Interpretability and explainability from an ethical AI perspective
It comes as no surprise that Google's WIT AI fairness example notebook comes with a biased dataset. It is up to us to use WIT to point out the unethical or moral dilemmas contained in the COMPAS dataset.
COMPAS stands for Correctional Offender Management Profiling for Alternative Sanctions. Judges and parole officers are said to score the probability of recidivism for a given defendant.
In this section, our task will be to transform the dataset into an unbiased, ethical dataset. We will analyze COMPAS with people-centered AI before importing it.
We will first describe the ethical and legal perspectives. Then we will define AI explainability and interpretability for our COMPAS example. Finally, we will prepare an ethical dataset for our model by improving the feature names of the dataset.
Our ethical process fits the spirit of Google's research team when they designed WITÂ for us to...
