Correlation Is Not Enough: Embedding Human Metadata for Individual Causal Discovery
Researchers have developed a method to improve the accuracy of causal discovery in Large Behavioural Models (LBMs) by addressing issues with embedding proximity. Standard biomedical language models incorrectly associate unrelated concepts, leading LBMs to infer false causal links. The proposed fix involves a contrastive learning approach using a knowledge graph to mine hard negatives, significantly improving the separation between related and unrelated concepts. This method also includes optimizations for faster inference using OpenVINO on Intel hardware. AI
IMPACT Enhances the reliability of AI models that infer causal relationships from user data, crucial for personalized applications.