A new study by Rice University researchers explores large language models (LLMs) as a tool for streamlining responsible research practices.
Led by Melissa Cantú, a second-year doctoral student in bioengineering, the project tested 27 publicly available LLMs from seven companies, including OpenAI, Anthropic, Microsoft and xAI. Each model was tested for its ability to generate citation diversity reports, which serve as a means for academic authors to assess the breadth and transparency of their research citation lists.
Using a curated dataset of nearly 200 authors whose demographic information was self-reported, the study shows that several models produced reliable analyses of citation lists with nearly 99% accuracy, generating publication-ready reports in minutes.
Cantú, who is developing new methods to make advanced cell-based immunotherapies more accessible, views this work as part of a broader commitment to fostering research communities and practices that support a multiplicity of perspectives. Cantú has taken several leadership roles at Rice to support the Latine community in STEM, hoping to expand opportunities for future generations of researchers.
“Citation practices might seem like a small detail, but they can have a large-scale impact on whose contributions get recognized and built upon in a field, potentially leading to bias,” Cantú said. “It is important to reflect on whose voices are being amplified by our work via our citations because citation counts shape the visibility and career success of underrepresented and early career scientists. When ensuring citation balance can now take minutes with LLMs instead of hours with conventional tools, researchers can realistically integrate this reflection into their regular manuscript preparation process. In health care-related fields such as bioengineering, this can ultimately support more inclusive scientific discourse that better reflects the populations served by our engineering innovations.”
While the study highlights practical benefits of using artificial intelligence to reduce barriers in research software, the authors also note important safeguards. Automated demographic predictions can raise concerns about accuracy and privacy, and self-reported data remain the most reliable standard. Nonetheless, the findings, published in Nature Machine Intelligence, suggest that free, widely available tools can help embed reflection and accountability more seamlessly into everyday scientific practice.
- Peer-reviewed paper:
-
LLMs as all-in-one tools to easily generate publication-ready citation diversity reports | Nature Machine Intelligence | DOI: 10.1038/s42256-025-01101-y
Authors: Michael King and Melissa Cantú
https://doi.org/10.1038/s42256-025-01101-y