The Danger of Flawed Assumptions: How AI Can Harm Those Who Don't Fit the 'Normative' Body Type

Thursday - February 22 2024, 14:46 UTC - 9 months ago

The use of AI in motion capture systems can be potentially harmful for those who do not fit the 'normative' body type, as these systems are often based on flawed and outdated assumptions. This issue can be addressed by examining how these assumptions are built into technology and working towards more inclusive and diverse representations of human bodies.

When it comes to using artificial intelligence (AI) in various applications, the assumption that the human body fits a certain 'normative' type can have harmful consequences for those who do not fit that body type. This is the finding of a new study posted to the arXiv preprint server by researchers from the U-M School of Information and Center for the Study of Complex Systems.

The study focuses on the use of AI in motion capture systems, which are used to track and simulate the movement of people, animals, and objects. These systems collect data using sensors or cameras, which is then used to create 'digital skeletons' that can be used in activities ranging from video game animation to diagnosing health conditions.

Unfortunately, as the study points out, these systems often rely on flawed and stylized assumptions about what is considered 'typical' or 'representative' for a human body. This has been a longstanding issue, with the authors revealing that even early practices dating back to the 1930s relied heavily on healthy, adult men as the baseline for human bodies and movements. Even modern, state-of-the-art systems still rely on these outdated assumptions, with some even using the bodies of deceased men as stand-ins for live movement. This results in distorted and inaccurate representations of different body types and movements.

The consequences of these flawed assumptions can be serious, as the study points out. Similar to how early color photography only captured light skin tones, these flawed assumptions in AI can lead to harm for those who do not fit the 'normative' body type. This can be seen in cases such as the use of crash test dummies based on normative male bodies, which have been found to cause higher injury rates for women and children.

To address this issue, the researchers offer an analytical framework that can be applied to other technologies. This involves examining how assumptions are built into hardware and AI, how bodies are represented in AI systems, and how outdated and unfounded assumptions shape present-day technologies. By paying attention to these factors, we can work towards creating AI systems that are more inclusive and safe for all body types.