CausaCore is FERZ's patent-pending implementation of the Causal Determinism pathway. The architecture supports independent operation as a deterministic causal modeling framework, and within the FERZ stack supplies causal structure that governance engines bind to, audit, and enforce. The architecture specifies multi-engine causal modeling with a deterministic engine-selection function, generating causal structure across multi-tier and cross-domain systems. Every result carries an explainable causal pathway that is independently traceable from input to causal conclusion.
The architecture produces the same causal structure, and every causal claim is traceable to the evidence and reasoning that produced it.
Pharmaceutical and clinical research where causal pathways span hierarchical tiers (molecular, cellular, tissue, organ, organism) and require deterministic reasoning across the hierarchy.
Financial risk modeling, portfolio analysis, and market forecasting where causal effects propagate across individual, institutional, and systemic levels.
Government policy analysis and crisis response where decisions require modeling causal effects across political, economic, and social domains simultaneously.
Cross-domain research where the causal connections that matter (such as biology to economics, or policy to public health) cannot be analyzed within any single domain in isolation.
Causal Determinism is the architectural mechanism by which CausaCore assures determinism. The assurance principle is deterministic causal modeling across complex multi-tier and cross-domain systems: by design, under the same input data, the same engine selection, and the same parameters, the architecture produces the same causal structure, and every causal claim is traceable to the evidence and reasoning that produced it.
CausaCore is designed to generate causal structure across two dimensions. Within a domain, the architecture specifies modeling across hierarchical tiers (such as molecular, cellular, tissue, organ, and organism in medical research). Across domains, the architecture specifies horizontal causal connections enabling reasoning that spans, for example, biology, economics, and policy in a single analysis. Determinism rests on a deterministic engine-selection function and a validation layer, not on probabilistic scoring.
CausaCore implements four causal modeling primitives. Each is optimized for a different data and reasoning profile. The architecture specifies a deterministic engine-selection function, S(D,R,C), that routes each input to the appropriate engine based on the data's certainty, completeness, and dimensionality characteristics. The selection is determined by the input properties; the same input under the same parameters always selects the same engine.
Engine 1
Causal Nexus Algorithm (CNA)
What it doesDeterministic causal inference across well-documented causal relationships. The architecture propagates causal effects through known structure, producing a causal pathway when the underlying relationships are established.
Optimal forEstablished molecular pathways, documented economic relationships, policy tracking against known causal models.
Engine 2
Bayesian Network Inference (BN)
What it doesProbabilistic causal inference where evidence is incomplete or uncertainty is high. The architecture grounds probabilistic outputs against the deterministic foundation through a validation layer, so probabilistic claims remain replayable.
Optimal forEarly-stage medical research, emerging markets, novel policy situations, scenarios with significant data gaps.
Engine 3
Graph-based Counterfactual Reasoning (GCR)
What it doesCausal pattern recognition in high-dimensional graphs where direct propagation is intractable. The architecture grounds neural pattern outputs against the deterministic foundation, so counterfactual claims remain traceable.
Optimal forHigh-dimensional genomics, large social networks, multifactorial diseases, complex pattern detection.
Engine 4
Agent-Based Simulation (ABS)
What it doesCausal modeling of dynamic system behavior over time, where agent interactions generate emergent causal effects. The architecture grounds simulation outcomes against the deterministic foundation, so emergent causal claims remain reproducible.
Optimal forPolicy implementation modeling, market dynamics, epidemic modeling, and other dynamic system simulations.
CausaCore is designed to operate as a causal modeling layer over heterogeneous data. The architecture specifies a heterogeneous data ingestion layer (multi-LLM orchestration is one implementation approach within it), routes the input through the S(D,R,C) engine-selection function to the engine best suited to the data and reasoning profile, and produces a causal structure along with an explainable causal pathway documenting how the structure was derived. The architecture is model-agnostic and does not depend on any single LLM provider.
CausaCore sits between data input and downstream consumers. The S(D,R,C) selection function routes each input deterministically to one of the four engines, and the architecture produces a causal structure with an explainable causal pathway plus an append-only signed causal record for every result.
CausaCore can operate as a standalone causal modeling architecture. Within the FERZ stack, the architecture is designed to supply causal structure that governance engines bind to, audit, and enforce: LASO(f) and DELIA treat the causal structure as the modeled domain over which their governance rules apply. The architecture supports both modes without rework, because the causal pathway and signed causal record are the same artifacts in both deployment configurations.
CausaCore addresses three categories of risk that arise when AI systems generate causal claims. Each is named below with the concrete failure mode the architecture is designed to block.
Black-box causal reasoning that cannot be audited
AI systems that produce causal claims (X causes Y, intervention A drives outcome B) without exposing the reasoning chain that produced the claim. CausaCore is designed to produce an explainable causal pathway for every result, traceable from input data through the engine selection and validation layer to the causal conclusion.
Domain-siloed analysis that misses cross-domain causal effects
Causal modeling tools that operate within a single domain (clinical, economic, political) and cannot represent causal connections that span domains. The architecture specifies both multi-tier intra-domain modeling (such as molecular through organism in medical research) and cross-domain causal connections (such as biology to economics to policy), so causal analyses are not artificially constrained by domain boundaries.
