View Of Causation Research Articles

A survey of out‐of‐distribution generalization for graph machine learning from a causal view

AbstractGraph machine learning (GML) has been successfully applied across a wide range of tasks. Nonetheless, GML faces significant challenges in generalizing over out‐of‐distribution (OOD) data, which raises concerns about its wider applicability. Recent advancements have underscored the crucial role of causality‐driven approaches in overcoming these generalization challenges. Distinct from traditional GML methods that primarily rely on statistical dependencies, causality‐focused strategies delve into the underlying causal mechanisms of data generation and model prediction, thus significantly improving the generalization of GML across different environments. This paper offers a thorough review of recent progress in causality‐involved GML generalization. We elucidate the fundamental concepts of employing causality to enhance graph model generalization and categorize the various approaches, providing detailed descriptions of their methodologies and the connections among them. Furthermore, we explore the incorporation of causality in other related important areas of trustworthy GML, such as explanation, fairness, and robustness. Concluding with a discussion on potential future research directions, this review seeks to articulate the continuing development and future potential of causality in enhancing the trustworthiness of GML.

A Causal View on Bias in Missing Data Imputation: The Impact of Evil Auxiliary Variables on Norming of Test Scores

Among the most important merits of modern missing data techniques such as multiple imputation (MI) and full-information maximum likelihood estimation is the possibility to include additional information about the missingness process via auxiliary variables. During the past decade, the choice of auxiliary variables has been investigated under a variety of different conditions and more recent research points to the potentially biasing effect of certain auxiliary variables, particularly colliders (Thoemmes & Rose, 2014). In this article, we further extend biasing mechanisms of certain auxiliary variables considered in previous research and thereby focus on their effects on individual diagnosis based on norming, in which the whole distribution of a variable is of interest rather than average coefficients (e.g., means). For this, we first provide the theoretical underpinnings of the mechanisms under study and then provide two focused simulations that (i) directly expand on the collider scenario in Thoemmes and Rose (2014, appendix A) by considering outcomes that are relevant to norming and (ii) extend the scenarios under consideration by instrumental variable mechanisms. We illustrate the bias mechanisms for two different norming approaches and exemplify the procedures by means of an empirical example. We end by discussing limitations and implications of our research.

Identifying roles of formulas in inconsistency under Priest's minimally inconsistent logic of paradox

It has been increasingly recognized that identifying roles of formulas of a knowledge base in the inconsistency of that base can help us better look inside the inconsistency. However, there are few approaches to identifying such roles of formulas from a perspective of models in some paraconsistent logic, one of typical tools used to characterize inconsistency in semantics. In this paper, we characterize the role of each formula in the inconsistency arising in a knowledge base from informational as well as causal aspects in the framework of Priest's minimally inconsistent logic of paradox. At first, we identify the causal responsibility of a formula for the inconsistency based on the counterfactual dependence of the inconsistency on the formula under some contingency in semantics. Then we incorporate the change on semantic information in the framework of causal responsibility to develop the informational responsibility of a formula for the inconsistency to capture the contribution made by the formula for the inconsistent information. This incorporation makes the informational responsibility interpretable from the point of view of causality, and capable of catching the role of a formula in inconsistent information concisely. In addition, we propose notions of naive and quasi naive responsibilities as two auxiliaries to describe special relations between inconsistency and formulas in semantic sense. Some intuitive and interesting properties of the two kinds of responsibilities are also discussed.

CIS2N: Causal independence and sparse shift network for rotating machinery fault diagnosis in unseen domains

Intelligent fault diagnosis (IFD) based on deep learning (DL) has demonstrated its powerful performance to promote the reliability and safe operation of rotating machinery. In industrial applications, working condition changes frequently, leading to models’ performance degeneration due to out-of-distribution (OOD) samples, such as vibration signals with unseen rotating speeds, loads, etc. To address this issue, domain generalization fault diagnosis (DGFD) has attracted increasing attention. However, mainstream DGFD methods are implemented from statical dependence, which fail to excavate intrinsic causal mechanisms and have limited performance and reliability. From a causal view, the causal independence and sparse shift network (CIS2N) is proposed to diagnose via causal mechanisms. Specially, a structural causal model (SCM) representing causal relations of features in IFD is given. Based on the SCM, it is derived that features are jointly independent and feature deviation between intervened sample pairs is sparse. The dependence among features is measured via the covariance matrices constructed by high-level features mapped by random Fourier features (RFF). Meanwhile the L1-distance between pairs of features with the same causal features (intervened pair) is calculated and optimized to realize shift sparsity. The effectiveness of the proposed CIS2N on DGFD is demonstrated on two real helical gearbox datasets.

