Lune

ACL Findings2021

Rationalization through Concepts

Diego Antognini, Boi Faltings

Abstract

Automated predictions require explanations to be interpretable by humans. One type of explanation is a rationale, i.e., a selection of input features such as relevant text snippets from which the model computes the outcome. However, a single overall selection does not provide a complete explanation, e.g., weighing several aspects for decisions. To this end, we present a novel self-interpretable model called ConRAT. Inspired by how human explanations for high-level decisions are often based on key concepts, ConRAT extracts a set of text snippets as concepts and infers which ones are described in the document. Then, it explains the outcome with a linear aggregation of concepts. Two regularizers drive ConRAT to build interpretable concepts. In addition, we propose two techniques to boost the rationale and predictive performance further. Experiments on both single-and multi-aspect sentiment classification tasks show that ConRAT is the first to generate concepts that align with human rationalization while using only the overall label. Further, it outperforms state-of-the-art methods trained on each aspect label independently.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

6 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • D-Separation for Causal Self-Explanation cites Learning to Faithfully Rationalize by Construction
  • D-Separation for Causal Self-Explanation cites Towards Trustworthy Explanation: On Causal Rationalization
  • D-Separation for Causal Self-Explanation cites Invariant Rationalization
  • D-Separation for Causal Self-Explanation cites An Information Bottleneck Approach for Controlling Conciseness in Rationale Extraction
  • Learning to Faithfully Rationalize by Construction cites Learning to Deceive with Attention-Based Explanations
  • Towards Faithful Explanations: Boosting Rationalization with Shortcuts Discovery cites D-Separation for Causal Self-Explanation
  • Towards Faithful Explanations: Boosting Rationalization with Shortcuts Discovery cites Invariant Rationalization
  • Towards Faithful Explanations: Boosting Rationalization with Shortcuts Discovery cites An Information Bottleneck Approach for Controlling Conciseness in Rationale Extraction
  • Interpretable Concept Bottlenecks to Align Reinforcement Learning Agents cites Concept Bottleneck Models
  • Towards Trustworthy Explanation: On Causal Rationalization cites Learning to Faithfully Rationalize by Construction
  • Towards Trustworthy Explanation: On Causal Rationalization cites Invariant Rationalization
  • Towards Trustworthy Explanation: On Causal Rationalization cites An Information Bottleneck Approach for Controlling Conciseness in Rationale Extraction
  • An Information Bottleneck Approach for Controlling Conciseness in Rationale Extraction cites Learning to Faithfully Rationalize by Construction
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers6

  1. Interactive Concept Bottleneck ModelsKushal Chauhan, Rishabh Tiwari, Jan Freyberg, Pradeep Shenoy et al.AAAI 2023 · 91 citations
  2. Interpretable Concept Bottlenecks to Align Reinforcement Learning AgentsQuentin Delfosse, Sebastian Sztwiertnia, Mark Rothermel, Wolfgang Stammer et al.NeurIPS 2024 · 32 citations
  3. D-Separation for Causal Self-ExplanationWei Liu, Jun Wang, Haozhao Wang, Ruixuan Li et al.NeurIPS 2023 · 29 citations
  4. Towards Trustworthy Explanation: On Causal RationalizationWenbo Zhang, Tong Wu, Yunlong Wang, Yong Cai et al.ICML 2023 · 25 citations
  5. Towards Faithful Explanations: Boosting Rationalization with Shortcuts DiscoveryLinan Yue, Qi Liu, Yichao Du, Li Wang et al.ICLR 2024 · 10 citations

Builds on5

  1. Concept Bottleneck ModelsPang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann et al.ICML 2020 · 1,233 citations
  2. Invariant RationalizationShiyu Chang, Yang Zhang, Mo Yu, Tommi S. JaakkolaICML 2020 · 232 citations
  3. Learning to Deceive with Attention-Based ExplanationsDanish Pruthi, Mansi Gupta, Bhuwan Dhingra, Graham Neubig et al.ACL 2020 · 17 citations
  4. An Information Bottleneck Approach for Controlling Conciseness in Rationale ExtractionBhargavi Paranjape, Mandar Joshi, John Thickstun, Hannaneh Hajishirzi et al.EMNLP 2020 · 13 citations
  5. Learning to Faithfully Rationalize by ConstructionSarthak Jain, Sarah Wiegreffe, Yuval Pinter, Byron C. WallaceACL 2020

Related papers

  1. SELFEXPLAIN: A Self-Explaining Architecture for Neural Text ClassifiersDheeraj Rajagopal, Vidhisha Balachandran, Eduard H. Hovy, Yulia TsvetkovEMNLP 2021 · 39 citations
  2. Knowledge-Grounded Self-Rationalization via Extractive and Natural Language ExplanationsBodhisattwa Prasad Majumder, Oana Camburu, Thomas Lukasiewicz, Julian J. McAuleyICML 2022 · 40 citations
  3. Flexible Instance-Specific Rationalization of NLP ModelsGeorge Chrysostomou, Nikolaos AletrasAAAI 2022 · 17 citations
  4. MARE: Multi-Aspect Rationale Extractor on Unsupervised Rationale ExtractionHan Jiang, Junwen Duan, Zhe Qu, Jianxin WangEMNLP 2024 · 2 citations
  5. Learning Robust Rationales for Model Explainability: A Guidance-Based ApproachShuaibo Hu, Kui YuAAAI 2024 · 11 citations
Dusk over the sea between two cliffs drawn in fine vertical lines