FAT Forensic Events

A 2021 Bristol Interactive AI Summer School (BIAS) session with a hands-on component

Learn more about FAT Forensics: Source Code Documentation

Resources: Slides Demo Jupyter Notebooks Slack

Practical Machine Learning Explainability

Surrogate Explainers and Fairwashing

A hands-on session at the 2021 Bristol Interactive AI Summer School (BIAS).
Where: The Bill Brown Suite, Queens Building, University of Bristol, Bristol, United Kingdom.
When: Monday, September 6th, 2021.

Table of Contents

About the Session

Surrogate explainability is a popular transparency technique for assessing trustworthiness of predictions output by black-box machine learning models. While such explainers are often presented as monolithic, end-to-end tools, they in fact exhibit high modularity and scope for parameterisation1. This observation suggests that each use case may require a bespoke surrogate built and tuned for the problem at hand. This session introduces the three core components of surrogate explainers: data sampling, interpretable representation and explanation generation in view of text, image and tabular data. By understanding these building blocks individually, as well as their interplay, we can build robust and trustworthy explainers. However, we can also misuse these insights to create technically-valid explainers that are intended to produce misleading justifications of individual predictions. For example, by manipulating the size and distribution of the data sample, and the grouping criteria of the interpretable representation, an automated decision may be shown as fair despite the underlying model being inherently biased. This overview of theory is complemented by a no-code hands-on exercise facilitated through an iPython widget delivered via a Jupyter Notebook.

FAT Forensics (Software)

To support the goals of this session, we employ FAT Forensics – an open source Python package that can inspect selected fairness, accountability and transparency aspects of data (and their features), models and predictions. The toolbox spans all of the FAT domains because many of them share underlying algorithmic components that can be reused in multiple different implementations, often across the FAT borders. This interoperability allows, for example, a counterfactual data point generator to be used as a post-hoc explainer of black-box predictions on one hand, and as an individual fairness (disparate treatment) inspection tool on the other. The modular architecture23 enables FAT Forensics to deliver robust and tested low-level FAT building blocks as well as a collection of FAT tools built on top of them. Users can choose from these ready-made tools or, alternatively, combine the available building blocks to create their own bespoke algorithms without the need of modifying the code base.

Schedule and Resources

The session lasts for 2 hours. The first part – 1 hour and 15 minutes – introduces surrogate explainers of text, image and tabular data, discussing their pros, cons and modularisation. The next part – 45 minutes – is devoted to a hands-on exercise demonstrating the importance of tabular surrogate parameterisation, data sampling and interpretable representation composition (discretisation of numerical features) in particular.

Duration Activities Instructor Resources
1.30pm BST
(75 minutes)
Introduction to modular surrogate explainers. Kacper Sokol slides
2.45pm BST
(45 minutes)
Hands-on with parameterising tabular surrogate explainers. Kacper Sokol Jupyter


Kacper Sokol

Kacper is a research associate with the TAILOR project at the University of Bristol. His main research focus is transparency – interpretability and explainability – of machine learning systems. In particular, he has done work on enhancing transparency of logical predictive models (and their ensembles) with counterfactual explanations. Kacper is the designer and lead developer of the FAT Forensics package.



  1. Kacper Sokol, Alexander Hepburn, Raul Santos-Rodriguez, and Peter Flach. 2019. bLIMEy: Surrogate Prediction Explanations Beyond LIME. 2019 Workshop on Human-Centric Machine Learning (HCML 2019) at the 33rd Conference on NeuralInformation Processing Systems (NeurIPS 2019), Vancouver, Canada (2019). https://arxiv.org/abs/1910.13016 

  2. Kacper Sokol, Alexander Hepburn, Rafael Poyiadzi, Matthew Clifford, Raul Santos-Rodriguez, and Peter Flach. 2020. FAT Forensics: A Python Toolbox for Implementing and Deploying Fairness, Accountability and Transparency Algorithms in Predictive Systems. Journal of Open Source Software, 5(49), p.1904. https://joss.theoj.org/papers/10.21105/joss.01904 

  3. Kacper Sokol, Raul Santos-Rodriguez, and Peter Flach. 2019. FAT Forensics: A Python Toolbox for Algorithmic Fairness, Accountability and Transparency. arXiv preprint arXiv:1909.05167. https://arxiv.org/abs/1909.05167