Conformal Prediction Interval Counterfactuals

This repository provides a demonstration of Conformal Prediction Interval Counterfactuals (CPICFs) -- a way of generating counterfactual examples for tabular data chosen to be informative to the recpient by choosing those with a large conformal prediction interval. It accompanies the submission of the 2025 COPA conference paper "Individualised Counterfactual Examples Using Conformal Prediction Intervals".

Conformal Counterfactual Generation Demo

Installation Instructions

For running the jupyter notebook locally:

1. Create a virtual environment for your operating system python3.10 -m venv env
2. Install requirements for notebook

pip install -r requirements.txt

and set up ipykernel

python3 -m pip install ipykernel
python3 -m ipykernel install --user --name CPICF

3. Open the jupyter notebook and select the CPICF kernel