Artifact: Verifiability Diagram

Worked Scenario 21 -- verifiability diagram -- exercises co-10, co-09.

This diagram ranks the output structures seen across Theme B by verification cost -- how long it takes a user to independently check a claim, from cheapest (top) to most expensive (bottom).

%% Color Palette: Blue #0173B2, Orange #DE8F05, Teal #029E73, Purple #CC78BC
%% Ranked top-to-bottom: cheapest verification cost first
graph TD
    A["Structured, per-field cite<br/>~5s per field -- ex-20"]:::teal
    B["Free text, clause link<br/>~10s -- ex-18"]:::blue
    C["Free text, doc-only cite<br/>several min -- ex-19"]:::orange
    D["Free text, no citation<br/>unverifiable, no check"]:::purple
 
    A -->|cheaper| B -->|cheaper| C -->|cheaper| D
 
    classDef teal fill:#029E73,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef blue fill:#0173B2,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef orange fill:#DE8F05,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef purple fill:#CC78BC,stroke:#000000,color:#FFFFFF,stroke-width:2px

Figure: four output structures, ordered by measured or estimated verification cost. The arrows read "cheaper than" -- each rung down the ladder costs a user meaningfully more time and effort to independently check, for no gain in the underlying answer's accuracy.

Verify: all four rungs appear in strictly increasing verification-cost order, and each rung cites the specific worked scenario that produced or diagnosed it -- satisfying co-10 and co-09's rule that verification cost is a rankable, not merely qualitative, property of an output's shape.

Key takeaway: Two answers with identical underlying accuracy can have wildly different real-world usability, entirely as a function of how their output is structured -- verification cost is a design variable, not a fixed cost of using a probabilistic feature.

Why It Matters: This ladder is the concrete argument for treating co-10 as the single highest-leverage pattern in this theme: moving one rung up costs a product team a one-time engineering investment in output structure and citation precision, and pays back on every single answer the feature produces, forever. A team auditing its own feature can place its current output format on this same ladder to see, concretely, how much investment separates it from the cheapest achievable rung -- turning an abstract "make citations better" goal into a specific, measurable rung to climb toward.

---

← Back to Theme B: Uncertainty, Confidence, and Provenance