Dana-Farber Cancer Institute has recently introduced an artificial intelligence application called GPT4DFCI. This system has been designed for use in various aspects of a medical center or hospital, including operations, administration, and research. However, it is important to note that GPT4DFCI is not authorized for direct clinical care. The application operates within the digital confines of Dana-Farber, ensuring privacy, security, and compliance with HIPAA regulations. GPT4DFCI was introduced gradually over the course of the past year to an increasing number of users, with detailed instructions outlining the responsibilities of users to verify and validate the output generated by the system.
Upon surveying initial users of GPT4DFCI, it was found that the primary uses of the application included information extraction, searches, and answering general knowledge questions. Users also reported utilizing the tool for document summaries, research paper synopses, and drafting written materials such as letters, meeting minutes, and presentations. However, concerns were raised regarding the potential for inaccurate or false outputs, as well as ethical and policy compliance. It is recognized that while generative AI tools hold promise in healthcare, they also come with inherent risks of bias, inaccuracy, incompleteness, and potential misuse.
The decision to implement GPT4DFCI at Dana-Farber was driven by the belief that prohibiting generative AI tools entirely would hinder learning and innovation, both of which are crucial to the institution’s mission. To mitigate risks and foster discovery, a diverse Governance Committee comprising representatives from various sectors of Dana-Farber was established. This committee oversaw the technical, ethical, and policy considerations surrounding the implementation of GPT4DFCI. By taking a multidisciplinary approach to governance, Dana-Farber aimed to ensure that the implementation of AI technology was conducted responsibly and with the best interests of patients and stakeholders in mind.
As Dana-Farber continues to navigate the landscape of AI technology in healthcare, the institution is committed to sharing its experiences and technical insights with other healthcare organizations considering similar initiatives. By sharing knowledge and lessons learned, Dana-Farber hopes to contribute to the collective understanding of how AI applications can be effectively integrated into medical settings while upholding standards of ethics, security, and patient care.