Causal Reasoning Research Articles

Human socio-technical evolution through the lens of an abstracted-wheel experiment: A critical look at a micro-society laboratory study.

Micro-society experimental setups are increasingly used to infer aspects of human behavioural evolution. A key part of human society today is our dependence on, and use of, technology-whether simple (such as a knife) or complex (such as the technology that underpins AI). Previously, two groups of researchers used an abstracted-wheel experiment to explore the evolution of human technical behaviour, reaching fundamentally different outcomes. Whereas one group saw their results as indicating social learning only (void of causal understanding), the other inferred non-social technical reasoning as part of human technical behaviour. Here we report on the third generation of the micro-society abstracted-wheel experiment. We argue that causal reasoning is inseparable from both social learning and technical reasoning, and that these traits probably co-evolved into the current human socio-technical niche. Based on our outcomes, we present a critical assessment of what this experiment may (or may not) reveal about the evolution of human technical behaviour. We show that the abstracted-wheel experiment reflects behavioural output only, instead of testing for cognition. It is therefore limited in its ability to inform on aspects of human cognitive evolution, but it can provide useful insights into the interrelatedness of social learning, technical reasoning, and causal reasoning. Such a co-evolutionary insight has the potential to inform on aspects of human socio-technical evolution throughout the Pleistocene.

A Model-Based Approach to the Problem of Variable Choice

Abstract This article distinguishes causal modeling, metaphysical, epistemic, and modeling reasons for variable/parameter choice. I argue in favor of justifying variable/parameter choices by appealing to modeling reasons concerning the limitations of the available measurements, experimental data, modeling techniques, and modeling frameworks. I use this “tyranny of availability” to identify normative criteria for variable/parameter choice that apply across most scientific modeling contexts and investigate their metaphysical and epistemological implications.

Cause and Effect: Can Large Language Models Truly Understand Causality?

With the rise of Large Language Models (LLMs), it has become crucial to understand their capabilities and limitations in deciphering and explaining the complex web of causal relationships that language entails. Current methods use either explicit or implicit causal reasoning, yet there is a strong need for a unified approach combining both to tackle a wide array of causal relationships more effectively. This research proposes a novel architecture called Context-Aware Reasoning Enhancement with Counterfactual Analysis (CARE-CA) to enhance causal reasoning and explainability. The proposed framework incorporates an explicit causal detection module with ConceptNet and counterfactual statements, as well as implicit causal detection through LLMs. Our framework goes one step further with a layer of counterfactual explanations to accentuate LLMs' understanding of causality. The knowledge from ConceptNet enhances the performance of multiple causal reasoning tasks such as causal discovery, causal identification, and counterfactual reasoning. The counterfactual sentences add explicit knowledge of `not caused by' scenarios. By combining these powerful modules, our model aims to provide a deeper understanding of causal relationships, enabling enhanced interpretability. Evaluation of benchmark datasets shows improved performance across all metrics, such as accuracy, precision, recall, and F1 scores. We also present CausalNet, a novel dataset specifically curated to benchmark and enhance the causal reasoning capabilities of LLMs. This dataset is accompanied by code designed to facilitate further research in this domain.

Causal Inference in Natural Language Processing: Application Status and Future Outlook

Abstract. Causal reasoning has received much attention in recent years. Unlike statistical learning that focuses on the correlation between variables, causal inference can analyze the causality between variables and avoid false causal relationships caused by confounding factors. The use of causal inference to help solve Natural Language Processing (NLP) problems has made great progress. Most existing studies have focused on exploring the application of causal relationships in downstream tasks of NLP, achieving good results. This article conducted a comprehensive survey, focusing on investigating the advantages of causal inference methods in solving specific NLP problems and applications, as well as their practical advantages compared to deep learning methods. Intended to provide researchers in the NLP field with more detailed perspectives and recommendations. Specifically, this article first introduces the relevant concepts of causal inference, including causal inference and related concepts of causal relationships, as well as the differences between causal inference and statistical machine learning; Then, existing research on the use of causal relationships in downstream applications was summarized, including bias removal, stock price prediction, document dialogue based, and continuous few shot learning. Finally, a summary was made on the development of future causal relationships.

