Healthcare | Michael Halassa | Psychiatry

Why I Tell My Patients “I Don’t Know”

michaelhalassa — Mon, 16 Mar 2026 22:41:02 +0000

https://michaelhalassa.substack.com/p/why-i-tell-my-patients-i-dont-know

Medical training teaches us to project confidence, offer quick diagnoses, and provide clear explanations. Patients come seeking answers, and there’s real pressure to have them ready.

But psychiatry lives in uncertainty. We’re trying to understand how our most complex organ system goes awry using tools that are still remarkably crude. We lack the brain equivalent of a chest X-ray, an EKG, or blood tests for kidney and liver function.

So I find myself saying “I don’t know” fairly often. And when I do, something interesting happens: patients engage more deeply in their own care, and they feel less burdened by false expectations.

Doctor, What’s my Diagnosis?

Psychiatric diagnosis is messy. The DSM looks authoritative, but anyone who uses it clinically knows how artificial those boundaries can be. For example, ruling out manic symptoms in what appears to be a mood disorder may not be straightforward because for many reasons including simple misunderstanding of what is being asked.

So… in certain situations, when patients ask “What exactly is my diagnosis?” I’ve learned to say: “I don’t know for certain right now, but here’s what I’m considering and why.”

Then I walk them through my reasoning. I explain that some psychiatric diagnoses require pattern recognition over time, and that cross-sectional snapshots can be misleading. This honesty opens up space for collaboration, and we become partners trying to figure out what’s happening together.

Are you sure this is the right medication for me?

That same openness applies when we discuss medication management.

It’s a question I hear often: “Will this medication work for me?” Sometimes it’s helpful to cite studies and response rates. But in many cases, saying “I don’t know if this is the right medication for you, but I have a plan if it’s not” sets the right tone.

It shifts the conversation. Instead of setting up false expectations, we’re starting a collaborative process. The patient knows we’re learning together, which makes them more likely to give honest feedback about effects and side effects.

When a medication doesn’t work, there’s no sense of failure or broken promises. There’s just information. Valuable information that helps us understand their particular brain’s coalition of systems and move toward something more effective.

What is causing this?

The hardest question is often the simplest: “What’s going on with my brain?”

Some clinicians launch into explanations about neurotransmitters and chemical imbalances; what medical training teaches us to say. But, as I’ve written before, that level of explanation rarely connects to behavior or lived experience.

Instead, I use the coalition framework. the brain is a coalition of semi-autonomous systems, like apps on a smartphone or subcommittees in a parliament. I might say: “I don’t know exactly what’s happening, but here’s how I think about brains, and let’s figure out together which systems might be struggling.”

Then I explain that their brain is like a coalition of different systems, each optimized for specific tasks. There’s a system that tracks rewards and motivates behavior. There’s a system that builds predictions about the world and tries to minimize uncertainty. There’s an executive system that coordinates between all the others.

This gives us a framework to understand symptoms without pretending to know more than we do. Someone experiencing depression might have a reward system that’s become pessimistic about future outcomes, while their prediction system remains stuck on negative expectations. Someone with anxiety might have an uncertainty-monitoring system that’s become hypersensitive.

While not the definitive explanations patients may be used to hearing, they’re working hypotheses that connect to experience and suggest intervention strategies.

This applies to patients who may ask about “mechanism”, and I’ve learned to be honest about the limits of our knowledge while still offering useful frameworks.

The Therapeutic Power of Uncertainty

Something counterintuitive happens when you admit uncertainty: patients relax. The pressure to have all the answers immediately disappears. Instead of feeling like failures when initial treatments don’t work perfectly, they become curious collaborators.

“I don’t know yet,” followed by “but here’s how we’ll figure it out,” creates a different kind of therapeutic relationship; one built on genuine partnership, not expert authority.

I am absolutely not advocating for abandoning expertise. I still bring my training, knowledge of research, and clinical experience to every interaction. But I hold that expertise as hypotheses to be tested rather than certainties to be proclaimed.

Making Space for Learning

The coalition framework helps here too. I can explain that just as their brain is learning and adapting, our understanding of their specific brain is also evolving. Each medication trial, each therapy session, each mood tracking entry gives us more information about how their particular systems respond.

Sometimes the reward system responds quickly to interventions. Sometimes the prediction system needs more time and different approaches. Sometimes the executive system needs strengthening before other interventions can be effective.

This framing makes treatment feel less like a series of failures when things don’t work immediately, and more like a systematic exploration of their brain’s unique coalition of systems.

The Practice Transformation

Saying “I don’t know” more often has made me a better psychiatrist. It’s made my patients more engaged in their care. It’s reduced the pressure I felt to have immediate answers to enormously complex questions.

Most importantly, it’s created space for the kind of honest collaboration that actually helps people get better. When patients trust that I’ll tell them what I don’t know, they also trust what I do know.

The three words haven’t made my job easier, but they’ve made it more honest. And in a field where so much remains uncertain, honesty might be the most therapeutic thing we can offer.

