A new study published in npj Digital Medicine provides evidence that a personalized, artificial intelligence–powered brain stimulation system can enhance sustained attention in people completing cognitive tasks at home. The system uses non-invasive electrical brain stimulation and an adaptive algorithm to tailor stimulation to each person’s baseline attention level and head size. The results indicate that this personalized approach was especially effective for individuals who initially had lower attention performance, suggesting potential for future use in both everyday settings and clinical populations.
Maintaining attention over long periods is essential for many real-world tasks—whether it’s staying alert while driving, focusing on schoolwork, or remaining engaged during meetings. When attention falters, the consequences can range from missed learning opportunities to dangerous accidents. And for people with neurological or psychiatric conditions like attention-deficit/hyperactivity disorder, depression, or brain injury, the stakes are even higher.
Despite this, existing methods to improve attention, such as cognitive training or medications, have limitations. Brain stimulation, especially methods involving low electrical currents like transcranial electrical stimulation, has shown some promise in boosting cognitive performance. But most studies use fixed stimulation protocols that do not account for differences between individuals. Results have been inconsistent, likely because a single dose or method does not work equally well for everyone.
Additionally, most previous studies have been conducted in controlled laboratory settings, which reduces their relevance to real-life environments. In this context, the research team aimed to address two key challenges: how to personalize brain stimulation for each person, and how to deliver it effectively in a real-world setting such as a participant’s home.
“Most of our research is lab-based and doesn’t always translate outside the lab, where we hope it will work. We have noticed, along with other researchers, that not everyone benefits from the stimulation, so how can we improve that? There’s also a broader societal challenge in maintaining attention at a good level,” said study author Roi Cohen Kadosh, head of the School of Psychology and professor of cognitive neuroscience at the University of Surrey.
To develop a personalized system, the researchers created a framework that uses artificial intelligence to adapt stimulation settings for each individual. The system centers on a technique called transcranial random noise stimulation. Unlike more intense or invasive methods, this technique uses painless, low-intensity electrical currents delivered through electrodes placed on the scalp.
The researchers trained a machine learning algorithm using what’s called personalized Bayesian optimization. This method adjusts the intensity of the stimulation by analyzing each person’s baseline attention score and head circumference—a factor that affects how much current reaches the brain. As more people used the system, the algorithm refined its recommendations, improving over time.
The researchers tested the system across three experiments. To measure sustained attention, the researchers used a task developed by the US Air Force Research Laboratory. Participants acted as air traffic controllers monitoring aircraft on a screen. They had to press a button for “safe” aircraft paths and withhold their response for “critical” paths that could lead to collisions. Only about 11 percent of the trials were critical, mimicking the low-frequency but high-risk situations faced in real monitoring jobs.
The main performance metric was a score called A-prime, which reflects the participant’s ability to distinguish safe from unsafe trials. This score became the target the AI system tried to optimize by adjusting stimulation intensity. An improvement in this score after stimulation indicated better sustained attention.
In the first experiment, 103 participants used the neurostimulation device at home while completing the air traffic control simulation. Participants measured their own head size, completed a baseline attention task, and then received stimulation adjusted by the algorithm. The system identified an inverted U-shaped pattern in the data—suggesting that there is an optimal intensity for stimulation, and going above or below that level tends to worsen performance. The optimal dose also varied based on head size and baseline attention levels.
The findings provide evidence that “we can improve people’s attention outside of the lab using personalized neurostimulation to maximize the benefit,” Cohen Kadosh told PsyPost.
The second experiment used simulated data to compare the personalized approach against other optimization methods, including random selection and non-personalized algorithms. The personalized method outperformed both alternatives, although its advantage narrowed as noise in the data increased. This suggests that the AI system is more effective when the input data are reliable, which has implications for future deployment in uncontrolled environments.
In the third experiment, the researchers directly compared the personalized stimulation method to a fixed-intensity stimulation and a sham condition. This phase included 37 participants who completed three sessions, each with a different type of stimulation.
Only those with lower initial attention scores showed significant improvement from the personalized approach. Those who already performed well at baseline did not see further gains. This pattern “fits with the idea that there is a limit to how much we can improve cognition,” Cohen Kadosh said.
Participants did not report differences in side effects across conditions, and stimulation intensity did not predict discomfort or adverse outcomes. This suggests the method is safe for at-home use, at least in healthy adults.
While the study suggests that personalized brain stimulation can enhance attention in people with lower baseline performance, several limitations remain. The device and algorithm were tested only in healthy young adults, so it is unclear whether the same benefits would occur in people with attention-related conditions such as ADHD or long COVID. The researchers say that future work will need to extend to clinical populations.
Some participants also experienced technical issues, including device malfunctions and task crashes, leading to data exclusions. In particular, the remote nature of the experiment posed challenges for ensuring consistent engagement, as participants could lose motivation during repetitive tasks. The researchers note that about 27 percent of the sessions had to be excluded due to highly variable baseline performance, likely reflecting fluctuating motivation or attention.
Another limitation is that the algorithm was developed to optimize overall performance rather than track moment-to-moment fluctuations in attention. While this makes the system easier to use and more scalable, it may not capture the full complexity of how attention waxes and wanes in real time.
The study also raises broader ethical questions about the future use of cognitive enhancement technologies. The researchers point out that the personalized nature of their system might help reduce disparities, since it appears to work best for individuals starting from a lower performance baseline. Unlike some neuroenhancement tools that might widen performance gaps, this system could help narrow them. Still, questions about access, data privacy, and long-term safety will need to be addressed as the technology advances.
The study, “Personalized home based neurostimulation via AI optimization augments sustained attention,” was authored by Roi Cohen Kadosh, Delia Ciobotaru, Malin I. Karstens, and Vu Nguyen.
