New research published in journal neuron We provide evidence that giving people immediate, real-time feedback about their success during a motor task significantly improves their ability to control machines. This real-time enhancement tends to be particularly useful when users have limited visual or physical sensations to guide their movements. This study suggests a promising strategy to enhance rehabilitation techniques and improve functional use of advanced prosthetics in stroke patients.
Scientists and engineers are actively developing clinical interventions to help individuals restore or replace lost physical capacity. These interventions often rely on human-machine interfaces. Human-machine interfaces are systems that allow humans to interact with and control computers and robotic devices, such as virtual reality platforms used in stroke rehabilitation and robotic prosthetic arms for amputees.
To operate these technologies, users must generate a specific sequence of physical movements to achieve a goal, such as reaching to grab a glass of water. A major challenge in this field is that users often perform these actions with limited sensory feedback. When humans use robotic hands, they cannot feel the natural sensation of touch, and stroke patients often experience neurological deficits that impair vision and the physical sense of their own body movement.
Sensory instability makes it difficult to control devices smoothly and accurately. To address this issue, scientists have attempted to add artificial sensory feedback to these technologies. Even with artificial feedback, sensory information remains incomplete due to sensor noise and small delays in data transmission.
One proposed solution is to use reinforcement learning mechanisms. Reinforcement learning is a process by which people or systems learn how to make better decisions by receiving feedback about their successes or failures. In everyday life, this might be similar to getting a high score in a video game or hearing a soothing chime when you complete a digital task. It lets you know directly whether the goal has been achieved and has motivational value built-in.
In most previous exercise studies, reinforcing feedback was only given at the end of the task. This endpoint feedback can be confusing during complex and continuous actions. When a person fails in a multi-step movement, a single failure signal at the end does not explain that a particular part of the movement was wrong.
Pierre Vassiliadis, a researcher at University College London, conducted this study while studying at the Federal Institute of Technology in Lausanne, Switzerland, and wanted to investigate how timing affects this learning process. “We were interested in a simple question: In many motor tasks, people receive feedback only after the movement is finished, but the actual movement unfolds continuously,” Vasiliadis said.
“We thought that by emitting an explicit success signal in real time when someone is moving, we might be able to control the human-machine interface more effectively, especially in situations where normal sensory feedback is limited, such as in rehabilitation or prosthetic limb control,” Vasiliadis said. The researchers proposed that continuous delivery of these success and failure signals would provide better guidance.
To test this idea, the researchers designed a series of five experiments involving a total of 106 participants. In the first three experiments, healthy young adults completed a continuous tracking task. Participants grasped a special handgrip device to control a digital cursor on a computer screen and attempted to adjust their grip strength to hold the cursor inside a moving target for exactly 7 seconds.
The scientists manipulated visual feedback by hiding the cursor for varying amounts of time. In the fully visualized state, the cursor was always visible. In the low visual acuity condition, the cursor was visible for only about 35% of the trials.
To test real-time reinforcement, the color of the target on the screen changed based on the participant’s performance. The target always turns green to indicate success and red to indicate failure. This color change occurred in real time and was customized for each participant.
The software program constantly calculated the user’s average error over the past few attempts, and the user had to consistently outperform their recent performance to see a green success signal. Control trials featured targets flashing random, uninformative colors.
The first experiment involved 24 healthy adults and showed that reducing visual feedback significantly impaired their ability to control the cursor. The introduction of real-time reinforcement improved participants’ performance under all conditions. The benefit was significant in the low vision condition and relatively small in the full vision condition.
The researchers also measured skill retention and found that real-time reinforcement helped participants maintain their motor skill gains, especially when trained with less visual feedback. “Very simple signals, here performance-based real-time success cues, can help people learn how to better control their devices,” Vasiliadis says.
“We found this to be particularly useful when visual or somatosensory feedback is reduced, which is important because many patients and assistive technologies are operating under exactly such conditions,” Vasiliadis told PsyPost. “In other words, small, inexpensive changes to the feedback design can make human-machine interfaces easier to use and train.”