Understanding starts with truth. And sometimes, the most powerful thing we can say, clinician or patient, is: “I don’t know yet, but let’s figure it out together.”

Substance-Induced Psychosis: When Learning Algorithms Distort Reality

michaelhalassa — Sun, 14 Sep 2025 20:15:17 +0000

The Portal in the Basement

The EMTs who brought Stephanie to the inpatient unit looked genuinely unsettled. She’d been found at 3 AM, methodically photographing her neighbors’ houses with her phone, documenting what she described as “dimensional breach points.”

“There’s a portal to a different galaxy in the Johnsons’ basement,” she emphatically explained to me the next morning. “They’ve been in contact with an advanced civilization for months. I have over a thousand photos documenting the evidence.”

Stephanie was 58, a senior vice president at a medium-sized company. She had no previous psychiatric hospitalizations and no major mental illness in her family. She’d been successfully managing ADHD with stimulants for thirty years. I wondered what had changed to bring on psychotic symptoms all of a sudden.

Her son filled in the details when I called him. “She’s been working insane hours since the new product launch six months ago. Said she needed to stay sharp, couldn’t afford to fall behind. I think she was taking way more Adderall than prescribed, but she insisted her doctor had increased it.”

The medical record told a different story. Her last psychiatry appointment was eight months ago. Her stimulant prescription hadn’t changed in two years.

Three weeks into her admission, completely off stimulants, Stephanie still believed in the portal. Importantly, we had done a thorough rule out of psychosis secondary to autoimmune and neurological entities. Nonetheless, Stephanie had developed an elaborate cosmological theory involving interdimensional communication protocols and galactic surveillance networks.

The Persistence Problem

According to DSM-5-TR criteria, substance-induced psychotic disorder should resolve when the substance clears. The hallmark feature distinguishing it from primary psychotic disorders is temporal relationship: symptoms emerge during intoxication or withdrawal and disappear during sustained sobriety.

But Stephanie wasn’t following the script. Her psychotic symptoms had crystallized into a stable, internally consistent belief system that persisted weeks after stimulant discontinuation. She wasn’t alone in this persistence.

Large epidemiological studies suggest that 10-25% of substance-induced psychotic episodes don’t resolve as expected. Some patients transition to diagnoses like schizophreniform disorder or brief psychotic disorder. Others remain in diagnostic limbo: no longer substance-induced, not quite meeting criteria for primary psychotic disorders.

The conventional explanation focuses on “unmasking” underlying vulnerability. The substance supposedly reveals a predisposition that was always there, waiting to emerge. But this explanation feels unsatisfying. Why do some people develop persistent psychosis after chronic stimulant use while others don’t? What’s actually changing in the brain during those months or years of escalating use?

Understanding Stephanie’s case required moving beyond simple neurochemical explanations toward a computational framework that could explain the persistence, internal coherence, and treatment-resistance of her symptoms. Her eventual treatment design incorporated an algorithmic circuit framework, and the response suggested that persistent substance-induced psychosis might represent biased constraint satisfaction algorithms rather than a persistent hyperdopaminergic state. But demonstrating this required understanding how chronic stimulants alter the confidence estimates that feed into updating our models of reality.

From a Simple Molecular Narrative to the Complexity of Learning in the Brain

The standard story about stimulant-induced psychosis centers on dopamine. Stimulants block dopamine reuptake, leading to excessive dopamine signaling. This hyperdopaminergic state is what psychosis is. Case closed.

The idea that excessive dopamine underlies psychotic symptoms is supported by two major pieces of evidence. Neuroimaging of dopamine in the ventral striatum early in psychotic illness reveals excessive release, and traditional antipsychotic efficacy is tied to how well they block dopaminergic signaling. In primary psychotic disorders, the idea is that a hyperdopaminergic state is triggered by an underlying genetic vulnerability, and that environmental stressors can lower the threshold for such vulnerability to be expressed. Stimulant use may be one such environmental stressor. With Stephanie, the age of onset was inconsistent with such an interpretation.

But the bigger issue at stake is this: what does dopamine have to do with thinking in the first place, and how does this molecule actually work in that context?

The prefrontal cortex (PFC) houses our most sophisticated cognitive machinery: neural populations that maintain working memory, simulate future scenarios, and hold the beliefs that guide our decisions. These mental maps form the neural substrate of planning and inference, allowing us to navigate complex environments and make sense of ambiguous information. The prefrontal cortex, like other cortical areas, is composed of 80-85% excitatory neurons that use glutamate to communicate with one another. The remaining 15-20% are inhibitory neurons that perform functions like gating, filtering and normalization of signals transmitted among the excitatory glutamatergic neurons.

The prefrontal cortex engages in metalearning (learning how to learn) by discovering which strategies and representations work across different contexts and tasks. Rather than just memorizing specific stimulus-response patterns, it extracts abstract rules and principles that can be applied to novel situations.