From Generative AI to Objective-Driven Systems: A Paradigm Shift in Artificial Intelligence

Marwan Omar 1Illinois Institute of Technology Email: momar3@iit.edu Abstract The rapid advancement of Artificial Intelligence (AI), particularly in the domain of generative models, has led to impressive achievements in content creation and natural language processing. However, these models are inherently limited by their reliance on pattern recognition and lack of true understanding. In contrast, Objective-Driven AI offers a promising alternative by focusing on goal-oriented behavior, causal reasoning, and the development of world models. This paper explores the limitations of generative AI, highlighting its inability to grasp context, causality, and ethical considerations. It then presents the concept of Objective-Driven AI, emphasizing its potential to operate effectively in complex, real-world environments where understanding and reasoning are critical. The paper concludes with a discussion of future research directions, including advanced world modeling techniques, ethical AI, and robustness against adversarial attacks, which are essential for the further development of Objective-Driven AI systems. Keywords Objective-Driven AI, Generative AI, Causal Reasoning, World Modeling, Ethical AI, Artificial Intelligence, Adversarial Attacks, Machine Learning, Autonomous Systems

The Advantage of Board Game with Deep Reinforcement Learning and Causal Inference

With the continuous progress of artificial intelligence technology, the deep reinforcement learning method combining behaviourism and connectionism has shown fantastic performance far beyond the human level in chess games. This paper systematically introduces two mainstream chess games, including the standard algorithms of artificial intelligence technology in chess and Go, including Monte Carlo and deep reinforcement learning, and expounds on their algorithm principles and the core reasons for their strong learning ability. In addition, this paper discusses the limitations of existing methods in interpretability and puts forward the possibility of applying causal reasoning to deep reinforcement learning to solve interpretable problems. This study will provide a valuable reference for practitioners in related fields.

Models of Variability in Probabilistic Causal Judgments

AbstractMost theories of causal reasoning aim to explain the central tendency of causal judgments. However, experimental studies show that causal judgments are quite variable. In this article, we report the results of an experiment using a novel repeated measures design that demonstrate the existence of meaningful (i.e., not noise-related) within-participant variability. Next, we introduce and assess multiple computational cognitive models that serve as potential accounts of the sources of variability and fit those models to the new empirical data. We find that the Bayesian Mutation Sampler has the best fit to the data and is able to account for a number of unusual features of the response distributions (e.g., bi-modality), supporting the view that the stochastic sampling mechanism it posits reflects the cognitive processes via which people draw causal inferences. Additionally, our findings suggest that incorporating ‘non-reasoning’ processes, such as rounding and guessing, can improve the ability of models of causal reasoning to account for the observed response distributions. Overall, the study highlights the potential of computational modeling of full response distributions to shed light on the underlying mechanisms of human causal reasoning and identifies promising directions for future research.

Efficient Differential Diagnosis using Cost-aware Active Testing

The diagnosis of complex systems benefits greatly from a differential, multistep approach that narrows down the list of possible conditions or failures that share the same observable effects to a single root cause. We provide a suitable and practically applicable methodology for this. In extension to existing work, it covers all types of diagnostic actions, i.e., the observation of system properties, active testing, and system interventions like providing a dedicated diagnostic input or forcing the system into discriminating states, but also the replacement of components. Combining all these possible steps into one probabilistic and causal reasoning framework, we I) stepwise generate the diagnostic model systematically to correctly cover the interplay of observations and diagnostic interventions, and II) provide decision support based on counterfactuals for the selection of the next diagnostic step, countering the vast number of possible actions that arise in machine diagnostic processes, considering costs of actions to minimize the expected overall cost of the diagnosis.We developed and successfully tried our methodology for diagnosing cyber-physical systems in the high-tech industry, but we found that it supports more processes, such as computing intervention actions for autonomous robots.

