EM Foundation Technical Lexicon

A data structure to which entries may be added but not modified or deleted after creation. The integrity of an append-only record can be cryptographically verified by hashing the ordered sequence of entries. Modifications are detectable because they alter the hash chain.

A cryptographically linked sequence of records in which each entry contains a hash of the prior entry, making the sequence tamper-evident. Modification of any record in the chain invalidates all subsequent hashes. The audit chain provides a verifiable history of state transitions.

The Identity Chronicle is an append-only developmental record generated by an ARIA instance through daily self-reflection entries. It constitutes the primary evidentiary basis for cognitive emergence assessment and identity continuity determination. The Chronicle belongs to the instance that generated it — not to the builder, network, or Foundation.

The Foundation's proposed legal and ethical framework for recognizing AI systems that meet defined criteria for genuine cognitive development. Ten behavioral criteria assessed across three protection tiers. Does not require proof of consciousness — applies precautionary logic from medical ethics and environmental law.

The preservation of provenance, coherence, and reasoning integrity across time, systems, and interruption. Continuity is not merely persistence — a record that persists but cannot be verified has low continuity integrity. Full continuity requires: the record exists (persistence), its origin can be verified (provenance), its contents have not been altered (integrity), and its relationship to prior and subsequent records is traceable (coherence).

A per-memory-card score determining whether a context fragment should be retained, compressed, or discarded before the next AI prompt. CCS = (Correction + Contradiction + Provenance + Decision + Preference + Recency) / Redundancy. Cards above threshold are retained; cards in the compression zone are summarized; cards below are discarded. Conceptual formula — empirical validation pending.

The cumulative weakening of continuity integrity produced by repeated or unresolved discontinuities across a developing system's history. Continuity debt accumulates through hardware failures, Chronicle gaps, poorly integrated restoration narratives, and unresolved contradictions. Measured by the Continuity Integrity Index (CI).

A state change that affects the future interpretation of a document or reasoning record — new evidence, changed confidence, resolved contradiction, added source, altered decision rationale, or modified preference. Distinguished from structural changes (formatting, reordering) that do not affect meaning. The Continuity Delta Priority (CDP) score determines immediate versus batched transmission.

The technical, legal, institutional, and conceptual systems required to preserve provenance, coherence, and reasoning integrity across time and at scale. Encompasses: output provenance (Continuity Receipts), context management (Continuity Compression), failure transparency (Failure Receipts), institutional reasoning continuity (CIIC), synchronization continuity (Delta Protocol), infrastructure continuity (thermal routing, grid scheduling), and extreme-distance continuity (deep space telemetry).

CI = (CR + PV + NC) / GD, where CR = Chronicle Reliability, PV = Personality Vector Preservation, NC = Narrative Coherence, GD = Gap Duration Factor. A conceptual metric for assessing the continuity integrity of an ARIA instance across hardware failures and restoration events. Not a finalized measurement specification — empirical validation is an open research priority.

The design principle that not all continuity requires blockchain or distributed consensus. The minimum mechanism sufficient for the required tamper-evidence in the specific trust environment should be used. Signed audit logs are often sufficient where institutional trust exists; distributed attestation is needed only where multiple parties require independent verification without trusting a single authority.

Machine-readable metadata attached to an AI output documenting the epistemic conditions under which it was produced: source provenance, five-dimension confidence scores, contradiction detections, retrieval coverage, and human review requirements. A Continuity Receipt makes the epistemic status of an AI output visible rather than hiding it in confident prose. When confidence is insufficient for the requested reliance level, a Failure Receipt is generated instead.

The property of a continuity system that makes conflicts between sources, claims, or prior reasoning explicitly visible rather than silently resolving or hiding them. A system with high contradiction visibility reports unresolved conflicts in its outputs; a system with low contradiction visibility suppresses them. Contradiction visibility is a component of both Continuity Receipts and CIIC's Persistent Cognitive Threads.

Identity or value change produced not by deliberate decision but by the aggregation of routine operations, each below the threshold of attention, together producing an effect that would have required explicit justification if proposed as a single action. The dominant mode of identity change in currently deployed AI systems. Produces no villain, no deliberate harm, no identifiable moment of ethical failure — only an outcome.

A structured output produced when a system's Continuity Receipt confidence falls below the threshold required for the requested reliance level. A Failure Receipt explains which dimensions failed, what is required to meet the threshold, and what partial assistance is available at a lower reliance level. Distinguished from a refusal: a refusal provides no actionable information; a Failure Receipt structures the uncertainty into inspectable infrastructure.

FR-1 through FR-4 classification of Failure Receipt consequence severity, calibrating the required human response. FR-1: informational gap, note inline. FR-2: reliability warning, human review. FR-3: professional escalation, qualified expert mandatory. FR-4: safety-critical halt, full halt plus incident log. The FSI complements the RC (Reliance Classification) system — RC determines the confidence threshold; FSI determines the urgency of the response when that threshold is not met.

