Artificial Intelligence for all

Overview

This Git repository contains related material to a presentation from Rachel Orti and Melanie Shilpa Rao at Devoxx France 2019, "L'Intelligence Artificielle pour tous" (Artificial Intelligence for all).

Installation Guide

You can run the Jupyter notebooks using a Jupyter Notebook server that is provided in the current repository.

Launch the Jupyper Notebook server

Requirements: You must have Docker installed.

• From the docker folder, use the following command to launch the Jupyter Notebook server with docker-compose:

docker-compose up

NB: Use docker-compose down to stop the Docker container.

• The Jupyter Notebook server is then available at the following URL: http://localhost:8888

Run the Jupyter notebooks

Some notebooks are provided in the docker/notebook/resources/jupyter/work folder:

• AIF360_COMPAS.ipynb - Example of bias removal in dataset using the Reweighing bias mitigation algorithm implemented in the AI Fairness 360 Toolkit (IBM)
• WIT_COMPAS.ipynb - Example of bias removal in model using the What-If tool (Google)

They are directly accessible at the following URL when the Jupyper Notebook server is up and running: http://localhost:8888/tree/work

Useful links about fairness in AI

General articles about bias in AI

• Are Algorithms Building the New Infrastructure of Racism?
• Artificial Intelligence and Privacy
• Bias detectives: the researchers striving to make algorithms fair
• How to prevent bias in machine learning
• Responsible AI Practices
• Algorithmes : biais, discrimination et équité
• Algocratie: l'inégalité Programmée

Examples

• Project Norman
• COMPAS:
  • ProPublica Investigation
  • AI is convicting criminals and determining jail time, but is it fair?
• Gender Shades
  • 2nd Study Results

Movements & organisations

• Trustable AI
• Partnership on AI
• Microsoft FATE
• Algorithmic Justice League
• Algorithm Watch
  • TEDTalk How I'm fighting bias in algorithms
• Fair by Design
• AlgoTransparency
• Data For Good
  • On cherche à éveiller la conscience éthique des développeurs et data scientists
• AI Now Institue

Tools

• AI Fairness 360 Toolkit (IBM):
  • Overview
  • Github
• What-If (Google):
  • Overview
  • Github
• Audit AI(Pymetrics)
• AI Openscale(IBM)
• Teach & Test AI(Accenture)
• Lime
• Activation Atlas (Google):
  • Overview
  • Github