Welcome to this notebook on Responsible AI. This project aims to touch on the most important topics in the field of Responsible AI including bias, fairness, interpretability, error analysis and counterfactuals.
We use the famous UCI Adult Census Income dataset to apply the responsible AI concepts.
- ML models: After an in-depth exploratory analysis of the dataset we create two white-box models (Decision Trees) and one black-box model (Gradient Boosting).
- Interpretating models: This section aims to introduce common interpretability techniques including global (tree plots, feature importance, PDP plots) and local (SHAP, LIME) techniques
- Error Analysis: Here we compare the models based on fairness and create a bias-mitigated model using the fairlearn library. We further investigate potential fairness-accuracy tradeoffs and discover the sources of bias in the data.
- What if Counterfacticals: Inspired by Causal ML modelling we test how the models' prediction changes when we change individual features using the DiceML library.
- Documentation: We create a datasheet for the dataset and a model card for our most fair model.
- References
To run the notebook locally just install all required packages.
pip install -r requirements.txt