This metalearning happens through self-supervised learning processes, where the PFC uses the inherent structure of experience to generate its own training signals. For example, it might learn to predict future events based on current context, or identify which environmental cues reliably predict important outcomes. These prediction tasks don’t require external labels: the PFC generates supervisory signals from the temporal structure of experience itself.

What distinguishes the PFC from other cortical areas is its timescale for temporal integration. While sensory areas might integrate information over milliseconds to seconds, the PFC operates over seconds to minutes. This longer temporal window allows it to detect patterns and relationships that unfold across extended behavioral sequences, enabling the extraction of abstract rules that persist across changing contexts.

The PFC maintains extensive connections throughout the brain, but is particularly involved in loops with two subcortical regions: the thalamus and basal ganglia. Recent work, including Wang and colleagues’ influential DeepMind study, suggests that the PFC learns predictive models of the environment, but that changes in its connectivity patterns support metalearning rather than adapting to individual tasks. The way it adapts to individual tasks is by adjusting its activity patterns rapidly, not connectivity slowly. The thalamus appears to be the brain region that sends signals to rapidly adjust PFC activity patterns, helping it adapt its world models to the current environment. The thalamus receives inputs from regions like the cerebellum, critical for motor adaptation, and the hippocampus, which outputs compressed long-term memory episodes.

How does dopamine factor into this picture? Dopamine is expressed by neurons scattered in the midbrain and brainstem, with the midbrain populations most relevant to our discussion. The major success story in systems neuroscience is that dopamine neurons signal reward prediction errors in temporal difference learning algorithms: computing the difference between expected and actual outcomes to drive learning. In the classic formulation, the temporal difference error is δ = r + γV(s’) – V(s), where r is the immediate reward, γ is the discount factor, V(s’) is the predicted value of the next state, and V(s) is the current state’s value.

The largest concentration of dopamine receptors is in the striatum, the first station of the basal ganglia, where these prediction error signals are expected to adjust value representations encoded by striatal populations. However, two major developments have transformed our understanding of dopamine’s relationship to thinking and planning.

First, whatever happens in the striatum ultimately impacts the thalamus and then the PFC. Recent evidence suggests that dopamine-driven changes in striatal circuits influence how prefrontal world models get updated through cortico-basal ganglia-thalamic loops. This creates a pathway for reward prediction errors to systematically bias the metalearning processes that maintain our beliefs about reality.

Second, emerging research on distributional reinforcement learning reveals that dopamine neurons don’t just signal simple prediction errors (basic “better than expected” or “worse than expected” signals). Instead, they encode the full statistical distribution of possible prediction errors. Think of it like this: instead of just saying “that was surprising,” different dopamine neurons maintain different perspectives on what kinds of surprises to expect and how to weight them. Some neurons act as “optimistic” predictors that overweight positive prediction errors, while others act as “pessimistic” predictors that overweight negative prediction errors. Moreover, individual dopamine neurons encode different discount factors: some optimized for short-term rewards, others for long-term outcomes.

Stimulant use may change several critical aspects of this architecture. First, it could alter how reward prediction errors are computed: the pallidal signals feeding back to the ventral tegmental area (VTA, a midbrain dopamine-producing region) to compute prediction error differences would be acutely altered. This may impact different VTA neurons differently: those with optimistic versus pessimistic biases, and those with short versus long discount factors, could show differential vulnerability to stimulant-induced alterations.

Second, stimulant use may change how quickly striatal systems impact the updating of PFC world models through thalamic loops. If the normal temporal dynamics of this metalearning process are accelerated or biased, the PFC might begin incorporating unreliable distributional information into its predictive models of reality.

Here’s a potential mechanistic insight: distributional alterations may not create noisy signals so much as systematically distort the confidence estimates that guide belief updating. In healthy brains, the width of prediction error distributions appears to signal uncertainty. When you encounter something unexpected, wide distributions might tell the system “be cautious, gather more evidence.” Narrow distributions could signal high confidence: “update your beliefs strongly based on this information.”

Chronic stimulant use might bias this uncertainty signaling by creating artificially narrow distributions around extreme positive prediction errors. The system could receive signals that essentially convey “high confidence that this unexpected pattern is highly significant.” This could alter the normal process by which the brain decides how much to update its beliefs based on new evidence.

When patients notice unusual environmental patterns, their altered distributional system might generate high-confidence signals about the significance of these observations. Instead of the appropriate response (“this might be random, collect more evidence”), the metalearning algorithms could receive the message “this is definitely important, build strong beliefs around it.”

The thalamic gating system, which appears to filter which patterns get promoted to adjust beliefs and plans, likely relies on these confidence estimates to make gating decisions. Altered confidence estimates might cause it to gate spurious patterns as if they were reliable environmental regularities. Once gated into prefrontal circuits, these patterns could become the foundation for elaborate belief systems.

This sets up a crucial question: how does the brain actually construct unified belief systems from these distributed confidence signals? Understanding this process illuminates why Stephanie’s delusions were so resistant to contradictory evidence.