Causal reasoning in Software Quality Assurance: A systematic review

Context:Software Quality Assurance (SQA) is a fundamental part of software engineering to ensure stakeholders that software products work as expected after release in operation. Machine Learning (ML) has proven to be able to boost SQA activities and contribute to the development of quality software systems. In this context, Causal Reasoning is gaining increasing interest as a methodology to go beyond a purely data-driven approach by exploiting the use of causality for more effective SQA strategies. Objective:Provide a broad and detailed overview of the use of causal reasoning for SQA activities, in order to support researchers to access this research field, identifying room for application, main challenges and research opportunities. Methods:A systematic review of the scientific literature on causal reasoning for SQA. The study has found, classified, and analyzed 86 articles, according to established guidelines for software engineering secondary studies. Results:Results highlight the primary areas within SQA where causal reasoning has been applied, the predominant methodologies used, and the level of maturity of the proposed solutions. Fault localization is the activity where causal reasoning is more exploited, especially in the web services/microservices domain, but other tasks like testing are rapidly gaining popularity. Both causal inference and causal discovery are exploited, with the Pearl’s graphical formulation of causality being preferred, likely due to its intuitiveness. Tools to favour their application are appearing at a fast pace – most of them after 2021. Conclusions:The findings show that causal reasoning is a valuable means for SQA tasks with respect to multiple quality attributes, especially during V&V, evolution and maintenance to ensure reliability, while it is not yet fully exploited for phases like requirements engineering and design. We give a picture of the current landscape, pointing out exciting possibilities for future research.

Agency at a distance: learning causal connections

AbstractIn a series of papers, we have argued that causal cognition has coevolved with the use of various tools. Animals use tools, but only as extensions of their own bodies, while humans use tools that act at a distance in space and time. This means that we must learn new types of causal mappings between causes and effects. The aim of this article is to account for what is required for such learning of causal relations. Following a proposal by Grush and Springle, we argue that learning of inverse mappings from effects to causes is central. Learning such mappings also involves constraints based on monotonicity, continuity and convexity. In order for causal thinking to extend beyond space and time, mental simulations are required that predict the effects of actions. More advanced forms of causal reasoning involve more complicated forms of simulations.

Causal Explanations and the Growth of Understanding

Causal explanations are central to cognitive development. Once considered beyond the grasp of young children, causal reasoning is now understood to be present in infants, driving curiosity and learning. The search for causal explanations underlying recurring patterns is evident in nonverbal forms. Causal explanations are diverse, varying in granularity, spanning different content domains, and encompassing both deterministic and probabilistic forms. These explanations support counterfactual reasoning, diagnostic thinking, and learning. The emergence of causal explanatory thought in infants is suggested by their responses to anomalies and by the development of exploratory behaviors. In the preschool years, children build upon early abilities to request, evaluate, and generate explanations. The decline of causal explanation-seeking in young schoolchildren can be dramatic and seems to arise from a convergence of social, cultural, and motivational factors. Overall, the development of causal explanations is essential to cognitive growth and mature understanding, with enduring influences on learning and reasoning.

Extending a rational process model of causal reasoning: Assessing Markov violations and explaining away with inhibitory causal relations.

Causes sometimes decrease rather increase the probability of an effect, as when drinking coffee lowers the probability of sleep or an aspirin eliminates a headache. This research tests how two causal reasoning errors that have influenced the development of theories of human causal reasoning manifest themselves in the presence of inhibitory causal relations. Past research with generative causal relations (a cause makes its effect more probable) has shown that people violate the Markov condition, the pattern of independence that should obtain among causally related variables. And it has shown that they explain away-the phenomenon in which one should lower likelihood of one event when another is discovered to have occurred (e.g., exonerating one murder suspect when evidence against another is found)-too little or not at all. The new empirical findings reported here reveal that both sorts of errors manifest themselves when inhibitory causal relations are present although, unexpectedly, the direction of those errors sometimes reverses. Only the mutation sampler, a rational process model of human causal reasoning, correctly predicted these novel empirical findings. These results support the view that causal reasoning errors can be understood as arising from rational inference constrained by limited cognitive resources. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Examining and Supporting Mechanistic Explanations Across Chemistry and Biology Courses.

