Replication package — The Institutional Window: How Contract Law Bounds Liability Signaling of Human Fallback Capability under Generative AI (v1.3.1.1)

The paper maps a posted cap and agreed-damages term into a provider's retained exposure through four legal primitives: litigation viability, the penalty doctrine, displacement to the state or an indemnity pool, and mandatory limits on contractual caps, thereby characterizing the signaling content of a liability commitment. Relative to treating a liability commitment as a direct signal, this models the filtering role of legal institutions explicitly, showing that the same commitment corresponds to different retained exposure across legal environments. This is theoretical modeling work, presented in the original as a model core and an exact message-topology construction, with no empirical estimates reported.

Where the positive image is connected and sub-threshold commitments remain admissible, retained exposure has the message set {0} u [F, C]: low types pool at zero, intermediate types separate on a schedule anchored at F, and high types may pool at the ceiling; where law removes the zero-exposure region, separation begins at the bottom type. It delivers a conditional result on how admissibility of commitments changes the equilibrium message structure, distinguishing institutional settings with and without a zero-exposure region. Derived from the model; the package's acceptance harness re-derives the full grid of five occupations by twelve configurations and checks equilibrium and boundary conditions including the attained-boundary tie break, plus the separation of full, interior, ceiling-pooling and structural-zero regimes.

The package provides occupation and jurisdiction calibration vectors with their legal and empirical provenance, a claim-robustness engine over 88 occupation and 46 jurisdiction perturbations, a Saltelli global sensitivity analysis, figure code, an interactive occupation-by-jurisdiction simulator with a JavaScript/Python cross-check, and SCRIPT_TO_TABLE.md naming the producer of every reported object. It delivers the model, calibration, robustness and sensitivity analyses together with executable code, making each reported object traceable to a specific producer. The acceptance harness code/verify_all.py exits 0 only if every reported quantity reproduces, including robustness coverage reported claim-by-claim as 187 of 200 claim-cells, the thirteen English-rule cells left unsupported because a hole meets the candidate path and the three proved separately, the disclosed tipping values of the prevailing probability, and the sensitivity block against both its base seed and a second seed; the simulator's JavaScript core is an independent implementation and reproduces the package on 65 quantities, including the separating range of all sixty central cells and h_min for the five occupations.

Version 1.3.1 supersedes 1.2.0: the prevailing probability in enforcement participation is now an exposed module parameter (default 0.50, central numbers unchanged), disclosed in Appendix B and stress-tested as its own robustness family with tipping values reported; the sensitivity paragraph is restated with seed bands rather than a point ranking; the Austrian case authority is corrected; and the simulator was brought onto the current construction. It makes a previously implicit prevailing probability explicit and part of robustness testing, and restates sensitivity conclusions as bands rather than a point ranking. Supported by the version notes and by the acceptance harness checks on tipping values and the two-seed sensitivity block.

The package is aimed at readers who want to test or extend the model: it supports re-running equilibrium and boundary conditions on the grid of five occupations by twelve configurations, checking claim robustness with 88 occupation and 46 jurisdiction perturbations, examining sensitivity with the Saltelli analysis, and exploring occupation-by-jurisdiction combinations through the interactive simulator. Parameters are placed ordinally against published evidence rather than estimated, and jurisdiction vectors are stylized benchmarks rather than measured regimes, so results apply to institutional comparison within this model setting rather than to measured prediction for specific jurisdictions.

A careful reader would still watch how far the ordinal parameter placement determines the qualitative conclusions about the message set and separation schedule; how the gap between stylized jurisdiction benchmarks and measured regimes affects tipping values; how stable the tipping values of the prevailing probability are across settings now that it is an exposed module parameter; and how the width of the seed bands affects interpretation of the sensitivity conclusions. In addition, this reading is at summary scope, and Figures 4, 5 and 6 are shipped as PDFs in the package but are not regenerated by it, so figure-level details should be checked against the shipped files.