In a second experiment, we asked the same 24 participants to continue practicing the task with more varying levels of visual feedback to ensure that the improved performance was not simply due to participants reaching the absolute limits of their physical abilities. The third experiment involved an entirely new group of 24 healthy participants. This other group replicated the original study’s findings and confirmed that real-time reinforcement reliably improves motor control when visual information is lacking.
The fourth experiment investigated whether these benefits extended to other types of technology and other physical sensations. A new group of 40 healthy participants controlled a screen cursor using electrical signals generated by their muscles. The researchers attached surface electrodes to participants’ biceps, and they had to tense the muscles without moving their arms to control the cursor.
Participants did not move their arms completely, which resulted in a lack of a sense of natural body movement. To compensate for this lost sensation, the researchers used a small motorized device pressed into the participants’ palms to provide artificial tactile feedback. The motorized device applied physical pressure that matched the amount of muscle force produced by the participants.
The scientists then selectively reduced the visual feedback on the screen and the physical pressure on the hands. The researchers found that reducing either visual or tactile feedback worsened participants’ ability to control muscle interfaces. Adding real-time color changes to indicate success or failure significantly reduced these errors.
Real-time reinforcement improved performance in all conditions with limited sensory feedback. This suggests that this strategy works for different types of machine interfaces.
In a fifth experiment, the scientists tested the strategy on a clinical population. They recruited 18 older adults who had suffered a stroke that caused long-term disability. Patients completed the original handgrip task using their impaired hand, and researchers individualized the physical difficulty of the task to each patient’s specific abilities.
Similar to healthy participants, stroke patients also showed significant improvements in real-time motor control when receiving reinforcement in low visual acuity conditions. Surprisingly, real-time enhancement actually decreased patients’ performance when they had full vision of the cursor. The researchers suspect that flashing success and failure colors may act as a distracting information overload for patients with brain lesions when complete visual information is already available.
Unlike healthy young adults, stroke patients did not show sustained retention of motor skills after the training session ended. When the reinforcing color was removed, performance returned to baseline levels. This memory deficit may be related to age-related learning disabilities or to certain brain changes caused by the stroke itself.
To understand how real-time reinforcement changes human behavior, researchers analyzed physical fluctuations in participants’ movements. When humans practice physical skills, we tend to adjust our movements based on what has happened before. When a behavior fails, people usually try something else. This increases behavioral variability in a process known as exploration.
Once an action is successful, people tend to repeat the exact same action, which reduces physical variation in a process known as exploitation. The data revealed that under conditions of low visual feedback, real-time reinforcement caused participants to strongly leverage successful actions. They firmly established a winning strategy and repeated it accurately, but in full vision reinforcement led them to explore more often after failure.
“Our information-theoretic analysis suggests that the main effect of real-time reinforcement is not to cause further exploration after a failure, but to stabilize the motor commands that were just working, allowing them to more effectively exploit successes,” Vasiliadis said.
“We also found that this success-related stabilization was associated with subsequent learning, indicating a specific mechanism by which reinforcement improves motor skill acquisition,” Vasiliadis said.
Although this study provides evidence for a new training strategy, it has several limitations. Training sessions were incredibly short, with some conditions having fewer than 20 trials. This short period likely explains why stroke patients were unable to retain their new skills after completing their practice, and means future studies should test whether longer training sessions spread over multiple days result in lasting physical improvements.
Another limitation involves how the researchers manipulated physical and visual sensations. While intermittently turning off a screen cursor or a robot’s touch sensors creates a predictable environment for scientific experiments, real-world sensory loss caused by neurological damage or stroke is far more confusing. Scientists need to test real-time enhancement in more unpredictable and natural environments to confirm clinical utility.
The study, “Real-time enhancement of human-machine interface control,” was authored by Pierre Vassiliadis, Daniel Leal Pinheiro, Lisa Fleury, Silvestro Micera, Solaiman Shokur, and Friedhelm C. Hummel.
