# bLIMEy: Surrogate Prediction Explanations Beyond LIME

**Authors:** Kacper Sokol, Alexander Hepburn, Raul Santos-Rodriguez, Peter Flach  
**arXiv:** [1910.13016](https://arxiv.org/abs/1910.13016)  
**Date:** 2019-10-29  
**Summary page:** https://www.summarizepaper.com/en/paper/1910.13016v1/

## Key points

- Surrogate explainers are important in interpreting black-box machine learning predictions in eXplainable Artificial Intelligence (XAI)
- Surrogate explainers can be applied to various types of data such as images, text, and tabular data
- Local Interpretable Model-agnostic Explanations (LIME) algorithm is often mistakenly considered the ultimate solution for surrogate explainability
- The paper titled "bLIMEy: Surrogate Prediction Explanations Beyond LIME" proposes a principled algorithmic framework for building custom local surrogate explainers
- The authors demonstrate how different component choices impact the functional capabilities of the resulting explainer using LIME as an example
- The proposed bLIMEy framework allows tailoring of surrogate explainers according to specific needs and requirements
- This research contributes to advancing XAI methods and promoting transparency in machine learning models.

## Summary

In the field of eXplainable Artificial Intelligence (XAI), surrogate explainers play a crucial role in interpreting black-box machine learning predictions. These explainers are versatile as they can be applied to various types of data such as images, text, and tabular data. They are also model-agnostic and can be retrofitted, making them highly valuable for understanding complex models. The Local Interpretable Model-agnostic Explanations (LIME) algorithm is often mistakenly considered as the ultimate solution for surrogate explainability. However, in this paper titled "bLIMEy: Surrogate Prediction Explanations Beyond LIME," authors Kacper Sokol, Alexander Hepburn, Raul Santos-Rodriguez, and Peter Flach propose a principled algorithmic framework that empowers the community to build custom local surrogate explainers. The authors demonstrate how to decompose the family of surrogate explainers into algorithmically independent and interoperable modules. By doing so, they highlight the influence of different component choices on the functional capabilities of the resulting explainer. They specifically focus on LIME as an example within this framework. This research expands on existing knowledge by providing a systematic approach to constructing local surrogate explainers for black-box model predictions. The proposed bLIMEy framework allows researchers and practitioners to tailor these explainers according to their specific needs and requirements. By offering a deeper understanding of how different components impact the functionality of surrogate explainers, this work contributes significantly to advancing XAI methods and promoting transparency in machine learning models.