Supplementary dataset for "LLM-Mediated Psychometric Trust Assessment for Counterparty Risk Management"

This dataset contains the supplementary materials for the paper “LLM-Mediated Psychometric Trust Assessment for Counterparty Risk Management”. The dataset supports an exploratory case study comparing LLM-derived psychometric trust scores with the 2023 Axios Harris Poll 100 corporate reputation benchmark. The package includes cleaned benchmark data, item-level GPT-4, GPT-4o, and repeated GPT-4 trust scores, merged summary data, reproducible Python and R analysis scripts, generated statistical outputs, and figure source data. The analysis reports aggregate Pearson and Spearman correlations, MAE, RMSE, item-level diagnostics, cross-run comparisons, and Cronbach’s alpha reliability estimates for the LLM-derived trust-item scales. The LLM assessments were conducted on 15 January 2025 using the ChatGPT user-visible GPT-4 and GPT-4o model selections. Exact backend model snapshots and independent source manifests were not retained, and the dataset should therefore be interpreted as supporting an exploratory LLM-mediated reputation/trust extraction study rather than a fully reproducible retrieval-augmented evidence review.