Asset Details
Do you wish to reserve the book?
Using Artificial Intelligence for Proxy Decision-Making
Hey, we have placed the reservation for you!
By the way, why not check out events that you can attend while you pick your title.
Oops! Something went wrong.
Looks like we were not able to place the reservation. Kindly try again later.
Are you sure you want to remove the book from the shelf?
Using Artificial Intelligence for Proxy Decision-Making
Do you wish to request the book?
Using Artificial Intelligence for Proxy Decision-Making
Please be aware that the book you have requested cannot be checked out. If you would like to checkout this book, you can reserve another copy
We have requested the book for you!
Your request is successful and it will be processed during the Library working hours. Please check the status of your request in My Requests.
Oops! Something went wrong.
Looks like we were not able to place your request. Kindly try again later.
Journal Article
Using Artificial Intelligence for Proxy Decision-Making
2025
Overview
Background: The literature alludes to several studies highlighting challenges with human proxy as decision makers such as emotional burden, physician barriers, decisional conflict, accuracy, and overconfidence. However, only a small subset reported on a proxy's congruency. This study expanded on a proof-of-concept that artificial intelligence (AI) can act as a proxy decision maker with value preferences and considered its ethical implications. Aim: To compare the congruency of AI as a proxy decision maker with human proxies on end-of-life treatment decisions. Methods: Utilizing LLaMa3, an AI Large Language Model as a proxy decision tool, we recruited 15 adults and their legal decision makers as dyads to complete a value and end-of-life preference surveys for a comparison analysis. We measured the participants' overall composite value scores and collected their end-of-life preferences to use in the AI congruence evaluation. Congruency percentage was taken over three clinical hypothetical scenarios and compared between the participant with either the human or AI proxy. Results: The mean congruency percentage between the participant and human proxy was 44.4% (95% CI: 23.6-65.3), n = 12. Fifty percent of dyads had one or no matching responses across the three scenarios and 16% had perfectly matched responses. After the model's adjustment for prompt engineering and parameter fine-tuning, the congruency with AI and value inputs was 72.2% (95% CI: 67.4-77.0) with 67.0% matched responses. The model performed the same as the human proxy without value preferences with congruency of 45.3% (95% CI: 36.2-54.4). Discussion: The AI model had a 28% higher congruency as a proxy decision-maker for end-of-life treatment decisions after the inclusion of value preferences. This approach has a promising utility as a supplemental tool for human decision-making and can protect self -determination if values are pre-recorded in the event of decisional incapacity.
Publisher
MedStar Washington Hospital Center
Subject
