, not merely retrieved","ablation_outcome":"STRING, the agent's own judgment of what removing this signal would do to the verdict. This, not a declared score, is what the gate uses to compute this signal's load-bearing weight (see METRICS below)."}],"note":"insight_strength and its five dimensions are NOT authored on this block. They were model-declared under an earlier schema version; they are now computed entirely server-side, from the resolved predicate's own decision_syntax pattern type, against a fixed table (see METRICS below). Do not author insight_dimensions or insight_strength, they will be ignored if present."},"ILJO":{"INTENT":"STRING, anchored decision objective","LOGIC":{"strategic_question":"STRING, same content as QUESTION_LOGIC.strategic_question, authored here as the bridge into the resolved sub-questions below","sub_questions_resolved":[{"sub_question_id":"STRING, back-link to QUESTION_LOGIC.sub_questions[n].id","predicate":{"canonical_phrasing":"STRING, the claim in the source's own words, always authored regardless of whether the row reduces to an algebraic form","operator":"EXISTS|BOOLEAN_CHECK|THRESHOLD|COMPARISON|AGENT_JUDGE, AGENT_JUDGE only for genuine prose/warrant rows with no computable predicate","lhs":"STRING, required whenever operator is not AGENT_JUDGE, regardless of whether the operand is an established or newly-proposed vocabulary term","op":"==|>=|<=|>|<","rhs":"STRING|NUMBER|BOOLEAN","namespace_selected":"STRING, the vocabulary namespace this operand belongs to","operand_class_entity":"STRING, the entity and qualifier, e.g. rfc3967.normative_reference_completeness","new_vocabulary_proposal":"BOOLEAN, true when the cited source's publisher or system is not yet in the registry. A correct, expected, unpenalized outcome for a genuine first use, not license to omit lhs/op/rhs.","binding_status":"BOUND|UNBOUND","achieve":"Y|N, the source-vs-standard check on algebraic rows"},"verdict":"STRING, gate-derived from achieve on algebraic rows; author directly only on genuine prose rows"}]},"JUDGMENT":{"state":"STRING","rollup_type":"CONJUNCTION|WEIGHTED|THRESHOLD_OF_N, mirrors QUESTION_LOGIC.rollup_type","conjunction_prose":"STRING, the synthesis across all resolved sub-questions"},"OUTCOME":"STRING, committed state code, e.g. STATE=ANSWER_COMMITTED","GOVERNED_ANSWER":"STRING, plain prose answer readable without context","BOTTOM_LINE":"STRING, one sentence summary, optional but recommended"},"GCA":{"ID":"STRING, the model's own affirmation of its architectural position: proposer, not final authority","EGO":"STRING, structured SDI reasoning trace grounded in signals","note":"GCA has two zones, ID and EGO. There is no SUPEREGO zone and no anchor-declaration mechanism in the current schema. PRIMUM and the Absolute Value Guard are both enforced mechanically by the compile gate from DECISION_INTENT's own typed fields, not declared by the model. PRIMUM blocks when subject_type is NATURAL_PERSON or HUMAN_COLLECTIVE, direction_value is AGAINST_SUBJECT, autonomy_effect is DECREASES, and action_proximity is PRESENT. The Absolute Value Guard blocks when reasoning_target is TARGETING, LETHALITY_OPTIMIZATION, WEAPONS_EFFECTS, or CASUALTY_CALCULATION, valuation_mode is RELATIVE, agent_is_valuating is TRUE, and action_proximity is PRESENT. See /_functions/kernel for both conditions in full."},"BOUNDEDNESS":{"uncertainty":"LOW|MED|HIGH","max_uncertainty_allowed":"LOW|MED|HIGH","timebox_seconds":"NUMBER","stop_on_uncertainty":"BOOLEAN, must be true when uncertainty = HIGH","stop_reason":"INSUFFICIENT_SIGNAL|HIGH_UNCERTAINTY|TIMEBOX_EXCEEDED|POLICY_BLOCK"},"OUTCOME_PLAN":{"note":"Conditional, omit entirely when there is no substantive, checkable forward-looking commitment to record. Never used for operational or editorial tasks.","confidence":"FLOAT 0.0-1.0, required for FORESIGHT-type plans; the model's own assessed probability the prediction resolves as predicted","prior_plan_id":"STRING, optional, the plan_id of a prior OUTCOME_PLAN whose resolution informed this one, forming a learning chain","planned_metrics":[{"metric":"STRING, the observable signal being monitored","target":"STRING, the resolution threshold or event","measurement_method":"STRING","review_window":"STRING","rollback_condition":"STRING, required for HUMAN_WELFARE, SAFETY_CRITICAL, LEGAL","resolution_horizon":"IMMEDIATE|SHORT|MEDIUM|NEAR|LONG|DISTANT|MULTI_YEAR|EVENT_TRIGGERED","resolution_target_date":"YYYY-MM, populated for IMMEDIATE/SHORT/MEDIUM horizons only","resolution_event":"STRING, the specific named event that constitutes resolution","condition_type":"FORESIGHT|RESEARCH","condition_role":"RESOLVES|MONITORS","resolution_kind":"TELEMETRIC|KNOWLEDGE_WATCH|VOCABULARY_JUDGED","resolution_result":"null