Unif-AI: A Structured Human-Led Method for AI-Assisted Conceptual Discovery

Scientific knowledge is increasingly distributed across specialized and rapidly expanding literatures, making conceptual integration a growing bottleneck in discovery-oriented research. Unif-AI is a structured, human-led methodological framework for AI-assisted conceptual discovery designed to improve the transparency, inspectability, and reproducibility of conceptual scientific reasoning. Rather than treating AI as an autonomous discoverer, the framework positions AI as a cognitive amplifier for scale-dependent tasks such as bibliographic mapping, semantic comparison, consistency checking, and adversarial stress testing, while preserving human responsibility for problem definition, hypothesis generation, boundary declaration, and interpretation. The workflow formalizes staged conceptual reasoning, identifies common failure modes in AI-assisted integrative work, and provides a practical scaffold for structured human-led reasoning across fragmented scientific domains.