The analytic multiverse of AI-assisted directed content analysis: a framework, open tools, and a screening study

Code, materials and derived data for the preprint “The analytic multiverse of AI-assisted directed content analysis: a framework, open tools, and a screening study” (van Allen, 2026). In the screening study, two large language models coded 40 open-access smoking-cessation trial reports for 94 behaviour change techniques under a half-fraction factorial design (32 configurations, 38 runs). Every run was scored against Human Behaviour-Change Project annotations. The project contains the prospective multiverse declaration, the prompts and codebook, the report × code decisions of every run, the reference, the figures, and analysis scripts that reproduce every reported result without new model calls. No report text or quotations are included. The tools and declaration can be adapted to run comparable multiverse studies on other data.