Partner: Kate MacDuffie

SSEC Engineers: Niki Burggraff, Apoorva Sheera, and Ying-Hsiang Huang

Research Goals and Domain

The Bioethics Team from the University of Washington Institute of Translational Health Sciences (ITHS) consults researchers and personnel on how to ethically engage with research participants and human subjects who inquire about their research studies. Researchers who conduct human subjects research often face the challenge of whether and how to share study results with participants. It is crucial that participants can learn the results of the research they participate in, especially in an accessible format and a language that they can understand. Currently this process lacks guidance and relies on the efforts of individual researchers.  

Software Problem

The ITHS Bioethics Team aims to change this by developing a set of tools for investigators that enables them to upload a PDF of an academic paper and have a system generate lay summary of research findings. This artificial intelligence will translate clinical research documents into accessible explanations for participants, while simultaneously protecting sensitive research data. The resulting Large Language Model (LLM) needs to be capable of answering participants’ questions in a way that is customizable across research studies and avoids hallucinations.  

Software Solution

SSEC is developing a Retrieval Augmented Generation (RAG) pipeline that doesn’t require fine-tuning of models and focuses on reducing hallucinations. This will effectively leverage sources in the research documents database for the interface to cite while generating responses for the participant. It will be hosted in Microsoft Azure. 

Impact

With this system, researchers across Washington, Wyoming, Alaska, Montana, and Idaho will be able to communicate results to potentially thousands of participants with ease.