The Computational Architecture of Belief Coherence

The persistence of Stephanie’s delusional beliefs reveals a fundamental computational challenge: how does the brain construct unified models of reality from distributed, potentially contradictory evidence? This problem has been formalized as coherence maximization, a well-studied algorithmic framework that illuminates how biased confidence signals can lead to systematically distorted worldviews.

Coherence maximization is formally defined as a constraint satisfaction problem, where mental representations either fit together (cohere) via positive constraints or resist fitting together (incohere) via negative constraints. The brain seeks configurations that maximize the satisfaction of these constraints: essentially finding the most internally consistent interpretation of available evidence.

The computational complexity of this process is significant: coherence problems are computationally intractable for large systems, requiring approximation algorithms similar to those used for traveling salesman problems or neural network optimization. This computational challenge explains why specialized neural circuits evolved to handle belief integration.

The brain implements coherence maximization through what computational neuroscientists call active inference: a framework where organisms maintain generative models of their environment and continuously update these models to minimize “free energy,” a measure of surprise or model-environment mismatch. Under this framework, beliefs are not passive representations but active hypotheses that guide both perception and action, with the system working to maintain internal consistency while accommodating new evidence.

This algorithmic framework appears in multiple computational domains. The ECHO (Explanatory Coherence by Harmonic Optimization) algorithm, developed for scientific reasoning, uses connectionist networks where propositions are represented as nodes and coherence relationships as weighted connections. The network settles into configurations that maximize overall constraint satisfaction. Modern machine learning systems face similar challenges: large language models must maintain consistency across vast knowledge bases, and this coherence-seeking behavior emerges naturally from their training objectives.

The evolutionary utility of coherence-seeking becomes clear from this computational perspective: organisms that can rapidly construct consistent internal models from fragmentary evidence gain survival advantages through improved prediction and decision-making. However, this same mechanism becomes problematic when the input signals (the confidence estimates about environmental patterns) become systematically biased.

When chronic stimulant use distorts the distributional properties of prediction errors, the coherence maximization system receives corrupted input: patterns that should be flagged as low-confidence noise instead arrive with high-confidence signals demanding explanatory integration. The system treats these spurious patterns as reliable environmental regularities requiring coherent explanation, leading to elaborate belief systems that represent optimal solutions to a fundamentally corrupted constraint satisfaction problem.

This computational framework explains why persistent substance-induced delusions resist simple contradictory evidence. The beliefs aren’t random false ideas; they’re optimal coherence solutions given systematically biased confidence inputs. Breaking these belief systems requires restoring the underlying computational processes that generate appropriate confidence estimates in the first place, not just presenting contradictory evidence.

Given this understanding, traditional psychiatric approaches that focus solely on blocking neurotransmitter receptors miss the deeper computational dysfunction.

Why Single-Receptor Models Miss the Circuit Story

This algorithmic circuit framework reveals why traditional dopamine-receptor models struggle to explain persistent substance-induced psychosis. The standard approach focuses on blocking D2 receptors with antipsychotics to suppress psychotic symptoms, but this doesn’t address the underlying computational dysfunction that generates biased belief systems.

The problem goes beyond “too much dopamine signaling.” It’s systematically altered distributional learning: corrupted confidence estimates that feed into coherence maximization algorithms, leading to elaborate but internally consistent delusional frameworks. These beliefs persist because they represent optimal solutions to a constraint satisfaction problem operating on fundamentally biased inputs.

A purely receptor-based approach treats the neurochemical symptoms rather than the computational dysfunction. Blocking dopamine receptors may reduce the intensity of aberrant signals, but it doesn’t restore the distributional properties that allow metalearning circuits to distinguish reliable environmental patterns from noise. The coherence maximization system continues operating on the same corrupted confidence estimates, just with dampened intensity.

Understanding how circuits implement distributional learning algorithms and how chronic stimulants systematically bias these implementations suggests more targeted interventions that address the computational roots of persistent delusions rather than just their neurochemical expression.

With this framework in mind, Stephanie’s treatment required a fundamentally different approach than standard antipsychotic protocols.

Treatment Through an Algorithmic Circuit Lens

Understanding Stephanie’s symptoms through the distributional learning framework suggested we needed a multi-pronged therapeutic approach. The altered prediction error distributions driving her persistent delusions couldn’t be fixed by simply blocking dopamine receptors. We needed to restore the natural diversity and distributional properties of dopaminergic signaling itself.

Standard antipsychotics like haloperidol or risperidone work by dampening aberrant dopamine signals in striatal circuits. A 2019 systematic review of six randomized controlled trials found that various antipsychotics (aripiprazole, haloperidol, quetiapine, olanzapine, and risperidone) were all effective at reducing both positive and negative symptoms of amphetamine-induced psychosis. But this dampening approach addresses only part of the distributional learning problem.

