A new study published on the arXiv preprint server shows that using inaccurate descriptions of the human body may pose potential safety risks for individuals who do not fit into these body types when artificial intelligence is used in certain applications.

Assistant Professor and co-author of the study, Abigail Jacobs from the School of Information and the Center for Complex Systems at the University of Michigan, stated that these flawed assumptions define the standards of the human body and are incorporated into artificial intelligence through motion capture.

This research demonstrates how artificial intelligence plays a crucial role in the design, development, and implementation of motion capture systems, which can infer the movements of humans, animals, and objects in space.

These systems collect data using sensors or cameras, which are then modeled on computers to create a "digital skeleton" for video game animations, health diagnostics, or workplace ergonomics simulations.

Jacobs said, "These systems are used for applications such as designing safe manufacturing workshops, augmented reality, and autonomous vehicles. They rely on programmed and flawed assumptions, such as assumptions about whose body is 'standard' or 'representative'."

In the same study, Jacobs and her colleagues delved into historical practices from the 1930s, revealing a concerning trend of relying excessively on healthy adult males to represent "typical" bodies and movements.

Over time, these assumptions have been integrated into modern software, potentially causing harm in how motion capture systems represent bodies. This is similar to how color photography predominantly captures light-colored skin, resulting in poor representation of individuals with darker skin tones.

Jacobs stated, "Think about the practice of crash test dummies based on the bodies of standard male adults in history, which led to higher injury rates for women and children. Considering the scope of motion capture systems, individuals with bodies that deviate from the 'standard' may be at risk of harm."

Jacobs mentioned that this study establishes an analytical framework that can be applied to other technologies, focusing on how assumptions are built into hardware and artificial intelligence, how bodies are represented in AI systems, and how hidden assumptions (often outdated and unfounded) shape today's technology.