From Static Repositories to Agentic Knowledge Webs: ResearchTwin and the S-Index for Federated Human-AI Research Discovery

The exponential growth of scientific literature, datasets, and code repositories has created a discovery bottleneck that impedes knowledge synthesis and reproducibility. Traditional dissemination formats – static PDFs, siloed code hosting, and fragmented data repositories – fail to represent the interconnected narrative of modern research, while conventional metrics such as the H-index neglect contributions from reusable code and shared datasets. We present ResearchTwin, an open-source federated platform that transforms a researcher’s scholarly output into a conversational digital twin, with a preliminary evaluation of its deployed prototype. The system uses a Bimodal Glial-Neural Optimization (BGNO) architecture comprising a Multi-Modal Connector Layer, a Glial Layer for caching and rate management, and a Neural Layer implementing Retrieval-Augmented Generation with a provider-agnostic LLM backend. We formalize the S-index, building on our earlier QIC framework, into a composite metric that extends FAIR principles – via a binary accessibility/licensing gate, field-normalized impact scoring, and geometric collaboration scaling – to quantify multimodal research impact. A case study comparing two researchers with similar H-indexes but substantially different S-indexes demonstrates that the metric captures dimensions of impact – particularly dataset and code contributions – invisible to citation-based measures alone. ResearchTwin exposes an inter-agentic discovery API using Schema.org typed responses and HATEOAS navigation, enabling AI agents to discover cross-lab synergies. A three-tier federated architecture preserves data sovereignty while enabling global discoverability.