The algorithmic framework suggests that successful treatment requires two complementary strategies: first, reduce the impact of biased prediction error signals on coherence maximization circuits; second, restore the capacity for healthy distributional learning by normalizing the statistical properties that generate appropriate confidence estimates.

For the first component, we chose aripiprazole over traditional D2 antagonists. Its partial agonism at dopamine receptors could theoretically provide more nuanced modulation: dampening excessive signals while preserving some baseline dopaminergic function needed for normal distributional learning. The systematic review noted that aripiprazole showed particular effectiveness for negative symptoms, which might reflect its ability to maintain residual dopaminergic signaling rather than completely blocking the system.

But medication alone wouldn’t restore healthy distributional learning. Stephanie’s constraint satisfaction algorithms needed to relearn how to process confidence estimates appropriately. This required tackling the underlying cause: her dependence on chronic stimulants that had systematically biased the distributional properties feeding into her coherence maximization system.

Restoring Distributional Diversity: Replacement and Modulation

The algorithmic circuit framework pointed toward two additional interventions that could help restore normal distributional learning properties.

First, we needed to address Stephanie’s underlying ADHD without continuing to bias her prediction error distributions. Wellbutrin (bupropion) offered a promising alternative. Unlike amphetamines, which create massive positive prediction errors through dopamine reuptake blockade, bupropion provides more modest, sustained increases in dopaminergic and noradrenergic signaling. Its mechanism might preserve more natural distributional properties while still providing therapeutic benefit for attention deficits.

The hypothesis here is that Wellbutrin could serve as replacement therapy: providing enough cognitive enhancement to manage her ADHD symptoms while allowing her biased distributional learning circuits to gradually renormalize. Instead of the extreme positive prediction errors from chronic stimulants, she would experience more naturalistic dopaminergic signaling patterns that could support healthy confidence estimation.

Second, we cross-titrated aripiprazole to KarXT, a combination of xanomeline and trospium that targets muscarinic receptors. The algorithmic framework suggests this might work by modulating the inputs to dopaminergic circuits rather than directly blocking dopamine receptors. My early clinical experience with this cholinergic modulation appears to support its efficacy in stimulant-induced psychosis.

Cholinergic signaling plays crucial roles in regulating the context-dependent release of dopamine. By modulating muscarinic receptors, KarXT could potentially help restore more natural patterns of dopaminergic signaling: not by suppressing all dopamine activity, but by helping circuits generate more appropriate distributional responses to environmental inputs.

This represents a fundamentally different therapeutic approach: instead of just dampening aberrant signals, we’re trying to restore the circuit mechanisms that generate healthy distributional learning in the first place.

But even optimal pharmacological intervention addresses only half the problem. The other half involves helping patients’ constraint satisfaction algorithms relearn how to process confidence information appropriately.

Circuit Rehabilitation: A Future Direction

Beyond pharmacological interventions, patients with biased constraint satisfaction algorithms may need active rehabilitation. Traditional cognitive behavioral therapy focuses on changing thoughts through rational examination. But persistent delusions aren’t irrational thoughts; they’re outputs from coherence maximization systems operating on systematically biased confidence estimates.

Future therapeutic interventions might look quite different from standard CBT. Rather than challenging the content of delusional beliefs directly, sessions could focus on the process of evidence evaluation itself. Imagine sitting with a patient and working through their observations systematically. Not “you’re wrong about the portal” but “let’s think about all the possible explanations for what you noticed.” What other reasons might account for changes in your neighbor’s lighting patterns? How confident should we be in each explanation? What additional evidence would help us distinguish between them?

The goal wouldn’t be to convince patients they’re wrong. It would be helping their constraint satisfaction algorithms practice processing confidence information appropriately: distinguishing between high-confidence and low-confidence inferences, calibrating degrees of belief to strength of evidence, and maintaining uncertainty when evidence is ambiguous.

Such approaches might reveal that patients’ observations aren’t entirely false. They may have noticed real environmental patterns. But their biased learning algorithms assign extreme confidence to elaborate explanations when the evidence actually supports much simpler, more probable alternatives.

By systematically examining the distributional properties of evidence (the range of possible explanations and their relative probabilities), therapeutic interventions could potentially help these circuits begin distinguishing signal from noise again.

Beyond Chemical Imbalance

Stephanie’s case illustrates why algorithmic circuit psychiatry offers superior explanatory power to traditional “chemical imbalance” models. The chemical imbalance framework struggles to explain key features of her presentation: why her symptoms persisted weeks after stimulant clearance, why her delusions were internally coherent rather than random, and why standard dopamine blockade provided only partial relief.

The algorithmic framework provides a more comprehensive explanation. Chronic stimulants didn’t create “too much dopamine.” They systematically biased the distributional properties that feed confidence estimates into constraint satisfaction algorithms. Her elaborate interdimensional theory wasn’t a symptom of broken brain chemistry but an optimal solution to a corrupted computational problem. Traditional antipsychotics dampened the signals but couldn’t restore the underlying distributional learning processes.

