A Game Theoretic Approach to Class-wise Selective Rationalization

By Shiyu Chang, Yang Zhang, Mo Yu, Tommi S. Jaakkola. 2019

66 citations recorded by the source corpus.

Selection of input features such as relevant pieces of text has become a common technique of highlighting how complex neural predictors operate. The selection can be optimized post-hoc for trained models or incorporated directly into the method itself (self-explaining). However, an overall selection does not properly capture the multi-faceted nature of useful rationales such as pros and cons for decisions. To this end, we propose a new game theoretic approach to class-dependent rationalization, where the method is specifically trained to highlight evidence supporting alternative conclusions. Each class involves three players set up competitively to find evidence for factual and counterfactual scenarios. We show theoretically in a simplified scenario how the game drives the solution towards meaningful class-dependent rationales. We evaluate the method in single- and multi-aspect sentiment classification tasks and demonstrate that the proposed method is able to identify both factual (justifying the ground truth label) and counterfactual (countering the ground truth label) rationales consistent with human rationalization. The code for our method is publicly available.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

BERTology Meets Biology: Interpreting Attention in Protein Language Models (2020, background): “Methods that learn to explain by extracting rationales from the input and only using those rationales to make a prediction are faithful by design [8, 13, 43, 62].”
ERASER: A Benchmark to Evaluate Rationalized NLP Models (2019, background): “Elsewhere, Chang et al. (2019) introduced the notion of classwise rationales that explains support for different output classes using a game theoretic framework.”
Learning to Give Checkable Answers with Prover-Verifier Games (2021, unknown): “Chang et al. [4] propose a six player game where the 4 generators select factual or counterfactual rationales, and the 2 discriminators aim to distinguish factual and counterfactual rationales.”
Explaining and Improving Model Behavior with k Nearest Neighbor Representations (2020, methodology): “Gradient-based saliency maps are faithful by construction and hence have been applied in various forms to neural models for interpretability (Lei, Barzilay, and Jaakkola 2016; Ribeiro, Singh, and Guestrin 2018; Chang et al. 2019; Bast-ings, Aziz, and Titov 2019).”
Generative causal explanations of black-box classifiers (2020, background): “[58] Shiyu Chang, Yang Zhang, Mo Yu, and Tommi Jaakkola.”
Explaining NLP Models via Minimal Contrastive Editing (MiCE) (2020, background): “…popular instance-based explanation methods produce highlights—segments of input that support a prediction (Zaidan et al., 2007; Lei et al., 2016; Chang et al., 2019; Bastings et al., 2019; Yu et al., 2019; DeYoung et al., 2020; Jain et al., 2020; Belinkov and Glass, 2019) that can be derived…”
An Information Bottleneck Approach for Controlling Conciseness in Rationale Extraction (2020, background): “Building on Lei et al. (2016), Chang et al. (2019) and Yu et al. (2019) consider beneﬁts like class-wise explanation extraction while Chang et al. (2020) explore invariance to domain shift.”
Paragraph-level Rationale Extraction through Regularization: A case study on European Court of Human Rights Cases (2021, methodology): “…the main focus is to build models that learn to provide proper Correspondence to: ihalk.aueb.gr justification for their decisions, similar to those of humans, (Zaidan et al., 2007; Lei et al., 2016; Chang et al., 2019; Yu et al., 2019) by requiring the models to satisfy additional constraints.”