Acting as causing change

AbstractThe paper defends a version of the view that agency is a causal power, the “causing view.” After sketching the view, and explaining how it differs from its rivals, various challenges are assessed. A family of objections says that causing change is neither necessary nor sufficient for acting. The second challenge centers on an Aristotelian thesis about the relation between an action (A's opening a window) and the corresponding passion (the window's being opened by A). The final objection concerns the dynamic nature of acting: the claim is that a causal view of agency cannot accommodate actions “in progress” or mere activity. I conclude that none of the objections examined presents unsurmountable problems for the causal view of agency which, at least in the version here defended, remains a highly plausible and attractive view.

Causality-aware social recommender system with network homophily informed multi-treatment confounders

Typical recommender systems utilize observed ratings of users as inputs to learn their preferences and aim to output recommendations of new items that users will like by predicting their potential ratings. The real world is driven by causality given that whether or not a user is made exposed to the item may significantly affect the rating. While some recent methods have taken a causal view to mitigate the confounding bias in observed rating data, few have recognized recommendation as a multiple causal inference problem that concerns numerous items simultaneously in practice. By framing the recommendation as a multiple causal inference problem, this paper develops a causality-aware social recommender that incorporates underlying social network structures with matrix factorization methods for deconfounding in networked observational data to enhance social recommendations. Leveraging social network information, which inherently confounds with the preferences of users, is particularly crucial but not trivial for inferring unobserved confounders in social recommender systems. Considering that connected users in a social network share similar attributes, we propose to incorporate the network homophily into the matrix factorization models through regularization to better learn the latent variables. The extracted latent variables then capture the network informed multi-treatment confounders to mitigate the confounding bias and improve the rating prediction accuracy. The models are estimated through a proximal gradient-based optimization framework, which not only eases the incorporation of the network structure constraints but also improves the computational efficiency of the proposed method. Simulation and case studies are conducted to validate the proposed method.

Debiased Graph Neural Networks With Agnostic Label Selection Bias.

Most existing graph neural networks (GNNs) are proposed without considering the selection bias in data, i.e., the inconsistent distribution between the training set with the test set. In reality, the test data are not even available during the training process, making selection bias agnostic. Training GNNs with biased selected nodes leads to significant parameter estimation bias and greatly impacts the generalization ability on test nodes. In this article, we first present an experimental investigation, which clearly shows that the selection bias drastically hinders the generalization ability of GNNs, and theoretically proves that the selection bias will cause the biased estimation on GNN parameters. Then to remove the bias in GNN estimation, we propose a novel debiased GNNs (DGNN) with a differentiated decorrelation regularizer. The differentiated decorrelation regularizer estimates a sample weight for each labeled node such that the spurious correlation of learned embeddings could be eliminated. We analyze the regularizer in causal view and it motivates us to differentiate the weights of the variables based on their contribution to the confounding bias. Then, these sample weights are used for reweighting GNNs to eliminate the estimation bias, and thus, help to improve the stability of prediction on unknown test nodes. Comprehensive experiments are conducted on several challenging graph datasets with two kinds of label selection biases. The results well verify that our proposed model outperforms the state-of-the-art methods and DGNN is a flexible framework to enhance existing GNNs.

Quantum Antitrust – A Unified Exclusionary Abuse Theory

Now that the judgments in Servizio Elettrico Nazionale and Unilever have made sense of previous case-law, a unified analytical framework can be predicated for all exclusionary abuses. It is made up of two limbs (artificiality/conduct deviating from competition on the merits and potential exclusionary effect/capability of foreclosing). The two limbs are two sides of the same coin, which consists of ascertaining whether the plausible rationale (in the sense of nature and economy) behind a dominant company’s conduct is to derive an advantage that equally efficient competitors cannot either derive by doing the same (first limb–artificiality) or offset by other means (second limb–potential effect). This exclusionary rationale, which is a cognitive state in the dominant company’s mind rather than an ontological reality, is objectivised by holistically judging the dominant company’s conduct against the backdrop of the relative efficiency of its competitors, which is the link between both limbs: if the inherent features of the company’s conduct and all the relevant circumstances surrounding it allow equally efficient competitors to either derive a similar advantage or to defeat its exclusionary effect by some other means, the conduct does not make anticompetitive sense (exclusion is not its plausible rationale). Owing to the very Darwinian nature of competition law, relative efficiency is the common “quantum” and, therefore, a useful yardstick for abuse from the historical, legal, teleological and practical perspectives, as well as from the points of view of causality and reality. Finally, plausibility would be the single standard of proof that could be rebutted by disproving either of the two limbs (in an Intel II-like fashion), either by providing an alternative non-exclusionary explanation that breaks the casual link or by putting forward an objective justification. All other tests and standards merely reflect the extent to which either the artificiality or the potential effects of the conduct can be presumed based on economic judgement or experience (as happened in merger control following CK Telecoms)–thus reconciling per se rules, consisting of an Art. 101-like cursory analysis, with a more economic approach.

Enhancing Out-of-distribution Generalization on Graphs via Causal Attention Learning

In graph classification, attention- and pooling-based graph neural networks (GNNs) predominate to extract salient features from the input graph and support the prediction. They mostly follow the paradigm of “learning to attend,” which maximizes the mutual information between the attended graph and the ground-truth label. However, this paradigm causes GNN classifiers to indiscriminately absorb all statistical correlations between input features and labels in the training data without distinguishing the causal and noncausal effects of features. Rather than emphasizing causal features, the attended graphs tend to rely on noncausal features as shortcuts to predictions. These shortcut features may easily change outside the training distribution, thereby leading to poor generalization for GNN classifiers. In this article, we take a causal view on GNN modeling. Under our causal assumption, the shortcut feature serves as a confounder between the causal feature and prediction. It misleads the classifier into learning spurious correlations that facilitate prediction in in-distribution (ID) test evaluation while causing significant performance drop in out-of-distribution (OOD) test data. To address this issue, we employ the backdoor adjustment from causal theory—combining each causal feature with various shortcut features, to identify causal patterns and mitigate the confounding effect. Specifically, we employ attention modules to estimate the causal and shortcut features of the input graph. Then, a memory bank collects the estimated shortcut features, enhancing the diversity of shortcut features for combination. Simultaneously, we apply the prototype strategy to improve the consistency of intra-class causal features. We term our method as CAL+, which can promote stable relationships between causal estimation and prediction, regardless of distribution changes. Extensive experiments on synthetic and real-world OOD benchmarks demonstrate our method’s effectiveness in improving OOD generalization. Our codes are released at https://github.com/shuyao-wang/CAL-plus .

Causality-Inspired Invariant Representation Learning for Text-Based Person Retrieval

Text-based Person Retrieval (TPR) aims to retrieve relevant images of specific pedestrians based on the given textual query. The mainstream approaches primarily leverage pretrained deep neural networks to learn the mapping of visual and textual modalities into a common latent space for cross-modality matching. Despite their remarkable achievements, existing efforts mainly focus on learning the statistical cross-modality correlation found in training data, other than the intrinsic causal correlation. As a result, they often struggle to retrieve accurately in the face of environmental changes such as illumination, pose, and occlusion, or when encountering images with similar attributes. In this regard, we pioneer the observation of TPR from a causal view. Specifically, we assume that each image is composed of a mixture of causal factors (which are semantically consistent with text descriptions) and non-causal factors (retrieval-irrelevant, e.g., background), and only the former can lead to reliable retrieval judgments. Our goal is to extract text-critical robust visual representation (i.e., causal factors) and establish domain invariant cross-modality correlations for accurate and reliable retrieval. However, causal/non-causal factors are unobserved, so we emphasize that ideal causal factors that can simulate causal scenes should satisfy two basic principles:1） Independence: being independent of non-causal factors, and 2）Sufficiency: being causally sufficient for TPR across different environments. Building on that, we propose an Invariant Representation Learning method for TPR (IRLT), that enforces the visual representations to satisfy the two aforementioned critical properties. Extensive experiments on three datasets clearly demonstrate the advantages of IRLT over leading baselines in terms of accuracy and generalization.

Identification of Necessary Semantic Undertakers in the Causal View for Image-Text Matching

Image-text matching bridges vision and language, which is a fundamental task in multimodal intelligence. Its key challenge lies in how to capture visual-semantic relevance. Fine-grained semantic interactions come from fragment alignments between image regions and text words. However, not all fragments contribute to image-text relevance, and many existing methods are devoted to mining the vital ones to measure the relevance accurately. How well image and text relate depends on the degree of semantic sharing between them. Treating the degree as an effect and fragments as its possible causes, we define those indispensable causes for the generation of the degree as necessary undertakers, i.e., if any of them did not occur, the relevance would be no longer valid. In this paper, we revisit image-text matching in the causal view and uncover inherent causal properties of relevance generation. Then we propose a novel theoretical prototype for estimating the probability-of-necessity of fragments, PN_f, for the degree of semantic sharing by means of causal inference, and further design a Necessary Undertaker Identification Framework (NUIF) for image-text matching, which explicitly formalizes the fragment's contribution to image-text relevance by modeling PN_f in two ways. Extensive experiments show our method achieves state-of-the-art on benchmarks Flickr30K and MSCOCO.

Revisiting the role of values in evidence-based education

Abstract Evidence-based practice in education involves basing decisions about what to do on evidence about the relative effectiveness of available interventions (e.g. programmes, products, practices). This article considers two influential critiques of evidence-based education (EBE) pertaining to its treatment of values. The ‘general critique’ condemns EBE for excluding values from decisions about what to do in education. The ‘specific critique’ condemns EBE for relying on a deterministic view of causality in education which disregards the complex, value-laden nature of educational contexts. I argue that virtually all versions of EBE escape the general critique, including the dominant intervention-centred approach that relies on experimental research to discover ‘what works’, because the predictions EBE aims to support are only one premise in a broader normative argument. Further, intervention-centred EBE can avoid much of the specific values-based critique because it is consistent with a probabilistic, rather than deterministic, understanding of causality. However, I argue that only a context-centred approach to EBE that relies on evidence about the specific target setting from local sources in addition to evidence from theory and mixed methods research can fully address the specific critique by accommodating critics’ descriptive claims about the nature of educational contexts.

Generalizing Graph Neural Networks on Out-of-Distribution Graphs.

Graph Neural Networks (GNNs) are proposed without considering the agnostic distribution shifts between training graphs and testing graphs, inducing the degeneration of the generalization ability of GNNs in Out-Of-Distribution (OOD) settings. The fundamental reason for such degeneration is that most GNNs are developed based on the I.I.D hypothesis. In such a setting, GNNs tend to exploit subtle statistical correlations existing in the training set for predictions, even though it is a spurious correlation. This learning mechanism inherits from the common characteristics of machine learning approaches. However, such spurious correlations may change in the wild testing environments, leading to the failure of GNNs. Therefore, eliminating the impact of spurious correlations is crucial for stable GNN models. To this end, in this paper, we argue that the spurious correlation exists among subgraph-level units and analyze the degeneration of GNN in causal view. Based on the causal view analysis, we propose a general causal representation framework for stable GNN, called StableGNN. The main idea of this framework is to extract high-level representations from raw graph data first and resort to the distinguishing ability of causal inference to help the model get rid of spurious correlations. Particularly, to extract meaningful high-level representations, we exploit a differentiable graph pooling layer to extract subgraph-based representations by an end-to-end manner. Furthermore, inspired by the confounder balancing techniques from causal inference, based on the learned high-level representations, we propose a causal variable distinguishing regularizer to correct the biased training distribution by learning a set of sample weights. Hence, GNNs would concentrate more on the true connection between discriminative substructures and labels. Extensive experiments are conducted on both synthetic datasets with various distribution shift degrees and eight real-world OOD graph datasets. The results well verify that the proposed model StableGNN not only outperforms the state-of-the-arts but also provides a flexible framework to enhance existing GNNs. In addition, the interpretability experiments validate that StableGNN could leverage causal structures for predictions.

Cross Modality Bias in Visual Question Answering: A Causal View with Possible Worlds VQA.