Understanding psychiatric symptoms as biased algorithms rather than chemical imbalances opens new therapeutic possibilities. Instead of treating medication and therapy as separate interventions targeting different domains, we can recognize them as complementary approaches working on the same computational substrate. Pharmacological interventions like cholinergic modulation help restore healthy distributional properties in the circuits that generate confidence estimates. Therapeutic interventions help retrain these same constraint satisfaction algorithms to process confidence information more appropriately. Both target the algorithmic dysfunction that generates pathological beliefs.

Stephanie’s recovery with cholinergic modulation and stimulant replacement suggests that restoring healthy algorithmic function may be more effective than suppressing aberrant chemistry. The brain implements sophisticated learning algorithms through specific circuit architectures. When we understand how these algorithms can be corrupted and restored, we move beyond the limitations of purely neurochemical approaches toward interventions that address the computational roots of psychiatric dysfunction.

Systems Neuroscience as a Foundation for Psychiatric Drug Discovery

michaelhalassa — Fri, 11 Jul 2025 04:55:38 +0000

The recent Innovation in Psychosis Therapeutics Summit in Boston revealed a clear truth: while the field celebrates breakthroughs like muscarinic agents, it’s equally clear that psychiatric drug development needs a neuroscience reboot. We’ve mastered molecules. What we lack is a model of the mind.

The molecular era of neuroscience has been productive. It gave us the tools to understand neurotransmitter systems, develop targeted receptor modulators, and generate narratives that we can explain to patients and families struggling to understand the burden of mental illness. The dopamine hypothesis, GABAergic interventions, and serotonergic medications established the scientific credibility of early biological psychiatry.

But as became clear throughout the summit discussions and pre-meeting workshops, the extreme focus on molecular details as the ‘mechanism of action’ misses a larger point. In contrast to other organ systems where knowledge of molecular and cell biology gives one a pretty reasonable understanding of organ function (Cytochrome P450 functioning in a hepatocyte tells you quite a bit about what the liver does in drug detoxification and actin sliding on myosin in a cardiomyocyte explains a lot of what the heart does), understanding action potentials and synaptic transmission tells us very little about how thinking works.

Think of it this way: studying individual brain cells and their chemical signals to understand mental illness is like trying to understand a movie by analyzing the pixels on your TV screen. You can learn a lot about how pixels work (their color values, brightness, refresh rates) but that won’t tell you whether you’re watching a comedy or a thriller, or why the plot doesn’t make sense. The story emerges from how all those pixels work together in patterns over time.

This is psychiatry’s fundamental challenge. We’ve become experts at the “pixels” (the molecular mechanisms, neurotransmitter systems, and individual brain cells). But mental illness isn’t a problem with individual pixels. It’s a problem with how the brain’s software processes information, makes decisions, and builds our sense of reality.

Building the Brain’s “Flight Simulator”

What we really need is something like a flight simulator for the brain—computational models that can show us how molecular changes ripple through neural circuits to affect thinking, emotion, and behavior. Just as pilots use flight simulators to understand how adjusting one control affects the entire aircraft’s performance, we need brain simulators to predict how a new medication will affect a person’s ability to think clearly, regulate emotions, or maintain stable beliefs about reality.

Take depression, for example. Molecular framing focuses on “low serotonin” or other types of “chemical imbalances.” But computational models indicate that certain forms of depression have more to do with how the brain learns from rewards and punishments. Imagine your brain has a built-in prediction system that’s supposed to help you learn from experience, when good things happen, it should update your expectations upward; when bad things happen, it should adjust appropriately. In depression, this system over-learns from negative experiences and under-learns from positive ones, creating a downward spiral of increasingly pessimistic predictions about the future.

Of course, this algorithm has a neural implementation—involving specific circuits, cell types, and neuromodulators—but the unit of analysis most relevant to symptoms and their relief is the algorithm itself, not the transmitter systems.

Understanding this algorithmic dysfunction opens up entirely new treatment possibilities. Instead of just trying to boost serotonin levels, we can target the specific computational processes that have gone awry.

Recent clinical trials are demonstrating exactly this approach. Researchers have used computational models to predict which patients with depression will respond to cognitive behavioral therapy by measuring how their brains process reward prediction errors during learning tasks (Rzepa et al., 2017). Other studies have shown that computational measures of effort-based decision-making can predict which patients will relapse after stopping antidepressants, identifying a persistent algorithmic dysfunction that outlasts mood symptoms (Berwian et al., 2020).

This isn’t just about having more treatment options. It’s about matching the right intervention to the right computational problem. Some patients might benefit most from medications that restore balanced reward learning. Others might need brain stimulation that resets dysfunctional prediction circuits. Still others might respond best to digital therapies that provide targeted algorithm retraining.

