Comprehensive Content Analysis Dataset of the Iranian Junior High School "Social Studies" Textbook(Grade 8)(Lesson 1-20): A Multi-Agent Adjudicated Protocol (CAP) Based on the William Romey Method

This dataset delivers a comprehensive, multi-modal, and peer-adjudicated content analysis of Lessons 1 to 20 of the Iranian National Middle School textbook, "Social Studies" (Grade 8), employing an enhanced adaptation of the William Romey Method. Developed as an open-science, empirical repository initiative at Farhangian University (Nasiabeh Campus, Tehran), this project aims to strengthen teacher agency by scaffolding pre-service teachers into the professional roles of Intelligent Supervisors. The collaborative repository is systematically divided into five specialized student-teacher research cohorts: Lessons 1-4 were rigorously audited by Farzaneh Mohammadi, Farzaneh Motamed, Nasim Azimi, and Hananeh Janparvar; Lessons 5-8 were audited by Fatemeh Torkashvand, Fatemeh Sedighi, Mahsa Ahmadifar, and Fatemeh Shahsavand; Lessons 9-12 were systematically evaluated by Soudeh Mombaini, Mobina Radmard, Mohaddeseh Kaleji, Mahdieh Rahimian Amiri, and Negar Ashrafi; Lessons 13-16 were analyzed by Hadiseh Rezavand, Mobina Sorkheil, Mina Jafari, and Fatemeh Moniri; Lessons 17-20 were evaluated by Elaheh Amandadi, Razieh Amiri, Zahra Sarkhanli, and Fatemeh Hojat. The diagnostic matrix conducts a deep-dive evaluation across critical curricular components: textual narratives, visual illustrations, integrated classroom activities, and sample evaluation questions. This strategic boundary intentionally captures the foundational socio-political, civil, and communicative pillars of the curriculum (such as civic cooperation, the judicial system, and digital literacy). To eliminate individual subjectivity and mitigate AI-generated systemic biases, the data generation pipeline utilized the 4-stage Collaborative Adjudication Protocol (CAP): (1) Methodological training on active/passive content dynamics; (2) Baseline AI-assisted auditing via structurally validated prompt sequences; (3) Double-blind peer human-AI cross-auditing involving four critical operational roles (First Auditor, Second Auditor, Prompt Engineer, and Documentation Manager); and (4) Final CAP reconciliation and multi-agent text/image adjudication. This benchmark corpus stands as Volume 8 of the Iranian School Textbook Auditing Repository.