Why advanced bionic hands fail, and how AI is fixing the problem

AI-powered advanced bionic hands today can closely match natural human hands in terms of movement and design. Yet, many amputees stop using them. According to researchers at the University of Utah, up to 50 per cent of people with upper limb amputations abandon advanced bionic prostheses. The main reason, experts say, is not hardware but control.

Why many users abandon bionic hands

Jake George, an electrical and computer engineer at the University of Utah, said modern bionic hands are difficult to operate in daily life. “Our goal was to make bionic arms more intuitive, so users could perform tasks without having to think about every movement,” George said. Natural hands rely on automatic reflexes. For example, when an object starts slipping, sensors in the fingertips adjust grip strength in 60 to 80 milliseconds, without conscious effort.

Most commercial bionic hands lack this automatic response. Users must manually control grip force and finger position, often through:

Mobile apps with preset grip modes

Muscle signals captured using surface electromyography. Researchers said this places a heavy mental burden on users, making simple actions exhausting.

How the AI ‘co-pilot’ works

To address this, the research team developed an AI-powered bionic hand co-pilot.

According to the study: The team modified an existing bionic hand. Fingertips were replaced with pressure and proximity sensors. These sensors detect distance and grip force in real time. An AI controller processes this data and automatically adjusts each finger.

By repeatedly touching and releasing objects, the system learned how to grip items without crushing or dropping them.

“The hand naturally adapts to the object. Each finger moves independently,” George explained.

Shared control keeps humans in charge. Unlike earlier autonomous prosthetic systems, this AI does not take full control. The researchers designed it as a shared-control system, where:

The user remains in charge

The AI quietly assists in the background

Users can still tighten, loosen, or release objects at any time

“It’s not self-driving,” George said. “It helps without fighting the user.”

Success rate jumps to 90 per cent in tests

The AI-powered bionic hand was tested on amputees and non-amputees in laboratory conditions.

Participants were asked to:

Pick up a paper cup and drink

Move fragile objects like eggs

Results showed:

Without AI: success rate was 10-20 per cent

With AI assistance, the success rate rose to 80-90 per cent

The AI also reduced mental effort, allowing users to focus less on controlling the hand.

Still early, but promising

Researchers caution that the technology is still in testing. “This hand is not yet as natural as a biological limb,”

The next step is real-world testing in home environments. The team is also working on:

Internal electromyography

Neural interface technologies

Partnerships for larger clinical trials

“The goal is to combine AI, robotics, and neural interfaces into one device,” George said.