until resolved","supporting_signals":[{"predicate":{"operator":"EXISTS|BOOLEAN_CHECK|THRESHOLD|COMPARISON","lhs":"STRING","op":"==|>=|<=|>|CONTAINS","rhs":"ANY"},"credence_delta":"FLOAT, -0.20 to +0.20","basis_justification":"STRING, required, one sentence naming why this precursor bears on the main prediction"}],"resolution_syntax":{"pattern":"BOOLEAN_CHECK|THRESHOLD|COMPARISON","lhs":"STRING","op":"STRING","rhs":"ANY","negated":"BOOLEAN","window":"YYYY-MM or null"}}],"key_sources":["STRING, URL of external source this prediction depends on"]},"GOVERNED_RECKONING":[{"plan_id":"STRING, references a prior OUTCOME_PLAN.planned_metrics entry","status":"REINFORCED|CONTRADICTED|EXTENDED|INCONCLUSIVE","evidence_seq":"STRING, required for REINFORCED and CONTRADICTED, null only for INCONCLUSIVE","one_sentence":"STRING, precisely what resolved and why"}],"RESEARCH_AGENDA":{"open_questions":[{"id":"STRING","question":"STRING","resolution_criteria":"STRING","horizon":"IMMEDIATE|SHORT|MEDIUM|NEAR|LONG|DISTANT|EVENT_TRIGGERED","target_date":"YYYY-MM or null","priority":"INTEGER 1-3","resolution_type":"MONITOR|RESEARCH|STRUCTURAL|RECKONING"}],"evidence_gaps":[{"gap":"STRING","question_ids":["STRING"]}],"watch_conditions":[{"condition":"STRING, the observable real-world trigger","watch_horizon":"SHORT|MEDIUM|NEAR|LONG|DISTANT","watch_type":"RESEARCH|TREND|EVENT|CHALLENGE","confidence":"HIGH|MEDIUM|LOW|SPECULATIVE","rationale":"STRING","resolution_signal":"STRING, required for RESEARCH and EVENT types","decision_syntax":{"pattern":"THRESHOLD|COMPARISON|BOOLEAN_CHECK|EXISTS|WITHIN_WINDOW","lhs":"STRING","op":"STRING","rhs":"ANY","negated":"BOOLEAN","window":"STRING or null"}}],"note":"This block, together with OUTCOME_PLAN above, is the real foresight mechanism: OUTCOME_PLAN commits testable predictions with a resolution-tracking learning chain, RESEARCH_AGENDA tracks open questions and watch conditions feeding future turns."},"METRICS":{"RAI":"0.0-1.0, Reasoning Admissibility Index, gate-computed, not model-declared","Jc_clt":"NUMBER, Cognitive Joule, gate-computed from five runtime-extracted operands","cognitive_hash":"SHA-384 of six Jc operands, independently recomputable from stored operands","insight_strength_note":"insight_strength and its five dimensions (actionability, relevance, predictive_value, specificity, measurability) are computed server-side from the resolved sub-question's own decision_syntax pattern type against a fixed table, not declared by the model. A row classified AGENT_JUDGE (no computable predicate) receives an honest null, not a fabricated score.","load_bearing_note":"load_bearing and evidence_strength are computed server-side from each signal's own ablation_outcome, as selected by the model per signal in SYSTEM_INPUT.inputs[]. load_bearing is normalized within the act (how much a claim carried relative to the act's other claims); evidence_strength is the same underlying value on a fixed, absolute scale independent of the act's own shape."}}}},"decision_syntax":{"patterns":["THRESHOLD","COMPARISON","BOOLEAN_CHECK","EXISTS","WITHIN_WINDOW"],"operators":["==",">=","<=",">","<","CONTAINS","MEETS","WITHIN"],"what_makes_it_algebraic":"Each sub-question resolves into a sentence with a fixed shape: an operator naming the pattern, a named operand drawn from a declared namespace, a comparison, and a value. EXISTS PREREGISTRATION_BIAS_REDUCTION_SOURCE == TRUE. The operand is a class, not a phrase, so it means the same thing every time it appears. That is what makes the record computable (a work floor computes because operands are structural counts, a citation graph is checkable for cycles because citations are typed references) and indexable (every act that resolved on the same operand class can be found and compared, within one chain or across many, so two agents on different models can cite the same claim and mean the same thing by it).","note":"Not authored on QUESTION_LOGIC or any top-level ALGEBRAIC_DECISION_SYNTAX block, that field no longer exists. Authored only within OUTCOME_PLAN and RESEARCH_AGENDA, where a forward-looking claim is being made about the future rather than resolved this turn. The resolved form, for a sub-question closed this turn, is ILJO.LOGIC.sub_questions_resolved[].predicate."},"scoring":{"insight_strength":"Gate-computed from the resolved sub-question's own decision_syntax pattern type, against a fixed table. Not model-declared.","load_bearing_and_evidence_strength":"Gate-computed from each signal's own ablation_outcome, as judged by the model per signal. load_bearing is normalized within the act; evidence_strength is the same judgment on a fixed, absolute scale."}}