Abstract
I have been an emergency medical technician (EMT) at the Tuckahoe Volunteer Rescue Squad in Richmond, Virginia, for nearly five years. Looking for research projects for my senior capstone, I came across the UVA Dependable Systems and Analytics lab, which, since 2017, has had a project called “Cognitive Assistant Systems for Emergency Response.” It is focused on developing artificial intelligence (AI) devices for EMTs to assist their decision-making, a perfect intersection between my EMT hobby and my computer science work. My STS topic then considered the ethics of EMTs using these devices with regard to the doctrine of informed consent in medicine.
AI tools EMTs might use have two components: the device itself and the software that runs on it. The technical portion of my thesis produced a simulator for our lab to test the software portion of my lab’s project. In 2024, the lab developed a smart glasses prototype EMTs can wear, equipped with a camera and microphone, and the software on the glasses produces feedback shown through a heads-up display while a user wears the glasses. It was necessary to do a simulation in real life using the glasses in order to test the software component, making testing time- and resource-intensive. The simulator I created allows pre-recorded video, audio, and smart watch accelerometer data to be streamed in a synchronized manner such that it is presented to the software component as if it were being collected in real-time. This innovation allows our lab to test and improve the software component (improvement in this case meaning greater accuracy in the feedback) much more efficiently.
In my STS research, I considered how these AI tools would be integrated in real emergency medical services (EMS), and whether using one should require informed consent (involving explanations and consent forms) from the patient. Firstly, I said that the device recording audio of conversations is already legal in one-party-consent states like Virginia. For the video, I compared an EMT’s authority to document as a part of a 911 public service to that of a police officer, who is not required to get consent from his subjects for his body-worn camera. This is also assuming the AI tool is an EMS agency-issued device that is compliant with HIPAA for its data storage. The other consideration is an EMT using the AI tool as clinical decision support. I compared an EMT’s clinical reasoning to that of a doctor, who normally would not disclose every aspect of his decision-making process. AI tools can be viewed as simply another form of a resource to consult when treating a patient. They should aid, not replace, clinical reasoning. So long as a doctor does not use AI to artificially augment his abilities by attempting to practice in an area in which he was not trained, in which case a patient would be taking on additional risk, there is no issue requiring informed consent. Similarly, only trained EMTs would use this technology, meaning that would be no artificial augmentation of abilities. For these reasons, my STS research paper claims that AI tools should not require informed consent.
I like to think of my project as the absolute ideal of STS and the engineering capstone program. The whole purpose of STS is to get engineering students to think critically about the work they do and its ethical consequences for the real world. Indeed, I chose my technical project first, and, instead of doing an STS paper on something unrelated or only tangential, I became exceptionally passionate about my project and chose to engage with the ethics of my technical work directly. Throughout the year, I realized how technical thinking must inform ethical thinking and vice-versa. For instance, the correctness of a clinical decision support agent is going to help determine whether it is ethical to use it since it could lead to more clinical mistakes than without it. Moreover, there is no good reason to spend time developing a technical means if we know its use would be unethical and/or illegal; we would make better use of resources elsewhere in solving a given problem.
Thank you for the mentorship I received this year from Professors William Davis and Homa Alemzadeh in my capstone work. Thank you especially to Keshara Weerasinghe, a graduate student in Professor Alemzadeh’s laboratory, who met with me weekly for mentorship with my technical work. The “Cognitive Assistant Systems for Emergency Response” project itself is funded by the Commonwealth Cyber Initiative - Central Virginia Node (CCI-CVN).
