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)
civic.icarm 
civic.icarm provides a unified, pedagogically-grounded R framework for Interpretable, Civic-Accountable, and Responsible Machine Learning (ICARM).
It is the computational backbone of the DataCitizen-Pro project a proposed DFG-funded research programme at Ludwigsburg University of Education (LUE) developing data literacy, statistical reasoning, and democratic judgment in civic and statistical education.
"Algorithmic decisions that affect civic life must be interpretable, auditable, and fair - not merely accurate." DataCitizen-Pro, DFG Sachbeihilfe 2026
Installation
# From CRAN (once accepted)
install.packages("civic.icarm")
# Development version from GitHub
install.packages("civic.icarm")
Quickstart
library(civic.icarm)
# Works with ANY tabular data - task auto-detected
m <- civic_fit(voted ~ ., data = civic_voting)
# Explain
ex <- civic_explain(m, data = civic_voting)
civic_plot_importance(ex)
# Fairness audit
fair <- civic_fairness(m, civic_voting,
outcome = "voted",
protected = "gender",
positive = "yes")
civic_plot_fairness(fair, metric = "tpr")
# Full accountability scorecard
civic_scorecard(m, civic_voting,
outcome = "voted",
protected = "gender",
positive = "yes",
project = "DataCitizen-Pro")
Key functions
| Function | Description |
|---|---|
civic_fit() |
Train any model - auto-detects binary, multiclass, regression |
civic_explain() |
Global feature importance |
civic_fairness() |
Group equity metrics across protected attributes |
civic_calibrate() |
Probability calibration diagnostics |
civic_compare() |
Side-by-side multi-model comparison |
civic_audit() |
Reproducible JSON audit trail |
civic_scorecard() |
Full civic accountability report |
DataCitizen-Pro connection
| Competency pillar | civic.icarm module |
|---|---|
| Data Literacy | civic_fit(), civic_audit() |
| Statistical Reasoning | civic_metrics(), civic_thresholds(), civic_calibrate() |
| Democratic Judgment | civic_fairness(), civic_scorecard() |
Built-in datasets
| Dataset | Rows | Task |
|---|---|---|
civic_voting |
1,000 | Binary classification |
civic_education |
800 | Regression |
civic_german_credit |
1,000 | Binary classification (fairness benchmark) |
Author
Prof. Dr. Olushina Olawale Awe Alexander von Humboldt Foundation Visiting Professor Statistical and Data Science Literacy Ludwigsburg University of Education (LUE), Germany [email protected]
Citation
@software{awe2026civicicarm,
author = {Awe, Olushina Olawale},
title = {{civic.icarm}: Interpretable, Civic-Accountable and
Responsible Machine Learning},
year = {2025},
url = {https://github.com/Olawaleawe/civic.icarm},
note = {R package v0.2.0. DataCitizen-Pro DFG Sachbeihilfe,
Ludwigsburg University of Education.}
}
Acknowledgements
Developed within the DataCitizen-Pro project submitted to the Deutsche Forschungsgemeinschaft (DFG) Sachbeihilfe programme. The Alexander von Humboldt Foundation is thanked for supporting the Visiting Professorship at LUE.