The Missing Piece: Systems Neuroscience

Here’s what’s been missing from the molecular-to-computational translation: systems neuroscience. Over the past two decades, this field has exploded with revolutionary tools and insights that completely change how we understand brain function. We can now record from hundreds of neurons simultaneously, manipulate specific cell types with optogenetics, trace connectivity patterns across entire brains, and interpret brain dynamics with unprecedented precision.

These advances have revealed something remarkable: the brain operates through large-scale computational principles that emerge from how circuits are organized and interact. We’ve discovered that the cortex implements hierarchical predictive processing—constantly generating predictions about incoming information and updating these predictions when they’re wrong. We’ve learned that the dopaminergic system implements temporal difference learning in the brain. We’ve found that the hippocampus works like a sophisticated pattern-completion system, able to reconstruct entire memories from partial cues by leveraging the same mathematical principles that power modern AI memory networks.

I have been fortunate to establish my lab around the time many of the technical advances in systems neuroscience had come to the fore. Using these tools and working with many talented students and collaborators, we made a series of surprising observations that challenged a long held dogma: the thalamus, considered a major sensory relay station in the brain, plays critical roles in higher cognition. In my own lab, we’ve used these tools to understand how the thalamus regulates cortical state switching—an operation fundamental to cognitive flexibility and psychiatric dysfunction.

In fact, most of the thalamus in our brains as humans is unlikely to play much of a role in sensory processing. Instead, it dynamically regulates cortical dynamics and implements context-dependent gating of information flow. This discovery emerged from combining well-controlled animal behavior (building on years of work by pioneers in the field), optogenetic manipulations, and high-density neural recordings.

The prefrontal cortex is a critical area in psychiatry because its neurons form coalitions that provide mental simulations, working memory and action plans. My lab among others discovered that inputs from the thalamus are critical for maintaining and switching prefrontal representations underlying these algorithmic processes. In essence, when you need to switch between different mental tasks, thalamic circuits provide the actual switching signals, determining the timing and specificity of cortical state changes.

This has profound implications for understanding cognitive deficits in disorders like schizophrenia. There is good neuroimaging evidence to suggest thalamic dysfunction in schizophrenia and we are in early stages trying to determine whether that may be related to the inability of patients to maintain accurate models of the world, revise their mental simulations when they are implemented or some combination of such processes. Close integration between animal and human work is key to making good progress.

Most importantly, this systems-level understanding opens new therapeutic possibilities. Rather than targeting broad neurotransmitter systems, we might develop interventions that specifically modulate thalamocortical dynamics. For instance, understanding how cholinergic signaling regulates thalamic gating could inform more precise pharmacological approaches. Similarly, targeted neuromodulation techniques could potentially restore proper state regulation in these circuits. However, translating these insights into clinical interventions will require careful validation of the computational models we develop in animals and their relevance to human psychiatric conditions.

Algorithmic Circuit Psychiatry: The Bridge We Need

This is where systems neuroscience becomes the essential bridge between cellular neuroscience and computational science. We can now connect specific molecular mechanisms to circuit dynamics to algorithmic functions—creating what I call “algorithmic circuit psychiatry.”

The framework works like this: cellular neuroscience identifies the molecular players (receptors, channels, neurotransmitters), systems neuroscience reveals how these molecules shape circuit dynamics and information processing, and computational science provides the mathematical frameworks to understand what algorithms these circuits implement. Instead of having three separate fields talking past each other, we can trace a coherent path from molecules to circuits to algorithms to symptoms.

Designing the Next Generation of Trials

The clinical application of this framework involves a systematic approach: first, we decompose patient symptoms using computational methods, fitting their behavioral data into mathematical models and extracting specific algorithmic parameters. Next, we use precision neuroimaging to identify the neural circuit alterations underlying these computational dysfunctions. Finally, we leverage mechanistic models built from animal studies to predict which pharmacological and behavioral interventions will restore healthy circuit-algorithm function in each individual patient.

This approach could fundamentally transform psychiatric treatment by replacing trial-and-error prescribing with mechanistically-informed precision medicine. Mental illness is not caused by broken molecules, but by maladaptive computations implemented in circuit dynamics. The treatment of the future won’t correct a chemical imbalance—it will recalibrate an algorithm.

Instead of cycling through different medications hoping something works, we could predict treatment response based on each patient’s specific pattern of circuit-algorithm dysfunction. The computational parameters tell us what’s broken, the neuroimaging reveals where it’s broken, and the mechanistic models suggest how to fix it.

The Path Forward: Evolution, Not Revolution

What’s most exciting about this moment is that we’re not throwing out decades of neuroscience research. Instead, we’re building on that solid molecular foundation to create more sophisticated, comprehensive approaches to psychiatric treatment.

This evolution is already transforming drug development in several ways. For smarter target identification, instead of hunting for individual molecules to drug, we can identify key bottlenecks in dysfunctional brain algorithms and ask what molecular interventions might restore healthy computational processes.

We’re also developing better animal models. Instead of relying on crude behavioral measures that don’t really capture human mental illness, we can focus on algorithmic functions that are conserved across species and ask whether potential treatments restore these core computational abilities.

