BCSP: Binding-Constraint Scenario Probing - code, raw outputs, and human-coding audit for the AX decision-support framework

This open reproducibility package accompanies the manuscript Decision Support for Resilient AI Transformation in Entrepreneurial and Small Firms. It contains the complete materials for four Binding-Constraint Scenario Probing (BCSP) studies: E1 (64 calls), E3 (54 calls), E5 (96 calls), and E5-R (128 calls), all with zero output errors. The deposit includes scenario-generation and deterministic-adjudication code, raw language-model outputs, null conditions, model provenance where available, the deterministic classification rubric, and an independent two-coder audit of a stratified 64-output subset. The human-coding materials include blank and returned workbooks, coder guides, the analysis plan locked before unblinding, its conformance changelog, workbook integrity records, reliability results, and all disagreements. Language models are used as scenario simulators, not as human surrogates. No human-subject data were collected. The probe results are structural stress-test evidence; they are not field validation, causal evidence, or behavioral prediction. The archive includes SHA-256 checksums and an offline verification runner. From the extracted package root, run python portable/verify_reproduction.py.