A team from the University of Chicago announced on Apr. 10 a new system that combines artificial intelligence and electrical muscle stimulation to help users perform complex, unfamiliar movements by physically guiding their muscles. The project, led by PhD students Yun Ho and Romain Nith under the supervision of associate professor Pedro Lopes, received the Best Paper Award at the ACM CHI 2026 conference.
The technology is designed to address situations where people need procedural knowledge—how to do something—rather than just factual information. Unlike previous electrical muscle stimulation (EMS) systems that were limited to pre-programmed tasks, this new approach uses advanced multimodal AI models to generate real-time movement guidance based on what users see and how they move.
The researchers describe previous EMS assistance as "highly-specialized… fixed, and non-contextual," meaning it could only help with specific actions anticipated by designers. In contrast, their "embodied AI" adapts guidance dynamically for each situation. According to the research team, this allows users to be guided through tasks such as opening child-proof pill bottles or operating unfamiliar devices without prior training.
During user studies cited in the paper, participants successfully completed various challenging physical tasks with support from dynamically generated muscle cues. When intentional errors were introduced by the AI for testing purposes, participants quickly noticed and adapted using their own intuition or by re-prompting the system. One participant said: "the body's intuitions help notice errors right away," highlighting an advantage over traditional instruction methods.
Lopes said: "This could be a game-changer, not only for tasks that are highly physical (such as learning physical skills required for working with manufacturing and materials, or learning musical instruments) but also in situations where users might be situationally impaired (e.g., multitasking and performing several gestures at once, or cannot see in the dark)."
Despite its promise, current limitations include electrode calibration challenges and discomfort associated with EMS use. Lopes acknowledged these issues: "While we are really excited about our system it is clearly just the first step... many needed improvements on AI reasoning... making EMS more comfortable and easy to wear... currently this is not something you can just wear on your everyday life." The team has open-sourced their code so others can build upon their work while emphasizing user control and safety throughout development.
