civic.icarm - Interpretable Civic-Accountable and Responsible Machine Learning

A general-purpose framework for Interpretable Civic-Accountable and Responsible Machine Learning (ICARM). Works with any clean tabular data and automatically detects whether a task is binary classification, multi-class classification, or regression from the target variable type. Provides a single unified entry point civic_fit() alongside tidy interfaces for global and local model explanations, group-level fairness auditing, probability calibration, multi-model comparison, threshold analysis, and reproducible audit trails. Designed to support the DataCitizen-Pro research agenda at Ludwigsburg University of Education: developing data literacy, statistical reasoning, and democratic judgment formation in civic and political teacher education. References: Biecek (2018) <doi:10.18637/jss.v085.i04>, Kuhn (2008) <doi:10.18637/jss.v028.i05>, Awe (2025) <https://github.com/Olawaleawe/civic.icarm>.