- Research using natural-language processing tools in machine learning is helping us understand conscious awareness and how drugs alter it.
- Mental health conditions may be the price we humans pay for the huge evolutionary advantage of having a default mode network.
- Psychedelics may work by lifting the filters of sensory input essential to DMN processes.
By Danilo Bzdok, M.D., Ph.D., Ph.D.
We aren’t going to completely understand the brain anytime soon. It’s just too complicated! But we can take what we do know and piece it together in new and better ways, to make new deductions about how our brains work and how drugs may be able to help people with mental health conditions. That’s what we’re working on in my lab.
Today, we have unprecedented big-data repositories and capabilities, including data about the human brain and capabilities we’ve developed in the course of building increasingly complicated systems.
It may sound surprising, but we can use some of the approaches developed towards artificial intelligence (AI) to investigate human intelligence! In fact, reframing questions in medical science as machine-learning problems can help us to view these issues in new ways and bring new insight to bear.
For instance: Today, research is showing the potential for psychedelics to alleviate symptoms of mental health disorders. But how are they working, and why? This is an important question that’s not unique to psychedelics but is in fact central to much of mental health treatment. And it’s one we’re working to help answer.
The default mode network
The brain’s default mode network (DMN)—a system we’ve begun to identify and investigate only in the 21st century—seems to be extremely relevant to how drugs works. The DMN is our brain’s deepest neural processing layer. When we’re not in a task-focused mind state (for example, when we’re staring out of the window instead of reading), this brain network is active.
Why is the DMN so important?
The DMN is the greatest consumer of energy of any network in the brain. This is interesting, especially because the brain overall already uses a great deal of energy (20% of all of the energy of the whole body, despite being only about 2% of the body’s mass). So the DMN is one of the most “expensive” parts of one of a human’s most “expensive” parts.
Moreover, the brain areas involved in the DMN have increased the most in size, evolutionarily speaking, compared with other parts of our brain, and compared with the brains of our nearest relatives, monkeys.
A system that’s always on and uses a great deal of energy both on a day-to-day basis and on an evolutionary time scale sounds like something very important. Why does the DMN matter so much?
One answer—backed by empirical evidence from neuroscience experiments—is that it’s working to model probability. Put simply, the DMN may help humans anticipate the future. The DMN helps us see the world from different perspectives, to come to solutions, to have strategic insight into how to value information in our environment and make smart decisions in the future.
A frog can’t plan to hunt a bug two weeks from now. It’s just not possible, given the mental capacities that are enabled by its brain circuitry. It can only process and react to what’s in its current sensory environment.
We can do much more than that, and this is an enormous evolutionary advantage. The DMN may make this possible and may be central to what defines human thought, intelligent behavior, and consciousness.
What does the DMN have to do with mental health and mental health treatments, including psychedelics?
The DMN—probably the deepest layer of the brain, unique to humans, consuming a massive amount of energy, and active by default—seems to be at the crossroads of creativity and of many or most major psychiatric disorders. It’s possible that mental health conditions are the downside of this massive evolutionary advantage—and, perhaps, adjusting this system can help such conditions.
The prevailing understanding of psychedelics is that they have a close relationship with the DMN. To make any sense of the world, our brains must filter out the vast majority of sensory input. One of the ways psychedelics may work is by lifting those filters, which have been built up over time, and which affect how our DMN processes and interprets.
How does this AI-based brain research work and what is it showing?
Because there is so much data to work with, machine-learning tools and approaches are helping us to understand how different brain receptor pathways and neurotransmitters are related to different experiences that people have after taking psychedelics.
They’ve allowed us, for example, to combine three disparate sets of information for the first time. First, we have thousands of text narratives of people describing experiences with psychedelics. Next, we have data on the known receptor affinities of currently known drug compounds. And finally, we have data on brain receptors that shows us where different receptors are and how dense they are throughout the brain.
We’ve trained a machine-learning system, using data analysis and natural-language processing, to figure out the components of an experience with a psychedelic drug. We can connect a person’s experience to the places in the brain that drug affects and posit that certain parts of the brain, when acted upon by specific drugs, cause specific effects.
Science isn’t quite sure yet exactly how drugs affect changes in consciousness. We haven’t even converged on a definition of what consciousness is yet! But the more we investigate the mechanisms of the brain, and the mechanisms of drugs that affect the brain, we may soon be able to better predict which drugs will help which people, helping researchers, clinical trials, and countless patients and those who care about them.
Danilo Bzdok, M.D., Ph.D., Ph.D., is Associate Professor, Department of Biomedical Engineering, Faculty of Medicine, McGill University; Canadian Institute for Advanced Research (CIFAR) Artificial Intelligence Chair at the Mila-Quebec Artificial Intelligence Institute (Montreal). Dr. Bzdok’s research centers on narrowing knowledge gaps in the brain basis of human-defining types of thinking, with a special focus on the higher association cortex in health and disease.