A Game-Theoretic and Introspective Framework for Interpreting Neural Predictions
First Author Affiliation / Address line 1 Affiliation / Address line 2 Affiliation / Address line 3、Second Author Affiliation / Address line 1 Affiliation / Address line 2 Affiliation / Address line 3
Abstract
Many recent self-explainable models for NLP adopt a two-step extractive framework where the first model generates a subset of the input as the explanation and the second model makes prediction over this explanation. In this paper, we first point out that all these works fall to a collaborative game formulation thus may suffer from unexpected behaviors, including (1) the two models can simply communicate with their own specific code on class labels, making the extracted explanations less meaningful; and (2) training of the first-step model may be difficult to capture sufficient information especially with less discriminative supervision. To deal with these problems, we propose a new three-player game framework to filter out the undesirable communications, and a new introspective model to encourage the explanation generation to be more aware of potential predictions. The new framework theoretically guarantees the good properties of the generated explanations, as well as gives improved empirical results on text classification tasks.
中文速览
现有的NLP自解释模型普遍采用"生成器提取关键文本片段、预测器基于片段做判断"的两步框架，但这种协作博弈机制存在严重缺陷：生成器可能只是用隐晦的编码规律把标签信息"偷传"给预测器，而非真正提取有意义的解释片段，同时生成器也难以从粗粒度标签中捕获足够的判别信息。为此，本文提出了一个三方博弈框架，在原有生成器和预测器之外引入第三方"补集预测器"，专门利用未被选中的文本来预测标签，迫使生成器必须把真正关键的信息纳入解释片段；同时设计了一个"内省生成器"，让生成器在提取片段时参考完整文本对应的预测标签，从而保证提取结果与真实判断依据一致。在文本分类任务上的实验表明，新框架显著减少了退化现象并提升了预测准确率，理论上也证明了三方博弈的均衡点满足解释片段应具备的充分性与非退化性等公理化性质。这项工作为可解释NLP提供了更严格的理论保障和更可靠的实践方案，对医疗等高风险决策场景尤具价值。
原文 arXiv:1910.13294;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1910.13294v2