Finding the Perfect Pace: How Speed Shapes AI Prosthetic Arm Acceptance

Prosthetic arm speed

Artificial‑intelligence‑powered prosthetic limbs are entering everyday life, but their success hinges on how naturally users experience them. Researchers created a virtual‑reality illusion where participants saw their own arm replaced by a robotic prosthesis, then examined how different movement speeds affected feelings of ownership, control, usability, and social perception.

The experiment revealed a clear trend: arms that moved either too fast or too slow felt detached and were judged as harder to use. In contrast, a moderate, human‑like speed—about one second to reach a target—produced the strongest sense that the prosthetic was part of the body.

From Direct Control to Intelligent Assistance

For people who have lost a hand or forearm, prosthetic devices are essential for daily tasks. Most current research focuses on translating bio‑signals such as EMG or EEG into precise movements. Meanwhile, advances in machine learning are enabling prostheses that can act autonomously, anticipating the user’s needs and moving on their own. That autonomy, however, can feel unsettling if the limb behaves in a way that feels foreign.

Virtual Reality Test of Speed and Embodiment

Building on prior work that links goal clarity with comfort, Harin Manujaya Hapuarachchi and colleagues asked whether the speed of an autonomous prosthetic arm influences acceptance. In a VR setup, participants watched an avatar whose left forearm was swapped for a robotic one. They completed a reaching task while the virtual arm moved toward a target at six different speeds, ranging from 125 ms to 4 s. After each trial, participants rated body ownership, sense of agency, usability (via SUS), and their impressions of the robot using the RoSAS scale (competence, warmth, discomfort).

The One‑Second Sweet Spot

  • At a 1‑second movement duration, scores for ownership, agency, and usability peaked.
  • Both the quickest (125 ms) and the slowest (4 s) conditions caused a sharp drop in those ratings.
  • Perceived competence was highest at moderate‑to‑fast speeds, while discomfort spiked at the fastest speed. Warmth showed no clear pattern.

In short, simply accelerating a prosthetic does not improve user experience. Matching the timing of natural human motion is far more important for making the device feel like an extension of the body.

Designing Human‑Like Robotic Augmentation

These findings suggest that future AI‑enhanced prostheses should prioritize human‑compatible timing over raw speed. Designers may need to tune movement profiles so they align with the brain’s expectations for a natural limb.

The lesson extends beyond arm prostheses. Supernumerary limbs, exoskeletons, and wearable robots could all benefit from motion that mirrors everyday human rhythm.

Long‑term studies are also planned. As users become accustomed to a tool, they often begin to incorporate it into their body schema. Over time, even a fast, high‑performance robotic limb might start to feel “normal” and fully embodied.

Virtual reality remains a powerful testbed for these investigations. It lets scientists evaluate new prosthetic concepts, control strategies, and psychological responses in a safe, repeatable environment before the technology reaches the market.

This work was supported by JSPS KAKENHI (JP22KK0158), the Murata Science and Education Foundation, JST (JPMJFS121), and MEXT (202334Z302).