Machine-Readable Accountability: eXtensible Business Reporting Language, Artificial Intelligence, and the Institutional Rewriting of Accounting Judgement
TL;DR
The study develops Machine-Readable Accountability as a bounded socio-technical construct organised around dynamic visibility, distributed judgement, and responsibility displacement as well as links market-efficiency research with studies of quantification, professional judgement, and digital governance.
Abstract
Machine-readable financial reporting changes how corporate disclosures become visible, verified, and answerable. Inline eXtensible Business Reporting Language (Inline XBRL) combines a human-readable report with embedded structured data, while artificial intelligence expands automated extraction and screening. The study develops Machine-Readable Accountability as a bounded socio-technical construct organised around dynamic visibility, distributed judgement, and responsibility displacement. A qualitative documentary analysis examines four primary archives: regulatory rules, official filing-evidence records, verification materials, and algorithmic-governance documents. Peer-reviewed studies serve as contextual framing and external corroboration. They do not form primary documentary observations. The corpus covers the mature United States Securities and Exchange Commission regime and the European Single Electronic Format from 2020 to July 2026, with earlier sources retained for historical grounding. A seven-family codebook guides analysis of classification, visibility, validation, judgement, audit, artificial intelligence, and accountability. Results arise from coded rules, official filing observations, assurance requirements, and cross-archive role mapping. Structured reporting reduces extraction costs, supports comparison, and permits automated quality checks. It also places institutional weight on taxonomy fit, extension design, validation logic, and software-mediated review. Documentary evidence supports the relocation of judgement across preparers, taxonomy designers, software vendors, auditors, and regulators. Evidence for deliberate narrative optimisation aimed at artificial intelligence remains indirect, so algorithmic answerability remains a bounded theoretical proposition. The framework links market-efficiency research with studies of quantification, professional judgement, and digital governance, and specifies controls for data lineage, extension approval, model documentation, human review, and responsibility assignment.