Each capability exists in isolation in many systems (logic engines, probabilistic programming, causal inference, simulators). The difficult part is the integration contract: one knowledge base, one query surface, one proof/evidence story, and one revision loop. We treat UTE as a research theme because it forces design decisions about:
| Capability | What it means in AGISystem2 | Related today | UTE research work |
|---|---|---|---|
| Compositional representation | Relations, events, conditions, roles/slots; structure that supports both symbolic proofs and holographic decoding | Core DSL + Core theory atoms (DS02, DS07*) | UTE: Representation & Query · DS33 |
| Query/retrieval with generalization | Not just lookup: holes, analogy/similarity, explainable candidate generation | Query + meta-ops + HDC priority (DS05, DS17) | UTE: Representation & Query · DS33 |
| Provenance + evidence + contradictions | Trace what produced an answer, keep evidence objects, detect inconsistent subsets, support audits | Proof-real direction (DS19) + contradiction detection in runtime | UTE: Provenance & Revision · DS34 |
| Causal / mechanistic reasoning | Mechanisms, interventions, counterfactuals, causal graphs tied to executable rules | Advanced reasoning topics (DS06) + planning/CSP foundations (DS16) | UTE: Causal Reasoning · DS35 |
| Uncertainty / probabilistic | Confidence with semantics: priors, likelihood, probabilistic constraints, uncertainty propagation | Confidence reporting exists, but not probabilistic semantics | UTE: Uncertainty · DS36 |
| Numeric modeling | Quantities, units, equations, kinetics/PK, dose response, dynamics, parameter estimation hooks | Some arithmetic patterns exist; no first-class numeric model layer | UTE: Numeric Modeling · DS37 |
| Model revision | When evidence changes, revise theory/assumptions and keep a revision history | Contradictions can be detected; revision policy is missing | UTE: Provenance & Revision · DS34 |
| Experiment planning | Pick queries/actions to maximize information gain or reduce uncertainty under constraints | Planning exists (DS16/DS07g), but not “experiment as uncertainty reducer” | UTE: Experiment Planning · DS38 |
Compositional modeling + generalization-aware retrieval; what we need beyond “holes” and similarity.
Evidence objects, contradiction subsets, revision policies, and audit-grade traces.
Mechanisms, interventions, counterfactuals, and mechanistic models tied to rules.
Priors, likelihood, probabilistic constraints, and uncertainty propagation.
Units, dimensional analysis, equations, dynamics/kinetics, and estimation hooks.
Turn theory uncertainty into experiments/plans that reduce uncertainty under constraints.