Ethical Tech CoLab · DUMBO rail-noise programme · Interactive artifact 2 of 3

Provenance-tagged 3D model of two East River rail crossings

A navigable three-dimensional model of the Manhattan and Williamsburg Bridge rail carriage, from whole-crossing context down to individual fastener components — in which every element is colour-coded and outlined by how well it is actually known, and nothing can be displayed without declaring its evidentiary basis. Companion to VISUAL-MODEL-FRAMEWORK.md.

Read this before you look at anything. Not one element of this model is MEASURED or DOCUMENTED in its three-dimensional position. Zero. The geometry you are about to navigate is inferred from photographs, from published narrative descriptions, and from general engineering practice. It is dimensionally schematic. It is not a survey, not a record drawing, and not fit to size a component from. Its purpose is to make the shape of the evidence gap legible, not to represent the bridges accurately. Press Hide INFERRED and Hide ASSUMED together and the frame goes empty. That empty frame is the finding.

Why this looks like a drawing and not a photograph. A photorealistic render of geometry that nobody has measured would communicate a confidence the evidence cannot support — the exact failure mode this programme has already had to correct five times. The renderer here is deliberately schematic: flat shading, visible edges, and a line style that encodes provenance. Solid outline = documented position. Dashed = inferred. Dotted = assumed. You can see the epistemic status of every surface from across the room.

Two fields, not one. Following Document 4 §5.4, each component carries Geometry Provenance (how the shape and position are known) separately from Verification State (how thoroughly the underlying source was read). A component can rest on a source rated 5/5 and still have ASSUMED geometry, because the source describes the element without locating it. That combination is the single most common condition in this model, and collapsing the two fields into one would hide it.

drag orbit · shift-drag pan · wheel zoom · click select

Components at this scale

What the callouts mark

Callouts are anchored to three-dimensional positions and follow the camera. They are placed at the points where the acoustic argument in Document 1 and Document 3 actually turns — not at the points that are visually interesting. Each states what is at stake and what is not known.

The evidence behind each element

What would move an element up the ladder

Provenance of this artifact. Built for the Ethical Tech CoLab DUMBO rail-noise programme. Renderer is hand-written: no external libraries, no network requests, no build step, no telemetry. The choice not to use a general-purpose 3D engine was deliberate — a schematic renderer that cannot produce a convincing photograph cannot accidentally imply a survey. Sources, ratings and the verification vocabulary are defined in VISUAL-MODEL-FRAMEWORK.md. Sibling artifacts: section-problem.html (2D sections) and acoustic-demo.html (measured level demonstration).

Licence. CC BY 4.0, consistent with the repository. Reuse freely with attribution; reuse the numbers only with their caveats attached.