Aditya Agrawal's position paper identifies a factual bias in Retrieval-Augmented Generation systems, which optimize for epistemic uncertainty reduction while ignoring aleatoric uncertainty in opinion-rich content. A survey of 34 major RAG benchmarks reveals only one addresses opinion synthesis. The paper introduces Opinion-Aware RAG (O-RAG), an architecture using LLM-based opinion extraction and entity-linked metadata. Evaluated across e-commerce forums and hotel reviews, O-RAG achieved 18-48% reduction in Wasserstein distance, +26.8% sentiment diversity, and 79.2% human evaluator preference.
No score is assigned. Sources and their independence are shown in the citation chain below.