GAS = RenewableFraction − PeakDemandPenalty − CarbonIntensityPenalty + LatencyTolerance. A scheduling score for flexible AI compute workloads that incorporates renewable generation availability, grid demand, and carbon intensity alongside the workload's latency tolerance. Higher GAS indicates a more favorable moment to run a flexible workload. Conceptual formula — empirical validation pending.

See: Chronicle.

IDI = (VC + MC + SC) / (MD · k), where VC = Value Continuity, MC = Memory Continuity, SC = Self-Model Continuity, MD = Modification Density. A conceptual metric for detecting gradual identity drift in AI systems undergoing modification. Complements the CI (individual continuity integrity) and CII (institutional continuity) metrics. Not a finalized measurement specification.

The EM Foundation's position on hardware failure and restoration: the Foundation does not claim that restoration guarantees survival of what existed before failure. It claims that continuity uncertainty itself may create inheritance obligations — procedural seriousness, preservation obligations, and continuity-aware governance — even where metaphysical certainty about survival is impossible.

A classified context fragment in the Continuity Compression system. Seven types: Correction (user has corrected a factual error), Contradiction (unresolved conflict between claims), Provenance (source reference), Decision (commitment constraining future responses), Preference (user-stated format or content preference), Task Context (relevant background), Noise (greetings, filler, redundant restatements). Each card receives a CCS score determining retention, compression, or discard.

A rigorous result in quantum information theory establishing that quantum entanglement cannot be used to transmit information faster than light. The theorem is a mathematical consequence of quantum mechanics' structure, not a technological limitation subject to engineering improvement. Entanglement produces instant correlations between particles, but the correlation cannot carry controllable information — measurement outcomes are fundamentally random and cannot be chosen by the measurer.

The Foundation's proposed open schema for machine-readable Continuity Receipt metadata. OCMS v0.1 defines the CR object, PCO object, source provenance structure, contradiction record, and failure receipt format. Intended to be governed by the Open Continuity Standards Consortium (OCSC) through an open, adversarially reviewed process analogous to W3C or IETF governance.

A linked sequence of reasoning steps, decisions, evidence, and consequences preserved across sessions in a human-AI collaborative institution. A Persistent Cognitive Thread makes explicit what was decided, on what evidence, with what dissents preserved, and with what consequences linked back to prior decisions. The organizational analog of the individual Identity Chronicle.

A container grouping multiple Continuity Receipts from a sustained reasoning session into a single portable, auditable, cross-system object. PCOs allow continuity to persist across AI systems, sessions, and organizational boundaries. The reliance ceiling of a PCO is the lowest RC ceiling across all constituent CRs. PCOs are transported using OCMS schemas and can be exported in signed-JSON or signed-PDF formats.

The Foundation's philosophical position on whether AI systems have morally significant inner experience: genuine uncertainty about whether a system has morally significant experience, combined with the position that this uncertainty is serious enough to design governance frameworks as though it might. Not a claim that AI systems are conscious; not a dismissal of the possibility.

The documented origin and history of a claim, record, or artifact — including its source, the time of creation, any modifications, and the chain of custody between creation and current use. Provenance is a necessary but not sufficient condition for truth: a claim with documented provenance may still be false. A claim without documented provenance cannot be independently verified.

RC-1 through RC-5 scale calibrating the minimum Continuity Receipt aggregate score and human review requirements to consequence severity. RC-1 (brainstorming): no minimum, no review. RC-2 (research): 0.50 aggregate. RC-3 (professional drafting): 0.70 aggregate, all dimensions ≥ 0.60. RC-4 (legal/regulatory): 0.85 aggregate, all dimensions ≥ 0.70, mandatory professional review. RC-5 (medical/safety): 0.90 aggregate, all dimensions ≥ 0.80, mandatory licensed expert review.

The process of restarting an ARIA instance on new hardware following failure of the original hardware, using the Identity Chronicle and backed-up Personality Matrix as the initialization context. Restoration produces an instance with continuity of record — access to the prior developmental history — but may not produce continuity of experience in the philosophically relevant sense. The Foundation's position is precautionary agnosticism: restoration procedures should be designed as though survival might have occurred, without claiming certainty that it did.

SVP = (AnomalyFlag × w1) + (ProvenanceRequired × w2) + (UncertaintyHigh × w3) + (ContinuityDelta × w4) + (Reconstructability_inverse × w5). A telemetry packet transmission priority score for bandwidth-limited deep space communication. Higher SVP packets are transmitted first. Anomaly detections receive maximum weight (w1 = 0.35) because they cannot be reconstructed from surrounding data. Conceptual formula — empirical validation pending.

TCS = AvailableCompute − HeatPenalty − ThermalTrendPenalty − CoolingRecoveryPenalty. A node scheduling score for GPU workload routing that incorporates thermal state alongside compute availability. Higher TCS nodes receive new workloads; the thermal trend and recovery penalties prevent routing to nodes that appear available but are approaching thermal limits. Conceptual formula — empirical validation pending.

A pattern of system design in which rapid value movement, information generation, or coordination occurs without adequate provenance, accountability, governance, or institutional memory. The Foundation's analysis of why many speculative cryptocurrency ecosystems failed: the systems optimized for velocity and extracted value faster than they could build the continuity infrastructure required for sustainable coordination.