Data and Coding Records for Data Risk Formation and Evolution in Human–AI Interaction for Health

This dataset supports the study Data Risk Formation and Evolution in Human–AI Interaction for Health. It contains an English-language workbook of 53 publicly documented cases, task-level and relation-level coding records, operational codebooks, reviewed factor–event and event–harm relations, network edges, node metrics, Bayesian-smoothed relation estimates, key risk paths, and the contextual risk-event distribution matrix. The data were derived from academic literature, institutional reports, regulatory materials, and publicly documented cases. Each record retains its source identifier and available URL. Full-text publications and third-party case documents are not redistributed because of copyright and access restrictions. The relation estimates and path scores describe relative support within the coded evidence. They do not represent real-world incident probabilities or causal effects.