Probabilistic causal claims without deterministic grounding
Causal claims that depend on stochastic inference and cannot be reproduced under the same inputs. The architecture specifies that every engine output is grounded against a deterministic validation layer; under the same input data, the same engine selection, and the same parameters, the architecture produces the same causal structure. Probabilistic and neural engines remain available where the data calls for them, but their outputs remain replayable.
The architecture specifies an explainable causal pathway for every CausaCore result. The pathway records the canonicalized input, the engine selected by S(D,R,C) along with the rationale for the selection, the causal structure produced, the validation outcomes against the deterministic foundation, and a cryptographic signature. The pathway and the signed causal record are produced together for every result.
The architecture specifies independent replayability: a reviewer with the input data and the model parameters can reproduce the causal structure. This is the definition of audit-grade evidence: not "we have logs" but "a reviewer can re-derive the causal claim."
CausaCore is one of five pathway engines, each implementing a distinct architectural pathway to deterministic governance. CausaCore carries the Causal Determinism pathway.
CausaCore can operate independently as a causal modeling framework. Within the FERZ stack, the architecture is designed to supply the causal structure that governance engines bind to, audit, and enforce. LASO(f) (Semantic Determinism) and DELIA (Constraint Determinism) compose with CausaCore by treating its causal structure as the modeled domain over which their governance rules apply. The three engines share an interoperability model and a common audit record.
CausaCore is protected by a filed utility patent. The architectural framework, the four-engine model, the S(D,R,C) deterministic engine-selection function, the deterministic validation layer, the explainable causal pathway, and the multi-tier and cross-domain scope are specified and protected by FERZ's filings. The engineering implementation is under active development. The capabilities described on this page represent the architectural design; specific deployments are scoped during engagement.
FERZ welcomes three kinds of conversation.
Architecture and licensing
For parties interested in the Causal Determinism pathway itself, including researchers, standards bodies, and prospective licensees of the underlying IP.
Design partner program
For research and analysis teams in healthcare, finance, government, or cross-domain research willing to shape a focused implementation against their causal modeling needs in exchange for early access to a working subset of CausaCore.
Development partnership
For parties interested in accelerating the broader engineering build through funding, technical contribution, or strategic alignment.
How does CausaCore integrate with existing AI systems?
CausaCore is designed to operate as a causal modeling layer over heterogeneous data, with a heterogeneous data ingestion architecture (multi-LLM orchestration is one implementation approach within it). The architecture is model-agnostic and does not depend on any single LLM provider. CausaCore can operate independently or compose with FERZ governance engines (LASO(f), DELIA) that bind to its causal structure for audit and enforcement.
How is CausaCore different from a vector database or knowledge graph?
CausaCore is a causal modeling architecture, not a storage or retrieval system. Vector databases and knowledge graphs represent relationships; CausaCore generates causal structure and verifies it against a deterministic foundation. The architecture specifies four causal modeling engines and a deterministic engine-selection function that routes each input to the engine best suited to its data and reasoning profile, producing an explainable causal pathway as the output artifact.
What makes CausaCore deterministic?
Causal Determinism is the assurance mechanism. By design, under the same input data, the same engine selection, and the same parameters, the architecture produces the same causal structure. Determinism rests on the S(D,R,C) deterministic engine-selection function and a validation layer that grounds all engine outputs against a deterministic foundation, not on probabilistic scoring.
What is the explainable causal pathway CausaCore produces?
The architecture specifies a verifiable record for every result, containing the canonicalized input, the engine selected by S(D,R,C) along with the rationale for the selection, the causal structure produced, the validation outcomes against the deterministic foundation, and a cryptographic signature. The architecture specifies independent replayability: a reviewer with the input data and the model parameters can reproduce the causal structure.
Does CausaCore support multi-tier intra-domain modeling or only cross-domain modeling?
Both. The architecture specifies modeling across hierarchical tiers within a domain (for example, molecular through organism in medical research; individual through systemic in economics) and across domain boundaries (for example, biology to economics to policy). Intra-domain hierarchy and cross-domain interaction are distinct architectural capabilities, and both are part of the Causal Determinism pathway.
How is CausaCore different from LASO(f) and DELIA?
CausaCore implements a different pathway and serves a different function. LASO(f) and DELIA are governance engines: they verify and authorize outputs and actions against codified rules. CausaCore is a causal modeling framework: it generates the causal structure across multi-tier and cross-domain relationships that downstream systems reason against. The three engines can operate independently. Within the FERZ stack, CausaCore is designed to supply the causal structure, and governance engines bind to, audit, and enforce decisions made on top of it.
What domains does CausaCore support?
The architecture is designed for cross-domain operation. Examples include pharmaceutical and clinical research, financial risk modeling, government policy analysis, and cross-domain research where causal connections span multiple domains. The architecture does not embed any single domain ontology; specific deployments configure domain and tier structure during engagement.
The Causal Determinism pathway and the CausaCore architecture are documented in the governance section and the FERZ corpus. See also the five pathway engines and the LASO(f) and DELIA pages for the governance engines that compose with CausaCore.