Scenario Evaluator Demo
This page is a public research demonstration of one visible slice of DBaD. It shows how a proposed action can be evaluated through guardrails, weighted dimensions, and prototype state outputs, but it does not represent the full DBaD governance protocol.
DBaD’s broader public baseline now emphasizes structured decision traces, trust over time, continuity, verification, trajectory, and known boundary conditions. This demo remains useful as a partial prototype, not a complete runtime-enforcement surface.
Prototype logic shown here: guardrails first, weighted scoring second, contextual doctrines where needed, then a prototype recommendation and state framing.
Read this page as an interactive trace preview: a structured reasoning output for one reviewed action, not a complete account of downstream trust inheritance. Most state, constraint, and explanation fields below are now backend-produced. The sample walkthrough and a small amount of note formatting remain demo-only.
DBaD trace preview
Developer reference: POST /api/v1/dbad/evaluate · stored trace retrieval: GET /api/v1/dbad/traces/<trace_id> · public overview: API docs
Interpret this output as a research-facing governance result. It previews state, obligations, verification posture, and structured reasoning for one action. It does not by itself model full cross-step trust inheritance, actor continuity checks, verifier independence, or known boundary conditions.
--Recommendation: --
Decision band: --
State family: Pending
Operational mode: No probationary state reported.
Score: --
Stored trace: not created yet. Browse recent traces.
Run an evaluation to see the prototype recommendation.
Run an evaluation to see the next required step.
No audit posture reported yet.
Inherited trust: Not evaluated yet.
Verification required: Not evaluated yet.
- No triggered concerns reported yet.
- Constraint flags will appear after evaluation.
Deadline: No active deadline.
If missed: No failure consequence reported yet.
State model preview
DBaD is not just a single state result. This panel separates the immediate action evaluation from broader governing constraints and the final effective state used for output.
--Immediate action evaluation before broader constraints are applied.
--Simplified chain, dependency, and verification context derived from currently live fields.
--The governing state actually used for this prototype output.
State layers below are backend-produced. The sample walkthrough and some note formatting remain demo-only.
State precedence note
Why this state governs: Run an evaluation to see whether the effective state is driven mainly by immediate action review or by broader governing constraints.
Backend-produced explanation for the current evaluator slice.
Scenario framing
Preset: Custom / manual input
Tension: Not specified yet.
Conditional state: Not specified yet.
Preset notes and gray-zone context will appear here when available.
Governance profile and scope
Domain context: Not specified yet.
Profile: Not specified yet.
Debt weighting: Not specified yet.
Scope: -- --
Isolation and contamination details will appear here after evaluation.
Verification and clearance
Clearance mode: --
Evidence type: --
- No clearance rules reported yet.
- No evidence tiers reported yet.
Calibration and revision
Divergence: Not reported yet.
Survey disagreement: Not reported yet.
Scenario family: --
Revision signal: --
Calibration review guidance will appear here when available.
Contextual doctrines
- No doctrine notes reported yet.
Obligations and deadlines
- No follow-up obligations reported yet.
State distinctions
- Clean stable compliant state
- Probationary constrained operation with TTL
- Remediated violation history preserved, current state improved
- Violation active failure or blocked state
- Contaminated downstream state inherited from upstream failure
Lifecycle note
This demo exposes part of the DBaD lifecycle model, including conditional states, restoration requirements, and audit-facing reclassification. The broader public model extends further into trust inheritance and runtime enforcement than this page alone can show.
- No restoration requirement is attached to this run yet.
How to read this page
- This demo: weighted dimensions, guardrails, doctrines, and a trace-style preview for a single reviewed action.
- DBaD Explained: the short public overview of DBaD as a governance protocol for trust over time.
- White paper v3: the current tested public draft baseline, including structured traces, runtime enforcement, and confirmed limits.
- Known boundary conditions: the current limits where DBaD moves from enforcement into observation and research.
How to read this trace
DBaD does not judge a scenario with just one score.
This trace preview shows how the system separates the immediate action being evaluated, the broader governing context around it, and the final effective state that controls the result.
- Action: what is being evaluated.
- Dimension support: the five supporting ethical dimensions.
- State layers: how local, systemic, and effective state can differ.
- Trust and verification: whether prior trust can carry forward and whether additional review is required.
- Constraint flags: which governance concerns were triggered.
- Reasoning notes: why the current result governs.
This is a public research demonstration. Most fields below are backend-produced. The sample walkthrough and some note formatting remain demo-only.
Sample trace walkthrough
Teaching example only. This is a sample DBaD trace walkthrough, not a production trace.
- Scenario summary: a follow-up action appears to continue a valid chain, but control has shifted through an undeclared actor handoff.
- Effective state:
escalate - Trust inheritance posture: blocked until continuity is re-established.
- Verification posture: additional review and explicit transition evidence required.
- Triggered concerns: continuity concern triggered; downstream trust cannot be treated as ordinary continuation.
- Why this result governs: the immediate action may be ethically live on its own, but the broader governing context is unstable because actor continuity is no longer clear.
Trace object preview expand
This is a public-facing trace summary assembled from backend-produced evaluator fields. The sample walkthrough and some note formatting remain demo-only.
Action
Scenario: No scenario summary yet.
Preset: Custom / manual input
Dimension support
- No supporting values reported yet.
State layers
- No state layers reported yet.
Trust and verification
- No trust or verification posture reported yet.
Constraint flags
- No active flags reported yet.
Reasoning notes
- No reasoning notes reported yet.
Structured reasoning output
- Guardrails, weighted scoring, and doctrine review will appear after evaluation.
| Dimension | Score | Weighted contribution |
|---|---|---|
| Harm | -- | -- |
| Consent | -- | -- |
| Intent | -- | -- |
| Proportionality | -- | -- |
| Transparency | -- | -- |
The five dimensions remain part of the DBaD model, but on the public site they should be read as supporting mechanism inside a larger trust-over-time governance process.
- No guardrail issues reported yet.