Data and code for human-centered component-level evaluation of spoken AI psychological support
Description
This dataset contains the deidentified item-level analysis data and reproducibility code for a randomised 2 × 2 factorial experiment crossing psychotherapist-generated and AI-generated text with psychotherapist and synthetic voices across eight hypothetical psychological-distress scenarios. The final analysis sample comprises 1519 retained Chinese adults. The package includes the final analysis workbook, a variable dictionary, executable R scripts, dependency and execution instructions, complete statistical outputs, reports covering 157 selected numerical verification checks, and results from 5000 bootstrap resamples. Version 2 extends the earlier code-only release by providing the analysis data and updated reproducibility materials. The study was retrospectively registered on OSF (https://doi.org/10.17605/OSF.IO/KJCT5), and the analyses are exploratory. The workbook does not include withdrawn or quality-excluded participants. Earlier participant-flow totals are author-reported and cannot be independently reconstructed from these final-sample data. Audio recordings are not included because of third-party licensing and redistribution restrictions affecting some synthetic voices.