# "Why Should I Trust You?": Explaining the Predictions of Any Classifier

**Authors:** Marco Tulio Ribeiro, Sameer Singh, Carlos Guestrin  
**arXiv:** [1602.04938](https://arxiv.org/abs/1602.04938)  
**Date:** 2016-02-16  
**Summary page:** https://www.summarizepaper.com/en/paper/1602.04938v1/

## Key points

- Machine learning models are often seen as black boxes, which hinders trust in their predictions.
- Understanding the reasons behind predictions is crucial for assessing trust in a model.
- LIME (Local Interpretable Model-Agnostic Explanations) is a novel explanation technique proposed by the authors.
- LIME provides interpretable and faithful explanations for predictions made by any classifier.
- LIME learns an interpretable model locally around each prediction to achieve this.
- The authors also introduce a method to explain models by presenting representative individual predictions and their explanations in a non-redundant manner.
- LIME's flexibility is demonstrated through its application to text and image classification models.
- The usefulness of explanations provided by LIME is validated through experiments conducted in simulated environments and with human subjects.
- Explanations empower users in various scenarios where trust is essential, such as deciding whether to trust a prediction, choosing between different models, improving an untrustworthy classifier, and identifying reasons why a classifier should not be trusted.
- Overall, LIME is presented as an effective solution for explaining machine learning model predictions in an interpretable and trustworthy manner.

## Summary

In the paper titled "Why Should I Trust You? ": Explaining the Predictions of Any Classifier, authors Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin address the issue of machine learning models being black boxes despite their widespread adoption. They emphasize the importance of understanding the reasons behind predictions in order to assess trust in a model. To tackle this problem, the authors propose LIME (Local Interpretable Model-Agnostic Explanations), a novel explanation technique that provides interpretable and faithful explanations for predictions made by any classifier. LIME achieves this by learning an interpretable model locally around each prediction. Additionally, the authors introduce a method to explain models by presenting representative individual predictions and their explanations in a non-redundant manner, framing it as a submodular optimization problem. The flexibility of LIME is demonstrated through its application to different models for text (such as random forests) and image classification (including neural networks). The usefulness of explanations is validated through novel experiments conducted both in simulated environments and with human subjects. The authors highlight that their explanations empower users in various scenarios where trust is essential. These scenarios include deciding whether to trust a prediction, choosing between different models, improving an untrustworthy classifier, and identifying reasons why a classifier should not be trusted. Overall, this paper presents LIME as an effective solution for explaining machine learning model predictions in an interpretable and trustworthy manner. By providing insights into the underlying reasons behind predictions, LIME enables users to make informed decisions based on these predictions and enhances trust in machine learning models.