A newly published study presents a proof-of-concept for using deep brain stimulation controlled with artificial intelligence (AI) in a closed-loop system to enhance cognitive control, suggesting it might be effective for a number of mental illnesses. That’s a lot to unpack, so let’s go back to the beginning. The most fundamental necessary to understand what is going on here is that your brain is a machine. It’s a really complicated machine, but it’s a machine none-the-less, and we can alter the function of that machine by altering its physical state.

This may seem obvious, but actually people are generally psychologically biased against this view. This may, in fact, be a consequence of brain function itself, which evolved to create a seamless stream of consciousness, an illusion of self unaware of all the subconscious processes that make up brain function. This is why we tend to interpret people’s behavior in terms of personality and conscious choice, when in fact much of our behavior is a consequence of subconscious processes. We are also biased to believe that people can think or will-power their way out of mental illness.

The more we understand about how the brain functions, however, the more it becomes apparent that the brain is just a glitchy machine, and lots can go wrong. Even when functioning within healthy parameters, there are many trade-offs in brain function, with strengths often coming at the price of weaknesses. We need to look out for our own interests, for example, but this comes at the price of anxiety and paranoia. But there are some brain functions that are so basic they are almost universally useful, and impairment of them can cause of host of problems. One such basic brain function is called cognitive control, which is essentially the ability to determine what thoughts and actions will be the focus of your brain’s attention.

As an aside, the most basic brain functions tend to be the same ones that most people are unaware of. If non-experts were asked to list things the brain does, cognitive control (or something like it) would probably not be on most people’s list. This is something that happens in the background, or is so fundamental to our stream of consciousness that the illusion does not let us notice it.

What, then, is cognitive control? Wakefullness requires that the brain experiences a constant loop of activity. There are, in fact, centers in the brain stem whose function is to constantly activate the cortex, to keep that cognitive train constantly moving. Without these stimulating centers you would be in a coma, even with a perfectly functioning cortex. In addition, we need to be able to shift gears, so that we are not stuck endlessly in one loop. That is cognitive control – a higher brain function that can consciously direct or thoughts, attention, and behavior. This conscious function, however, is dependent on deep brain structures and subconscious functions. If the deeper pathways are not functioning optimally, we may be consciously stuck. This is what happens in many mental illness states – those with anxiety may be unable to stop thinking about what makes them anxious, or those with major depression may stay focused on negative thoughts. But what if we can give cognitive control a kick in the pants? That is the focus of the current study.

The researchers used patients undergoing surgery to treat their epilepsy as a convenient population, because the surgeons would be in there anyway. They inserted wires to stimulate a part of the brain known as the internal capsule and striatum (specifically the dorsal segments), which is a white matter pathway deep in the brain. They compared two conditions – one in which the internal capsule (this is a bilateral structure) was stimulated at random. In the other condition AI was used to interpret EEG signals and determine when the subject was cognitively stuck. For the purposes of the study subjects were given a conflict task, and the stimulation was given when the AI determined they were stuck in the task. In either group stimulation of the internal capsule improved performance on the task, but the AI closed-loop condition performed much better.

This was a proof-of-concept trial, and now can move on to full clinical trials to determine efficacy. So it’s too soon to conclude that this approach will work, but these are encouraging early results. The other good news is that this treatment requires only off-the-shelf existing technology. Deep brain stimulation is already an established treatment for several conditions, and this approach can use the same technology, combined with the AI control to time the stimulation. Obviously, there is a lot that can go wrong here in a real-life setting, and therefore I expect the clinical research and tweaking will take some time. A lot will depend on how good the AI becomes at detecting unhealthy states that are feeding anxiety, depression, drug-seeking behavior, or other conditions, and not overcalling healthy brain functions, like focusing on completing a task. This possible problem might be solved with a simple off switch. Or perhaps it will improve task completion. Again, we need clinical trials to know for sure.

Regardless of how this specific intervention works out, it does seem likely that we are at the beginning of a new technology, using AI in a closed loop system to monitor brain function and alter it with targeted stimulation. There is no theoretical limit to this technology, which will improve as our understanding of brain function improves and as the underlying technology improves. This is another aspect of the brain-machine interface, and we are on the steep part of this technological curve.