To increase the generalization capability of VQA systems, many recent studies have tried to de-bias spurious language or vision associations that shortcut the question or image to the answer. Despite these efforts, the literature fails to address the confounding effect of vision and language simultaneously. As a result, when they reduce bias learned from one modality, they usually increase bias from the other. In this paper, we first model a confounding effect that causes language and vision bias simultaneously, then propose a counterfactual inference to remove the influence of this effect. The model trained in this strategy can concurrently and efficiently reduce vision and language bias. To the best of our knowledge, this is the first work to reduce biases resulting from confounding effects of vision and language in VQA, leveraging causal explain-away relations. We accompany our method with an explain-away strategy, pushing the accuracy of the questions with numerical answers results compared to existing methods that have been an open problem. The proposed method outperforms the state-of-the-art methods in VQA-CP v2 datasets. R2: Providing brief insights into the experimental setup and results would add valuable context for readers. In response to R2, we released the code and documentation for the implementation as follows. Our codes are available at https://github.com/ali-vosoughi/PW-VQA.

Some Remarks on Hume, Reichenbach, and Einstein’s views on Temporal and Causal Priority in Causation with Objections from Quantum Realm

Stunning new observations and theories deepen the concept of causation. Quantum theory is one of the most striking discovery of contemporary physics, and it challenges many facets of older theories, even though it lies on many assumptions that could not be empirically experimented. In this paper, I will recall the views of Hume, Reichenbach and Einstein about the issue of causation, whereas I will focus on causal priority and temporal priority criteria in causation. Nevertheless, one task that I perform is the interpretation of these ideas by arguing them within the quantum scope. That is, trying to manifest Hume, Reichenbach and Einstein views of causation that I think as close, but I shall avoid putting straight cuts on where they differ. Direction of causal processes is dependent on the direction of time in Hume’s sense while Reichenbach explained the direction of time with irreversible processes, even if he follows the temporal priority criteria of Hume. It is stated that there is nothing faster than light by Einstein, as a result, a cause could not pass the upper limit of light. Crudely put, their thoughts all differ from the principles of quantum, because causation (causation might denote a different meaning in quantum realm) may be able to travel faster than light and the common cause principle between events is rejected in quantum realm.

Military spending crowding out health and education spending: which views are valid in Egypt?

This study examines the relationship between government spending, specifically military spending, government spending on health, government spending on education, and economic growth in Egypt over the period from 1980 to 2021. The paper utilizes a Granger causality test to detect the directional relationship between spending components and GDP growth. Furthermore, an autoregressive distributed lag (ARDL) model, including error-correction models, was established to determine the long- and short-term relationships among these variables. The study contrasts Wagner’s versus Keynes’s views of the government spending relationship with economic growth, with a greater emphasis on Keynes’s argument for military spending enhancing economic growth. While previous studies have investigated the relationship between aggregate spending and economic growth from a single or two-way direction of causality, the present study contributes to the literature by exploring the possibility of military spending crowding out spending on health and education and detecting this causality and dynamic relationship. The empirical results support the Keynesian view of causality from all government spending components to economic growth. However, short- and long-term analyses revealed a negative relationship between military spending and economic growth. Moreover, the short-term impact of government spending on education and health on economic growth is negative, but positive in the long-term. Finally, the causality test revealed that military spending influences health and education spending. Additionally, a unidirectional relationship exists between military spending and health expenditure, which requires further research. The policy implication of this study shows that although spending is exogenous to economic growth, it does not align with the Keynesian view of inducing growth. Instead, spending has negative current and future implications for economic development.

How does “over-hype” lead to public misconceptions about autonomous vehicles? A new insight applying causal inference

Traffic accidents caused by drivers’ over-reliance on SAE L2 advanced driver assistance systems (ADAS) have become a new type of accident worthy of attention. There is growing concern that the “over-hype” about automation technology produced by manufacturers or the media might affect the formation of public misconceptions about autonomous vehicles (AVs), further influencing their unsafe interaction behaviors with the system. The purpose of our study was to explore whether and how “over-hype” in various contexts affects public misconceptions about AVs from the view of causality. We conducted a randomized controlled trial with a control group of 396 participants and a treatment group of 545 participants drawn from Zhenjiang, China. We applied Double Machine Learning (DML), ordered logit models, and significant difference tests to estimate the impact of “over-hype” in various contexts. The results found that the statement that “over-hype” caused public misconceptions was less rigorous. Significant causality varies among contexts of “over-hype” and aspects of misconceptions. Specifically, “over-hype” in fatal accident reports will lead to availability misconception about AVs, but not safety misconception, suggesting that participants believed AVs were available nowadays but not safe as human drivers. “Over-hype” that exaggerates ADAS capabilities has the greatest average treatment effect on public misconceptions, especially the safety misconception. We found that the young and the more educated are more likely to have safety and availability misconceptions under the “over-hype” of some contexts. However, the effects of “over-hype” varied less among groups with different socio-economic factors on the whole, which revealed the universality of the influence of “over-hype.” The results provided empirical evidence for regulating propaganda about AV technology and provided practical insights on how to introduce AV technology to the general public properly.

