AI Nutrition Labels, a New Tool Measuring AI's Impact on Human Wellbeing (www.media.mit.edu)

A team of researchers has launched "AI Nutrition Labels," a pioneering tool designed to measure the impact of artificial intelligence on human wellbeing. Developed as part of the MIT Media Lab's Advancing Humans with AI program and in collaboration with various academic institutions, this initiative addresses a significant gap in how AI technologies are evaluated. Instead of focusing solely on what AI can do, this framework, known as ImpactBench, assesses how AI influences human psychology and behavior, ultimately helping users make informed choices about technology that supports their wellbeing. ImpactBench evaluates AI models using 800 metrics tied to key areas such as user autonomy and psychological safety, distilling complex data into an accessible format similar to traditional nutrition labels. Early tests indicated that while AI systems showed promise in minimizing harm, they struggled to actively promote human flourishing, with a notable deficiency in supporting users' learning and agency. This research is particularly timely as it aligns with ongoing regulatory efforts, such as California’s Senate Bill 243, highlighting the urgency for independent data to guide effective regulation of AI technologies and their societal implications.