The Diligent Proxy — in-silico pilot materials and aggregated results (badge over-reliance experiment)

Supporting materials for the article "The Diligent Proxy: What an In-Silico LLM Pilot Can and Cannot De-Risk in a Pre-Registered Automation-Bias Experiment" (Behavior Research Methods, submitted). Contains the four synthetic record-state stimuli (sound, stale, replay, misleading) with seeded defects across three cue conditions (silent/badge/workspace), v1 and v2 stimulus sets, the review instrument and scoring rubric, and the aggregated per-cell results (APPROVE rates and Likert means by case x condition; 12 cells, n approx 20/cell) for both in-silico runs. This was an exploratory in-silico pilot using LLM reviewers; per-trial records and the one-off simulation harness were not retained, so results are provided at the aggregate level, from which every quantity reported in the article is reproducible. Fully synthetic stimuli; no human-subjects, proprietary, or personal data.