Measuring Association Between Labels and Free-Text Rationales

By Sarah Wiegreffe, Ana Marasovic, Noah A. Smith. 2020

219 citations recorded by the source corpus.

In interpretable NLP, we require faithful rationales that reflect the model’s decision-making process for an explained instance. While prior work focuses on extractive rationales (a subset of the input words), we investigate their less-studied counterpart: free-text natural language rationales. We demonstrate that *pipelines*, models for faithful rationalization on information-extraction style tasks, do not work as well on “reasoning” tasks requiring free-text rationales. We turn to models that *jointly* predict and rationalize, a class of widely used high-performance models for free-text rationalization. We investigate the extent to which the labels and rationales predicted by these models are associated, a necessary property of faithful explanation. Via two tests, *robustness equivalence* and *feature importance agreement*, we find that state-of-the-art T5-based joint models exhibit desirable properties for explaining commonsense question-answering and natural language inference, indicating their potential for producing faithful free-text rationales.

Subfields: Artificial Intelligence, Natural Language Processing

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

In Natural Language Processing

Citations in the corpus

Chain of Thought Prompting Elicits Reasoning in Large Language Models (2022, neutral): “That line of work typically focuses on natural language inference (Camburu et al., 2018; Yordanov et al., 2021; Bostrom et al., 2021), and produces explanations either simultaneously to or after the ﬁnal prediction (Narang et al., 2020; Majumder et al., 2021; Wiegreffe et al., 2021, 2022).”
Training Language Models with Language Feedback (2022, methodology): “Prior work explores ways of using explanations to train text classiﬁcation models, with mixed re-sults (Camburu et al., 2018; Stacey et al., 2021; Pruthi et al., 2021; Wiegreffe et al., 2021; Hase and Bansal, 2021; Lampinen et al., 2022, inter alia ).”
Reframing Human-AI Collaboration for Generating Free-Text Explanations (2021, background): “This line of research is complementary to works investigating the faithfulness of model-generated free-text explanations (Hase et al., 2020; Wiegreffe et al., 2021).”
Few-Shot Out-of-Domain Transfer Learning of Natural Language Explanations (2021, background): “There are three main focuses in NLE generation: quality improvement (Camburu et al., 2018; Narang et al., 2020; Valentino et al., 2022), NLE faithfulness (Kumar and Talukdar, 2020; Wiegreffe et al., 2021; Liu et al., 2019; Latcinnik and Berant, 2020), and transfer learning of NLEs.”
Few-Shot Self-Rationalization with Natural Language Prompts (2021, methodology): “Self-rationalization models (Narang et al., 2020; Wiegreffe et al., 2021) are currently based on T5 for at least two reasons.”
SELFEXPLAIN: A Self-Explaining Architecture for Neural Text Classifiers (2021, background): “A good model explanation should be (i) relevant to the current input and predictions and (ii) understandable to humans (DeYoung et al., 2020; Jacovi and Goldberg, 2020; Wiegreffe et al., 2020; Jain et al., 2020).”
Knowledge-Grounded Self-Rationalization via Extractive and Natural Language Explanations (2021, background): “Evaluating the faithfulness of explanations is a challenging open question for both ERs (Jacovi & Goldberg, 2021) and NLEs (Wiegreffe et al., 2021).”
The Unreliability of Explanations in Few-Shot In-Context Learning (2022, methodology): “It is also called a pipeline model in other literature on training models using explanations (Jacovi and Goldberg, 2021; Wiegreffe et al., 2021).”
Learning to Rationalize for Nonmonotonic Reasoning with Distant Supervision (2020, background): “However, whether or not attention weights provide reliable insights into the model’s decisionmaking process is debatable (Serrano and Smith 2019; Jain and Wallace 2019; Wiegreffe and Pinter 2019).”
Teach Me to Explain: A Review of Datasets for Explainable NLP (2021, background): “This makes them useful for explaining reasoning tasks where explanations must contain information outside the given input sentence or document (Camburu et al., 2018; Wiegreffe et al., 2020).”
Decomposition Enhances Reasoning via Self-Evaluation Guided Decoding (2023, unknown): “One approach is to generate rationales to enhance model interpretability (Zhou et al., 2020; Wiegreffe and Marasović, 2021; Wiegreffe et al., 2021).”
The Unreliability of Explanations in Few-shot Prompting for Textual Reasoning (2022, methodology): “Past work has built a series of pipeline models that first generate the explanations and then make predictions purely based on the generated explanations (Wiegreffe et al., 2021; Zhou and Tan, 2021; Chen et al., 2022).”
Rissanen Data Analysis: Examining Dataset Characteristics via Description Length (2021, background): “Such reasons include written explanations (Camburu et al., 2018; Rajani et al., 2019; Wiegreffe et al., 2020) or locating taskrelevant input words (Zhang et al., 2016b; Perez et al., 2019).”
Training Language Models with Language Feedback at Scale (2023, methodology): “Prior work explores ways of using explanations to train text classification models, with mixed results (Camburu et al., 2018; Stacey et al., 2021; Pruthi et al., 2021; Wiegreffe et al., 2021; Hase & Bansal, 2021; Lampinen et al., 2022, inter alia ).”
Rationale-Inspired Natural Language Explanations with Commonsense (2021, methodology): “Thus, our experiments confirm that REXC exhibits a strong association between task outputs and NLEs, passing the faithfulness sanity check as proposed in (Wiegreffe et al., 2020).”
Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes (2023, methodology): “…to the input examples and train the model to output differently based on the given task preﬁx (Section 3.2). man rationales can be used as gold standard labels to make models more interpretable by generating similar rationales (Wiegreffe et al., 2021; Narang et al., 2020; Eisenstein et al., 2022).”
Prompting Contrastive Explanations for Commonsense Reasoning Tasks (2021, background): “However, the space of possible free-form explanations is incredibly large, inherently ambiguous, and difficult to annotate or evaluate (Wiegreffe et al., 2020; Latcinnik and Berant, 2020).”
Training Language Models with Natural Language Feedback (2022, methodology): “Prior work explores ways of using explanations to train text classiﬁcation models, with mixed re-sults (Camburu et al., 2018; Stacey et al., 2021; Pruthi et al., 2021; Wiegreffe et al., 2021; Hase and Bansal, 2021; Lampinen et al., 2022, inter alia ).”
Teach Me to Explain: A Review of Datasets for Explainable Natural Language Processing (2021, background): “This makes them useful for explaining reasoning tasks where explanations must contain information outside the given input sentence or document [20, 128].”