Abu Hamid al-Ghazali in Our Times: How al-Ghazali is Interpreted and Transmitted in the Contemporary Turkish Context

Whether Abu Hamid al-Ghazali contributed to the decline of rationalist and scientific tradition in the Muslim world is a grand debate in Islamic studies. However, the literature on this question ventures only to construct a historical debate. Though al-Ghazali died in 1111, his opinions still influence Muslims today. So this research, rather than contribute to that historical debate, studies how al-Ghazali is interpreted in contemporary Turkey, and what this inter­pretation reveals about approaches to natural law, knowledge and philosophy (essentially, the study of conceptual lucidity and the valid procedures of reaso­ning). These three disciplines are chosen because al-Ghazali is accused of contributing to the decline of the scientific tradition in the Islamic world beca­use of his criticism of philosophy, his occasionalist view of causality, and his incorporation of inner knowledge into Sunnism. The research aims to answer ques­tions like: ‘Are al-Ghazali’s works referenced today to justify a hostile stance on philosophy?’ To achieve this goal, the research studies how al-Ghazali is interpreted by two Islamic social movements in Turkey: Işıkçılar and Erenköy. Studying these cases, the paper concludes that how al-Ghazali is interpreted and transmitted in Turkey contributes to (i) a highly sceptical stance on natural law, (ii) a deep suspicion of rational knowledge and the consequent belief in inner knowledge, and (iii) a highly critical stance on philosophy.

Instrumental Variable-Driven Domain Generalization with Unobserved Confounders

Domain generalization (DG) aims to learn from multiple source domains a model that can generalize well on unseen target domains. Existing DG methods mainly learn the representations with invariant marginal distribution of the input features, however, the invariance of the conditional distribution of the labels given the input features is more essential for unknown domain prediction. Meanwhile, the existing of unobserved confounders which affect the input features and labels simultaneously cause spurious correlation and hinder the learning of the invariant relationship contained in the conditional distribution. Interestingly, with a causal view on the data generating process, we find that the input features of one domain are valid instrumental variables for other domains. Inspired by this finding, we propose an instrumental variable-driven DG method (IV-DG) by removing the bias of the unobserved confounders with two-stage learning. In the first stage, it learns the conditional distribution of the input features of one domain given input features of another domain. In the second stage, it estimates the relationship by predicting labels with the learned conditional distribution. Theoretical analyses and simulation experiments show that it accurately captures the invariant relationship. Extensive experiments on real-world datasets demonstrate that IV-DG method yields state-of-the-art results.

Acquiring a new understanding of illness and agency: a narrative study of recovering from chronic fatigue syndrome

ABSTRACT Background The condition known as chronic fatigue syndrome or myalgic encephalomyelitis (CFS/ME) is poorly understood. Simplified medical models tend to neglect the complexity of illness, contributing to a terrain of uncertainty, dilemmas and predicaments. However, despite pessimistic pictures of no cure and poor prognosis, some patients recover. Purpose This study’s purpose is to provide insight into people’s experiences of suffering and recovery from very severe CFS/ME and illuminate understanding of how and why changes became possible. Methods Fourteen former patients were interviewed about their experiences of returning to health. A narrative analysis was undertaken to explore participants’ experiences and understandings. We present the result through one participant’s story. Results The analysis yielded a common plotline with a distinct turning point. Participants went through a profound narrative shift, change in mindset and subsequent long-time work to actively pursue their own healing. Their narrative understandings of being helpless victims of disease were replaced by a more complex view of causality and illness and a new sense of self-agency developed. Discussion We discuss the illness narratives in relation to the disease model and its shortcomings, the different voices dominating the stories at different times in a clinically, conceptually, and emotionally challenging area.