Human Oversight Dashboard
A locally validated synthetic prototype that helps a human reviewer inspect evidence, risk, priority, uncertainty, and proposed communication before recording a decision.
The problem
A draft is not a decision.
AI-prepared workflow language can omit context, overstate certainty, confuse recipients, or suggest actions that require human authority. Reviewers need the evidence and risk beside the proposed wording—not hidden in a separate log.
What Christian needed to solve
Create a visible review gate where a manager can inspect the scenario, edit exact wording, reject it, or stop and escalate before any hypothetical next step.
What Christian designed
A review environment with explicit boundaries.
The prototype organizes fictional scenarios by evidence, risk, priority, confidence, uncertainty, and missing context. It supports editable drafts, deterministic safer rewrites, exact-text approval, rejection, stop and escalation, local history, and public-safe synthetic export.
Approval does not mean sending
Approval records a local in-memory review state only. Nothing is sent, no live system is updated, and blocked or escalation-required scenarios cannot be approved.
Contribution
Christian’s role
Christian defined the domain problem, human-oversight objective, required review decisions, safety constraints, prototype boundaries, and public-evidence standard. He directed and reviewed AI-assisted implementation and validation under those constraints.
AI and tools
Implementation support
AI-assisted software development supported application implementation, test drafting, validation, and documentation. The rewrite behavior inside the prototype itself is deterministic—not a live model.
Evidence and validation
Validated locally across behavior and viewports.
These are engineering and accessibility checks, not business results or production-readiness claims.
Formatting, ESLint with zero warnings, TypeScript, production build, keyboard review, automated 200% page-scale review, reduced-motion review, and console/local-only network review also passed.
Technologies, by purpose
- Application: Next.js, React, TypeScript, CSS/Tailwind build tooling
- Testing: Vitest, Testing Library, Playwright browser acceptance
- Quality: ESLint, Prettier, type checking, production build
Known limitations
In-memory state, deterministic fixtures and rewrites, local-only export, no authentication, no persistence, no integration, no production control plane, no formal screen-reader certification, and open dependency advisories. This is not production-ready.
What Christian learned
Human oversight becomes more useful when evidence, uncertainty, exact wording, prohibited actions, and the decision record are presented as one workflow.
What could improve next
In a synthetic test environment, evaluate authenticated roles, durable policy-governed review records, and formal assistive-technology testing—without weakening no-send boundaries.