Emergent Medical Data: Health Information Inferred by Artificial Intelligence
School of Law
“Artificial intelligence can infer health data from people’s behavior even when their behavior has no apparent connection to their health. AI can analyze social media to track the spread of infectious disease outbreaks, scrutinize retail purchases to identify pregnant customers, and track people’s movements to predict who might attempt suicide. These feats are possible because in modern societies, people continuously interact with internet-enabled devices in homes, workplaces, schools, and public spaces, and these devices are increasingly designed for surveillance. Smart phones track people’s whereabouts, wearables monitor their physical activity, smart speakers record their voices, and surveillance cameras observe their facial expressions. Continuous daily exposure to these devices produces millions of digital traces, the electronic remnants of people’s interactions with technology.
Digital traces provide insight into who we are, what we have done, and what we might do. However, in their raw form, they are rarely very interesting or useful; one’s retail purchases and internet browsing habits are relatively mundane pieces of information. Before scientists, corporations, and government agencies can profit from them, they must transform those traces to enhance their value. Transforming digital traces into health information is called mining for emergent medical data (EMD) because, through analysis with AI, the connections between digital traces and people’s health emerge unexpectedly, as if by magic.
This Article argues that EMD should be viewed as a new type of health information, distinct from traditional medical data (TMD), which is transmitted voluntarily from patients to healthcare providers. It describes how EMD-based profiling and predictions are increasingly promoted as solutions to public health problems such as the opioid crisis, rising rates of suicide, and the high prevalence of gun violence. However, there is little evidence to show that EMD-based profiling works. Even worse, it can cause significant harm, and current health privacy and data protection laws contain loopholes that allow public and private entities to mine EMD without people’s knowledge or consent.
After describing the EMD mining process, and the benefits and risks of EMD, the Article proposes six different ways of conceptualizing this emerging technology. It concludes with preliminary recommendations for effective regulation. Potential options include banning or restricting the collection of digital traces, regulating EMD mining algorithms and limiting which entities can use them, restricting how EMD can be used once it is produced, and requiring ethics board approval for EMD mining research.”
You can reach original article from the link below: