TACITUS has one Conflict Ontology — 8 primitives that structure any dispute, from a workplace grievance to an armed conflict. The grammar is universal. The instantiation is domain-specific. This is what makes it a platform.
The same 8 primitives describe both a workplace grievance and a civil war. The grammar is shared. The semantics are domain-specific.
Interpersonal · Organizational · Diplomatic · Commercial · Governance
| Actor | Employee, Manager, HR Mediator, Union Rep, Client |
| Leverage | Reputation, tenure, social proof, information asymmetry |
| Constraint | HR policy, employment law, contract terms, cultural norms |
| Commitment | Settlement terms, action items, performance plans |
| Narrative | Fairness framing, blame attribution, organizational justice |
Interstate · Intrastate · Hybrid · Asymmetric · Grey Zone
| Actor | State, Armed Group, Proxy Force, IGO, Guarantor Power, Spoiler |
| Leverage | Territory, military force, sanctions, energy supply, UNSC veto |
| Constraint | IHL, UNSC mandates, geography, domestic opinion, election cycles |
| Commitment | Ceasefires, peace treaties, DDR programs, peacekeeping mandates |
| Narrative | Sovereignty claims, self-determination, atrocity framing |
The ontology is the moat. “Leverage” in a workplace dispute is reputation and social capital. In an armed conflict it’s territory, sanctions, and veto power. But both are typed LEVERAGE edges in the same knowledge graph — queryable by the same Cypher, scorable by the same KGE model, structurable by the same API. Every conflict ingested — from either domain — enriches the shared embedding space. This is the compounding data asset.
TACITUS is building in public. Here’s what exists today and what we’re building toward. #BuildingInPublic
Structure compounds. Every conflict ingested makes the next one easier to resolve.
The Conflict Ontology makes dispute data computable. 8 primitives, typed relations, temporal ordering, provenance. The ontology is the moat — it cannot be learned from data alone.
LLMs extract entities from unstructured sources and map them to the ontology. Ontology-Augmented Generation means the graph constrains the model — eliminating hallucination at the architectural level.
Knowledge Graph Embeddings learn cross-case patterns across both domains. Each conflict ingested enriches the shared embedding space. The compounding data asset.
TACITUS sits at the intersection of neurosymbolic AI, knowledge graph science, and computational conflict analysis.
Combining neural learning with symbolic reasoning for provenance-critical domains.
Kautz (2022), Garcez et al. (2019)
Translational and rotational approaches to encoding relational knowledge.
Bordes et al. (2013), Sun et al. (2019)
Interest-based negotiation, structural violence, peacebuilding frameworks.
Fisher & Ury (1981), Galtung (1969), Lederach (1997)
Event coding and early warning systems.
PLOVER, ACLED, UCDP
Message passing on typed multi-relational graphs.
Schlichtkrull et al. (2018)
Formal knowledge representation with OWL 2 DL and SHACL validation.
Guarino et al. (2009)
Workplace disputes, legal cases, commercial negotiations, institutional governance.
Peace process monitoring, early warning, document intelligence, conflict coding.