Causal mechanistic reasoning is a thinking strategy that can help students explain complex phenomena using core ideas commonly emphasized in separate undergraduate courses, as it requires students to identify underlying entities, unpack their relevant properties and interactions, and link them to construct mechanistic explanations. As a crossdisciplinary group of biologists, chemists, and teacher educators, we designed a scaffolded set of tasks that require content knowledge from biology and chemistry to construct nested hierarchical mechanistic explanations that span three scales (molecular, macromolecular, and cellular). We examined student explanations across seven introductory and upper-level biology and chemistry courses to determine how the construction of mechanistic explanations varied across courses and the relationship between the construction of mechanistic explanations at different scales. We found non-, partial, and complete mechanistic explanations in all courses and at each scale. Complete mechanistic explanation construction was lowest in introductory chemistry, about the same across biology and organic chemistry, and highest in biochemistry. Across tasks, the construction of a mechanistic explanation at a smaller scale was associated with constructing a mechanistic explanation for larger scales; however, the use of molecular scale disciplinary resources was only associated with complete mechanistic explanations at the macromolecular, not cellular scale.

Affordances of Electrostatic Potential Maps in Promoting Use of Electronic Features and Causal Reasoning in Organic Chemistry

Affordances of Electrostatic Potential Maps in Promoting Use of Electronic Features and Causal Reasoning in Organic Chemistry

Modeling of the cascading impacts of drought and forest fire based on a Bayesian network

The cascading impact of disastrous events has become an important focus of disaster research. In recent years, forest fires have occurred frequently in China, causing huge economic losses. Studying the cascading impacts of drought and forest fire is of great significance for reducing disaster risks. Taking Yunnan Province of China as an example, meteorological data and forest fire point data from 2005 to 2018 were collected and statistically analyzed. A Bayesian network model of the cascading impacts of drought and forest fire was established, enabling the prior probability and conditional probability of nodes to be determined. Based on this information, a probability prediction was established using causal reasoning. Finally, in the case test, the Brier score was used to test the accuracy of the model. The Brier test value was 0.305, which was less than the qualified threshold of 0.6. The results indicated that the Bayesian network model established in this study had a good prediction performance, which was basically consistent with the facts. The results provide an insight into the mechanism by which drought induced forest fires occur and will be of use in forest fire prevention work.

The association between Vitamin D serum level and sleep quality among pregnant women in Asia

Background and purpose: Studies have investigated that Vitamin D serum level is associated with sleep quality and circadian rhythms in pregnant women in Asia, but the results remained controversial. This systematic review is conducted to explore the association between Vitamin D serum level and sleep quality among pregnant women population in Asia. Methods: We conducted systematic literature review based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Relevant studies that discussed the association between Vitamin D serum level and sleep quality was retrieved from Pubmed, EBSCO, and Proquest. The outcome variable was sleep-quality which measured with Pittsburg Sleep Quality Index (PSQI) questionnaires and the independent variable was the serum 25(OH)D levels. A total 2,285 articles were excluded, leaving 3 final articles to be analyzed. The risk of bias was assessed with the New-Castle Ottawa Quality Assessment Scale (NOS). Results: Three studies included in this review with a total of 1,359 pregnant women in Asia, ranging from 18 to 68 years old. All three studies were controlled for covariates. Out of three studies, two studies showed a significant result of the association between Vitamin D serum level with sleep quality with a p-value <0,05. Causal reasons remained unexplained considering the studies were completed in cross-sectional and cohort design. Conclusion: This study gives an overview of the role of Vitamin D in the sleep quality of pregnant women in Asia. Future research should focus on conducting more comprehensive studies with stringent criteria to further explore this association in diverse Asian populations

Disentangling User Cognitive Intent with Causal Reasoning for Knowledge-Enhanced Recommendation

The primary objective of an effective recommender system is to provide accurate, varied, and personalized recommendations that align with the user’s cognitive intents. Given their ability to represent structural and semantic information effectively, knowledge graphs (KGs) are increasingly being utilized to capture auxiliary information for recommendation systems. This trend is supported by the recent advancements in graph neural network (GNN)-based models for KG-aware recommendations. However, these models often struggle with issues such as insufficient user-item interactions and the misalignment of user intent weights during information propagation. Additionally, they face a popularity bias, which is exacerbated by the disproportionate influence of a small number of highly active users and the limited auxiliary information about items. This bias significantly curtails the effectiveness of the recommendations. To address this issue, we propose a Knowledge-Enhanced User Cognitive Intent Network (KeCAIN), which incorporates item category information to capture user intents with information aggregation and eliminate popularity bias based on causal reasoning in recommendation systems. Experiments on three real-world datasets show that KeCAIN outperforms state-of-the-art baselines.

Navigating causal reasoning in sustainability science

When reasoning about causes of sustainability problems and possible solutions, sustainability scientists rely on disciplinary-based understanding of cause–effect relations. These disciplinary assumptions enable and constrain how causal knowledge is generated, yet they are rarely made explicit. In a multidisciplinary field like sustainability science, lack of understanding differences in causal reasoning impedes our ability to address complex sustainability problems. To support navigating the diversity of causal reasoning, we articulate when and how during a research process researchers engage in causal reasoning and discuss four common ideas about causation that direct it. This articulation provides guidance for researchers to make their own assumptions and choices transparent and to interpret other researchers’ approaches. Understanding how causal claims are made and justified enables sustainability researchers to evaluate the diversity of causal claims, to build collaborations across disciplines, and to assess whether proposed solutions are suitable for a given problem.

FairScene: Learning unbiased object interactions for indoor scene synthesis

In this paper, we propose an unbiased graph neural network learning method called FairScene for indoor scene synthesis. Conventional methods directly apply graphical models to represent the correlation of objects for subsequent furniture insertion. However, due to the object category imbalance in dataset collection and complex object entanglement with implicit confounders, these methods usually generate significantly biased scenes. Moreover, the performance of these methods varies greatly for different indoor scenes. To address this, we propose a framework named FairScene which can fully exploit unbiased object interactions through causal reasoning, so that fair scene synthesis is achieved by calibrating the long-tailed category distribution and mitigating the confounder effects. Specifically, we remove the long-tailed object priors subtract the counterfactual prediction obtained from default input, and intervene in the input feature by cutting off the causal link to confounders based on the causal graph. Extensive experiments on the 3D-FRONT dataset show that our proposed method outperforms the state-of-the-art indoor scene generation methods and enhances vanilla models on a wide variety of vision tasks including scene completion and object recognition.

Diffusion-Based Causal Representation Learning.

Causal reasoning can be considered a cornerstone of intelligent systems. Having access to an underlying causal graph comes with the promise of cause-effect estimation and the identification of efficient and safe interventions. However, learning causal representations remains a major challenge, due to the complexity of many real-world systems. Previous works on causal representation learning have mostly focused on Variational Auto-Encoders (VAEs). These methods only provide representations from a point estimate, and they are less effective at handling high dimensions. To overcome these problems, we propose a Diffusion-based Causal Representation Learning (DCRL) framework which uses diffusion-based representations for causal discovery in the latent space. DCRL provides access to both single-dimensional and infinite-dimensional latent codes, which encode different levels of information. In a first proof of principle, we investigate the use of DCRL for causal representation learning in a weakly supervised setting. We further demonstrate experimentally that this approach performs comparably well in identifying the latent causal structure and causal variables.