Reproducibility package for: Beyond AI Adoption: Functional Use Profiles and Institutional Conditions across Six Ibero-American Education Systems

This is the reproducibility package for the article "Beyond AI Adoption: Functional Use Profiles and Institutional Conditions across Six Ibero-American Education Systems". WHAT THE STUDY DOES The study is a secondary analysis of the OECD Teaching and Learning International Survey (TALIS) 2024, which for the first time asked lower-secondary teachers directly about their use of artificial intelligence in teaching. It covers 7,995 teachers who received the questionnaire form carrying that module in six education systems: Brazil, Chile, Colombia, Costa Rica, Portugal and Spain. Between 30 and 56 per cent of teachers report having used artificial intelligence in teaching in the previous twelve months, with the four Latin American systems above and the two Iberian systems below the six-system average. Among those who report using it, a latent class analysis of eight functional-use indicators identifies four profiles: narrow preparation use (45 per cent of adopters), learning-design use (23 per cent), assessment and monitoring use (14 per cent) and comprehensive integration (18 per cent). Weighted regressions with education-system fixed effects then ask which teacher and school conditions are associated with adoption, with the breadth of use and with profile membership. Professional learning on artificial intelligence is the condition most strongly associated with entering use at all (average marginal effect +40.5 percentage points, 95% CI 35.9 to 45.0), while collaboration in lessons is the condition most strongly associated with breadth of use (+0.95 of eight uses, 0.60 to 1.30) and with the comprehensive profile (+17.3 percentage points, 10.6 to 24.0), an association carried by Brazil and Costa Rica rather than shared across all six systems. All variances are estimated by balanced repeated replication with Fay's adjustment on 100 replicate weights. WHAT THIS DEPOSIT CONTAINS Twenty analysis scripts (twelve pipeline scripts driven by run_all.sh, eight supplementary), the 57 derived tables reported in the article and its supplement, the five figures in 600-dpi raster and vector form, a results digest that maps every number quoted in the article to the table it was read from, the environment manifest with all seven random seeds, the variable inventory and reserved-code recoding log, the questionnaire-form audit, an output-to-script map, SHA-256 checksums of every file, and a reproduction report documenting a clean re-run in which 17 of the 19 pipeline tables reproduced identically. WHAT IT DOES NOT CONTAIN The TALIS 2024 individual-level microdata are published by the OECD and are not redistributed here, and neither is the derived analytic file, which is a recoded subset of the same individual records. The deposit releases aggregate weighted estimates only. The file docs/OBTAINING_THE_DATA.md gives the download URLs, the exact commands, the expected record and variable counts and the SHA-256 checksums of the source files, so a replication can verify that it starts from identical data; obtaining them takes a few minutes and requires no application or registration. LICENSING Code is released under the MIT licence; derived tables, figures and documentation under Creative Commons Attribution 4.0 International. Nothing in this deposit licenses the OECD microdata, whose use is governed by the OECD's own terms.