Building Fair AI models tutorial at PyData New York, 2018

This tutorial uses an open source Python package named AI Fairness 360 or AIF360.

Please visit the above site and follow instructions to install the package.

Additionally, download datasets following instructions at https://github.com/IBM/AIF360/tree/master/aif360/data

Alternatively, instructions are provided below for manually installing AIF360 and downloading datasets using Conda on Windows

Create and activate environment

conda create --name aif360 python=3.5
conda activate aif360

Clone AIF360 from GitHub:

git clone https://github.com/IBM/AIF360

Install R-essentials for downloading MEPS data

conda install -c r r-essentials

Download datasets and place under appropriate folders under AIF360/aif360/data/raw by cloning this repository (NOTE: clone at same level as AIF360) and running the belowmentioned notebooks in the root folder

git clone https://github.com/monindersingh/pydata2018_fairAI_models_tutorial.git

Change to the root folder of just cloned repository and run

jupyter notebook pydata_datasets.ipynb
jupyter notebook pydata_meps_datasets.ipynb

Then, navigate to the root directory of the cloned AIF360 project and run:

pip install .

Finally, install the additional requirements as follows:

conda install ecos
pip install -r requirements.txt