SciSciGPT: Your AI Research Collaborator

SciSciGPT is an open-source, prototype AI research collaborator for the science of science (SciSci), developed by CSSI.

Rapid advances in data and computational tools have transformed the science of science. While the growing scale and complexity of datasets, coupled with the rapid evolution of analytical methods, have enabled new analyses, they have also made SciSci research more time-intensive and technically demanding.

Large language models (LLMs) offer new possibilities for human–AI collaboration. By leveraging LLM capabilities and integrating SciSci literature, datasets, and methodologies, we create SciSciGPT, a domain-grounded AI collaborator for the SciSci field. SciSciGPT’s multi-agent framework orchestrates diverse functionalities, including understanding user requests, breaking down tasks into concrete steps, retrieving relevant data, analyzing data, creating visualizations, comprehending the literature, evaluating its performance and making iterative improvements.

To learn more, read the full research article in Nature Computational Science (or read the research briefing here). You can also read a lay-person-friendly summary in Kellogg Insight, or watch the video below.