Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Odyssey of Commonsense Causality: From Foundational Benchmarks to Cutting-Edge Reasoning

Shaobo Cui, Zhijing Jin, Bernhard Schölkopf, Boi Faltings

TL;DR

This survey addresses the need for a consolidated view of commonsense causality at the intersection of knowledge, benchmarks, acquisition, and reasoning. It develops a taxonomy by commonsense type and uncertainty, reviews three primary data-collection approaches (extractive, generative, manual), and distinguishes qualitative from quantitative reasoning approaches, including LLM- and neuro-symbolic-based methods. It catalogs around 37 benchmarks and discusses the trade-offs, scalability, and explainability of data-collection methods, while outlining future directions such as context sensitivity, complex causal structures, temporality, probabilistic modeling, and multimodal data. By providing a pragmatic handbook and synthesizing insights across 200+ articles, the work aims to guide researchers and accelerate progress in commonsense causality in the era of large language models.

Abstract

Understanding commonsense causality is a unique mark of intelligence for humans. It helps people understand the principles of the real world better and benefits the decision-making process related to causation. For instance, commonsense causality is crucial in judging whether a defendant's action causes the plaintiff's loss in determining legal liability. Despite its significance, a systematic exploration of this topic is notably lacking. Our comprehensive survey bridges this gap by focusing on taxonomies, benchmarks, acquisition methods, qualitative reasoning, and quantitative measurements in commonsense causality, synthesizing insights from over 200 representative articles. Our work aims to provide a systematic overview, update scholars on recent advancements, provide a pragmatic guide for beginners, and highlight promising future research directions in this vital field.

The Odyssey of Commonsense Causality: From Foundational Benchmarks to Cutting-Edge Reasoning

TL;DR

This survey addresses the need for a consolidated view of commonsense causality at the intersection of knowledge, benchmarks, acquisition, and reasoning. It develops a taxonomy by commonsense type and uncertainty, reviews three primary data-collection approaches (extractive, generative, manual), and distinguishes qualitative from quantitative reasoning approaches, including LLM- and neuro-symbolic-based methods. It catalogs around 37 benchmarks and discusses the trade-offs, scalability, and explainability of data-collection methods, while outlining future directions such as context sensitivity, complex causal structures, temporality, probabilistic modeling, and multimodal data. By providing a pragmatic handbook and synthesizing insights across 200+ articles, the work aims to guide researchers and accelerate progress in commonsense causality in the era of large language models.

Abstract

Understanding commonsense causality is a unique mark of intelligence for humans. It helps people understand the principles of the real world better and benefits the decision-making process related to causation. For instance, commonsense causality is crucial in judging whether a defendant's action causes the plaintiff's loss in determining legal liability. Despite its significance, a systematic exploration of this topic is notably lacking. Our comprehensive survey bridges this gap by focusing on taxonomies, benchmarks, acquisition methods, qualitative reasoning, and quantitative measurements in commonsense causality, synthesizing insights from over 200 representative articles. Our work aims to provide a systematic overview, update scholars on recent advancements, provide a pragmatic guide for beginners, and highlight promising future research directions in this vital field.
Paper Structure (48 sections, 10 equations, 10 figures, 8 tables)

This paper contains 48 sections, 10 equations, 10 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Taxonomy and Benchmarks
  3. Classification by Commonsense Types
  4. Classification by Uncertainty Levels
  5. Causality Acquisition
  6. Extractive Methods
  7. Generative Methods
  8. Manual Annotation
  9. Comparison of Data Acquisition Methods
  10. Reasoning Over Causality
  11. Bypassing Uncertainty by Qualitative Causal Reasoning
  12. Measuring Uncertainty by Quantitative Causal Reasoning
  13. Future Research Directions
  14. Conclusion
  15. Real-World Applications of Commonsense Causality
  16. ...and 33 more sections

Figures (10)

  • Figure 1: Different aspects of commonsense causality and their link to different sections of this survey.
  • Figure 2: Taxonomy of commonsense causality in various aspects. The benchmarks, datasets, and methods in blue color are about counterfactual. Leaf nodes with different colors are associated with different sections of this survey.
  • Figure 3: A template of pattern matching from AltLex hidey-mckeown-2016-identifying.
  • Figure 4: Comparison of different causal strength metrics suppes-1973-probabilisticeells-1991-probabilisticpearl-2009-causality.
  • Figure 5: Overview of causal NLP tasks and required skill sets.
  • ...and 5 more figures