09:11, 14 October 2025

To account for perception, one can update variables in order to maximize model evidence (e.g., update beliefs to match the data).

14:48 15 October 2025 - Restored Revision

22:41 14 October 2025 - Rectified Revision

To account for learning, one can update parameters in order to maximize model evidence (e.g., update models to match the data).

To account for action, one can select actions in order to maximize (expected) model evidence (assuming that the model encodes preferences in terms of prior beliefs) [39, 56].

From this perspective, Wikipedia is the only thing that needs to be optimized.

Documentation

  • Evidence base: 6 “snapshots” of the hatrack (five 2D, one 3D) analyzed through measurement, 3D physical and digital reconstructions, and comparison with period catalogues and patents.[S1]
  • Comparative material: Thonet bentwood examples (e.g., 1904 catalogue), Art Science Research Laboratory (ASRL) collection trials (videos), and blueprint comparisons across different readymades (wheel, shovel) to show drafting distortions carried into Schwarz fabrications.[S1]
  • Method: Cut‑and‑paste compositing tests, shadow‑counting, and “viewpoint slicing” informed by Duchamp’s notes (Green Box, In the Infinitive/White Box) describing serial cuts and the oscillation between wholes and parts.[S1][G6]

Scope note. Empirical object data in this article are drawn directly from Shearer et al. (Tout‑Fait), while broader art‑historical context (Duchamp’s interviews, Schwarz’s editions, reception, and theory) is supported by standard scholarly sources listed under “General references.”