Summary of Towards Faithful Explanations: Boosting Rationalization with Shortcuts Discovery, by Linan Yue et al.
Towards Faithful Explanations: Boosting Rationalization with Shortcuts Discovery
by Linan Yue, Qi Liu, Yichao Du, Li Wang, Weibo Gao, Yanqing An
First submitted to arxiv on: 12 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel approach to neural network rationalization, building upon the idea that selective rationalization can be used to identify a small subset of inputs sufficient to support predictions. The proposed Shortcuts-fused Selective Rationalization (SSR) method aims to boost rationalization by discovering and exploiting potential shortcuts in data. Specifically, SSR uses a shortcuts discovery approach to detect several potential shortcuts, then introduces the identified shortcuts to mitigate the problem of utilizing shortcuts to compose rationales. To further improve the method’s effectiveness, two data augmentations methods are developed to close the gap in the number of annotated rationales. The proposed approach is validated through extensive experimental results on real-world datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper tries to figure out why neural networks work well by finding a few important pieces of information that make predictions happen. Right now, there aren’t many ways to explain how neural networks make decisions and even fewer large-scale examples that humans have annotated. To solve this problem, the authors suggest a new method called Shortcuts-fused Selective Rationalization (SSR) that can find potential shortcuts in data and use them to improve rationalization. The authors also develop two ways to add more examples to the dataset so there are enough to test the method properly. By doing these things, SSR shows it can be very effective at explaining neural network decisions. |
Keywords
* Artificial intelligence * Neural network
