In the world of technology and artificial intelligence (AI), the lack of clear information labels has become a concern for consumers. Just as nutrition labels on food products help consumers make informed choices, there is a need for similar labels on health technology to foster trust and enable better decision-making. This is particularly relevant in light of recent controversies surrounding AI governance and the OpenAI board coup.
For over 80 years, the United States has mandated labels on food and drugs to ensure safety. Now, tech companies are realizing the importance of labels for products that straddle the line between consumer gadgets and medical devices, such as fitness trackers. These devices, along with their associated algorithms, share similarities with regulated medical products and can have potential "side effects" that generate inaccurate data, especially among underrepresented populations. Labels can play a crucial role in helping users navigate the complex risk-benefit decisions associated with these products.
To envision what these labels could look like, HumanFirst CEO and co-founder Andy Coravos collaborated with researchers from Duke, Sage Bionetworks, and Mount Sinai. They proposed elements that would evaluate the risks and benefits of using mobile, connected, sensor-derived digital health technology. Similar to nutrition labels, these labels could highlight relevant metrics such as heart rate, body temperature, and data security practices.
Studies have shown that readable labels instill consumer confidence and increase willingness to pay for products that are labeled with trust-based indicators. The importance of labels extends beyond health technology to include privacy and security in smart devices. Some proposals suggest labeling devices to provide users with a clearer understanding of privacy practices. Apple, for instance, has introduced privacy labels for apps in its store.
Efforts are underway to develop labeling systems that extend from algorithms to source data. Researchers at Duke, Google, Dartmouth, and the FDA have drafted Model Facts labels to assist doctors in using machine learning models for clinical decisions. Harvard and MIT researchers have collaborated on the Dataset Nutrition Label Project to measure the completeness and inclusiveness of datasets. Companies like Twilio are also launching AI Nutrition Facts label generators to offer transparency on data usage.
The potential applications for health tech labels are vast. Consumers could use labels to compare and choose devices based on their specific needs, while physicians could utilize labels to assess the risks and benefits of remote monitoring products. Additionally, labels could revolutionize clinical trials by collecting quality-of-life and performance outcomes using sensors, ultimately leading to a more convenient approach to drug discovery and development.
While education is necessary to empower consumer decisions, centralizing objective and unbiased information through labels is the first step towards building a more human-centric healthcare infrastructure. By drawing lessons from the food and drug industries, we can document and address algorithm bias and adverse events, ultimately increasing confidence in health tech and enabling consumers to make informed choices.