EEG-Informed Machine Learning Framework for Causal Inference in News Reading

Zhangyi Tian^*

• The University of Manchester, Manchester, United Kingdom

The analysis and application of EEG signals in causal cognition during news reading represent a cutting-edge intersection of neuroscience and computational linguistics. Current advances emphasize the importance of understanding causal reasoning in complex and dynamic contexts, such as news media, to address challenges in semantic ambiguity, temporal dependency, and knowledge integration. Traditional methods often lack the ability to fully capture contextual and relational nuances, resulting in limited accuracy and interpretability in cognitive modeling. To overcome these limitations, this study introduces a Semantic Contextual Graph Network (SCGN) integrated with Dynamic Knowledge-Augmented Reasoning (DKAR). SCGN employs graphbased representations to model semantic dependencies in news text while incorporating temporal encodings for dynamic insights. Simultaneously, DKAR enriches the framework with external knowledge sources to dynamically align and infer causal relationships, leveraging multi-step reasoning for enhanced understanding. The combined approach significantly improves tasks such as entity recognition, relationship extraction, and causal reasoning, achieving state-of-the-art performance in experimental evaluations. The proposed methodology not only advances the computational modeling of causal cognition but also provides robust applications in real-world scenarios, including personalized content analysis and cognitive monitoring via EEG signals.