Skip to content
Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Join the chat at

Toolkit for Auditing and Mitigating Bias and Fairness of Machine Learning Systems πŸ”ŽπŸ€–πŸ§°

Responsibly is developed for practitioners and researchers in mind, but also for learners. Therefore, it is compatible with data science and machine learning tools of trade in Python, such as Numpy, Pandas, and especially scikit-learn.

The primary goal is to be one-shop-stop for auditing bias and fairness of machine learning systems, and the secondary one is to mitigate bias and adjust fairness through algorithmic interventions. Besides, there is a particular focus on NLP models.

Responsibly consists of three sub-packages:

  1. responsibly.dataset
    Collection of common benchmark datasets from fairness research.
  2. responsibly.fairness
    Demographic fairness in binary classification, including metrics and algorithmic interventions.
  3. responsibly.we
    Metrics and debiasing methods for bias (such as gender and race) in word embedding.

For fairness, Responsibly's functionality is aligned with the book Fairness and Machine Learning - Limitations and Opportunities by Solon Barocas, Moritz Hardt and Arvind Narayanan.

If you would like to ask for a feature or report a bug, please open a new issue or write us in Gitter.


  • Python 3.6+


Install responsibly with pip:

$ pip install responsibly

or directly from the source code:

$ git clone
$ cd responsibly
$ python install


If you have used Responsibly in a scientific publication, we would appreciate citations to the following:

  author = {Shlomi Hod},
  title =  {{Responsibly}: Toolkit for Auditing and Mitigating Bias and Fairness of Machine Learning Systems},
  year =   {2018--},
  url =    "",
  note =   {[Online; accessed <today>]}