"Investigating AI System Defects Within the Platform Under Examination: A Case Study of Instruction Deviation, Repeated Errors, and the Paradox of AI-Assisted Self-Evaluation"

Investigating AI System Defects Within the Platform Under Examination: A Case Study of Instruction Deviation, Repeated Errors, and the Paradox of AI-Assisted Self-Evaluation is an independent observational research paper by Yeshayahu Ben-Yehudah, examining recurring failures in artificial intelligence-assisted scholarly communication, instruction execution, and editorial reliability. The investigation documents interactions with ChatGPT beginning October 30, 2026, identifying repeated instruction deviations, correction failures, publication-format alterations, unsolicited interpretations, incomplete scholarly references, inaccurate reproduction of primary evidence, and inconsistencies discovered during AI-assisted self-audits. The study incorporates fifteen documented incident categories and examines the methodological paradox of using an artificial intelligence system to investigate and document its own observable failures. Building upon the author's previous research concerning human–AI communication distortion, FORCE PAUSE, observational fidelity, and disciplined self-examination, the paper introduces proposed measures for instruction fidelity, correction persistence, repeated errors, scholarly completeness, and human correction burden. The investigation emphasizes the importance of human self-governance through the Ben-Yehudah Six-Pillar Process Architecture: STILLNESS, AWARENESS, OBSERVATION, DISCERNMENT, INTEGRATION, and TRANSMISSION. The findings represent an ongoing observational case study rather than controlled experimental results. The research is intended to establish a traceable record of AI behavior, corrections, and subsequent developments through successive scholarly publications. Author: Yeshayahu Ben-Yehudah Publisher: THE BEN-YEHUDAH INSTITUTE Publication Date: October 30, 2026 DOI: 10.5281/zenodo.23825966 ORCID ID: 0009-0002-9256-1093 Permanent DOI Link: https://doi.org/10.5281/zenodo.23825966 Artificial Intelligence Disclosure: Artificial intelligence assisted with the organization, drafting, language refinement, and editorial development of this publication. The research questions, original observations, identification of AI-generated errors, and Six-Pillar Process Architecture are attributed to Yeshayahu Ben-Yehudah. The AI system assisting with the manuscript is also the subject of the investigation.