Abstract:It has been said, arguably, that causality analysis should pave a promising way to interpretable deep learning and generalization. Incorporation of causality into artificial intelligence (AI) algorithms, however, is challenged with its vagueness, non-quantitiveness, computational inefficiency, etc. During the past 18 years, these challenges have been essentially resolved, with the establishment of a rigorous formalism of causality analysis initially motivated from atmospheric predictability. This not only opens a new field in the atmosphere-ocean science, namely, information flow, but also has led to scientific discoveries in other disciplines, such as quantum mechanics, neuroscience, financial economics, etc., through various applications. This note provides a brief review of the decade-long effort, including a list of major theoretical results, a sketch of the causal deep learning framework, and some representative real-world applications in geoscience pertaining to this journal, such as those on the anthropogenic cause of global warming, the decadal prediction of El Niño Modoki, the forecasting of an extreme drought in China, among others.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is how to improve the interpretability and generalization ability of artificial intelligence algorithms through quantitative causality analysis. Specifically, the paper focuses on the following aspects: 1. **Application of Causal Analysis in Artificial Intelligence**: The paper explores how to incorporate causal analysis into artificial intelligence algorithms, especially deep - learning models, in order to improve the interpretability and generalization ability of these models. The importance of this problem lies in the fact that current artificial intelligence models often lack transparency and it is difficult to understand their decision - making processes, while causal analysis can provide a method to explain the behavior of the models. 2. **Formalization of Quantitative Causal Relationships**: The paper introduces the work of the author's team in quantifying causal relationships over the past 18 years and has established a set of strict formal methods for causal analysis. This method was initially developed from atmospheric prediction research, but has since been applied to multiple disciplines, such as quantum mechanics, neuroscience, financial economics, etc. 3. **Practical Application Cases**: The paper lists several practical application cases, demonstrating the effectiveness of quantitative causal analysis in solving practical problems. For example: - **Anthropogenic Causes of Global Warming**: Through causal analysis, it has been confirmed that carbon dioxide emissions are one of the main causes of global warming. - **Long - term Prediction of ENSO Modes**: By using causal analysis, the El Niño Modoki phenomenon has been successfully predicted, and the prediction time can even reach more than 10 years. - **Prediction of Extreme Drought in China**: In the 2022 China National Monthly Precipitation Prediction Competition, the extreme drought situation in the middle and lower reaches of the Yangtze River was successfully predicted using causal AI technology. 4. **Development of Causal AI**: The paper also discusses how to combine causal analysis with machine learning to construct more interpretable AI algorithms. For example, the information - flow - based fuzzy cognitive map (IF - FCM) proposed by Tyrovolas et al. can effectively remove spurious correlations in neural networks and enhance the interpretability and prediction ability of the model. In conclusion, this paper aims to solve the problems of interpretability and generalization ability of artificial intelligence models through quantitative causal analysis and demonstrate its practical application effects in multiple fields.

Quantitative causality, causality-guided scientific discovery, and causal machine learning

An Overview of the Quantitative Causality Analysis and Causal Graph Reconstruction Based on a Rigorous Formalism of Information Flow

Emergence and Causality in Complex Systems: A Survey on Causal Emergence and Related Quantitative Studies

Causal Deep Learning

Special Issue on Causal Discovery and Causality-Inspired Machine Learning

Causality Learning: A New Perspective for Interpretable Machine Learning

Causality between time series

Emergence and Causality in Complex Systems: A Survey of Causal Emergence and Related Quantitative Studies

Methods and tools for causal discovery and causal inference

Learning Causality and Causality-Related Learning: Some Recent Progress.

A Survey of Methods, Challenges and Perspectives in Causality

A Liang-Kleeman Causality Analysis based on Linear Inverse Modeling

A Survey of Learning Causality with Data

IEEE Transactions on Neural Networks and Learning Systems Special Issue on Causal Discovery and Causality-Inspired Machine Learning

Causal Machine Learning: A Survey and Open Problems

Visual Causality Analysis of Event Sequence Data.

Causality from Bottom to Top: A Survey

Normalized multivariate time series causality analysis and causal graph reconstruction

Causal reasoning in typical computer vision tasks

Deep Causal Learning: Representation, Discovery and Inference

Quantify the Causes of Causal Emergence: Critical Conditions of Uncertainty and Asymmetry in Causal Structure