Dr. Mona Sloane and Dr. Chirag Agarwal at the University of Virginia (UVA) are developing an LLM-based open-source AI chatbot to help researchers draft IRB protocols (IRB-SBS #7515). This project is a collaboration with the National Institute for Standards and Technology Research Protections Office (NIST RPO).
🤖 What does the chatbot do? It supports researchers as they work through questions asked as part of creating IRB protocols, engages them in principled reasoning about research ethics, and helps generate IRB protocol language as well as materials like consent forms, study information sheets, and recruitment language.
📄 How is the chatbot developed? To fine-tune a set of LLMs for the chatbot, the research team requires PIs to donate training data in the form of already approved IRB protocols and supplementary documents (active or expired).
🏛️ Where will the chatbot be hosted? Once the chatbot launches, it will be hosted and maintained by the National Institute for Standards and Technology’s (NIST) Research Protections Office. NIST’s Research Protections Office adheres to the same ethical principles as all IRB offices across the United States.
By donating your approved protocols, you help build an open-source tool that can make IRB protocol generation less time-consuming for empirical researchers everywhere.
If you would like to donate your IRB protocol(s) and supplementary documents (such as consent forms or study information sheets), please use this form to submit a PDF version of your previously approved protocol(s) and any supplementary document(s). Please ensure that you do not submit any IRB protocol(s) or supplementary documents that are considered "sensitive" studies (i.e. any research that is classified, contains proprietary information, etc.).
To ensure data privacy, all personally identifiable information (name, email, phone number, institution name, group leader, department chair, etc.) must be redacted from the documents prior to training. You may choose to redact this information yourself, or request that our research team does so for you. You will be asked for your preference during the donation process.
If you choose to not redact personally identifiable information from your documents, the research team will strip or mask any identifying information in the documents as well as strip any non-essential metadata from the donated materials, such as embedded geolocation information or file paths related to the donated documents. Once the donated materials are de-identified, they will be stored in a secure, IRB-compliant environment that is encrypted and password-protected, ensuring that no one but members of the research team have access to the donated materials.
Participants must be 18 or older, and all documents must be in English. The information that you give in the study will be handled confidentially per our approved IRB (IRB-SBS #7515).