This approach enables more meaningful biomarkers. Instead of simple blood tests or brain scans, we can develop assessments that capture how well someone’s brain algorithms are functioning, providing much richer information for treatment selection and monitoring progress.

Finally, understanding how different interventions work across levels opens up possibilities for rational combination therapies. We might pair a medication that fixes a molecular problem with brain stimulation that resets dysfunctional circuits and cognitive training that helps retrain maladaptive algorithms.

An Invitation to the Future

The conversations following my Boston summit report suggest that the field is ready for this evolution. Researchers across academia and industry are recognizing that our most exciting recent advances have come from thinking about mental illness as a multi-level problem requiring multi-level solutions.

This isn’t about abandoning the rigorous molecular research that brought us this far. It’s about using that foundation to build something much more powerful—treatments that are informed by molecular mechanisms, guided by circuit-level insights, and targeted toward restoring the algorithms that generate healthy thinking and emotion.

We have the molecular foundation. Circuit-level insights are maturing rapidly. Computational frameworks are emerging from labs around the world. The clinical need remains as urgent as ever.

The pieces are finally in place for a new generation of psychiatric treatments—ones that don’t just manage symptoms, but recalibrate the brain’s computational machinery for healthy thinking, feeling, and action.

The time for integration is now.

A Blueprint for Algorithmic Psychiatry: Revolutionizing Mental Health Treatment

michaelhalassa — Thu, 12 Sep 2024 03:44:57 +0000

Mental health care has long been dominated by the Diagnostic and Statistical Manual of Mental Disorders (DSM), a classification system that guides clinicians in diagnosing and treating psychiatric illnesses. While the DSM has provided a standardized approach to understanding mental disorders, it also has its limitations. Psychiatric diagnoses often represent clusters of symptoms rather than underlying mechanisms, leading to treatment approaches that may not address the root causes of mental illness. Enter algorithmic psychiatry, a new framework poised to revolutionize the field by focusing on latent computational features and predictive models to enhance treatment response.

The Traditional DSM-Based Approach: A Quick Overview

The DSM-based approach relies on categorizing psychiatric disorders into discrete entities based on observable symptoms and patient history. While this method has been instrumental in standardizing diagnoses, it tends to overlook the complex, multidimensional nature of mental illnesses. As a result, treatment options are often generalized and may not be effective for everyone within a diagnostic category. This can lead to trial-and-error prescribing, delayed treatment response, and, in some cases, inadequate care.

What Is Algorithmic Psychiatry?

Algorithmic psychiatry represents a paradigm shift in mental health treatment by moving beyond symptom-based diagnoses to focus on the underlying latent computational features of psychiatric illnesses. These features are independent constructs that can be targeted more precisely with various interventions, such as pharmacology, therapy, and neurostimulation. The ultimate goal is to put treatment response at the forefront of psychiatric care, using predictive models to guide clinical decisions.

The Foundation: Latent Computational Features

At the heart of algorithmic psychiatry is the concept of latent computational features. These are unobservable variables that underlie the observable symptoms of mental illness. Unlike the DSM categories, which group symptoms into broad diagnoses, latent features are more granular and can be more directly linked to specific brain processes, cognitive functions, and behavioral patterns.

For example, instead of diagnosing a patient with generalized anxiety disorder (GAD) based solely on their reported symptoms, an algorithmic psychiatry approach might identify specific latent features such as cognitive rigidity, hypervigilance, or impaired emotional regulation. These features can then be targeted with more precise treatments, potentially leading to faster and more effective symptom relief.

Clinical Interviews and Behavioral Tasks: Building the Model

The initial identification of latent features can begin with a comprehensive clinical interview, focusing not just on symptomatology but also on cognitive and emotional processes. To augment this, patients can undergo well-controlled behavioral tasks designed to probe specific mental functions. For example, tasks focusing on reasoning might help identify cognitive biases or decision-making impairments, while tasks focusing on emotions could reveal difficulties in emotional regulation or processing.

These behavioral assessments provide the first layer of data that can be fed into predictive models. The goal is to create an initial framework that can guide treatment decisions even before more complex data—such as biological markers or neuroimaging—becomes available.

Augmenting the Model: The Role of Data

As more data becomes available, the predictive models in algorithmic psychiatry can be continuously refined. This data can come from a variety of sources, including:

Biologics: Genomic, transcriptomic, and proteomic data can provide insights into the molecular underpinnings of mental illness. For example, certain genetic markers might be linked to specific latent features, guiding pharmacological interventions.
Organoids: These lab-grown mini-brains can model specific neural circuits, offering a window into the cellular and molecular mechanisms underlying psychiatric disorders.
Neuroimaging: Advanced imaging techniques can reveal structural and functional abnormalities in the brain, helping to refine the identification of latent features.
Wearable Technologies: Devices that track physiological and behavioral data in real-time can offer continuous monitoring of a patient’s mental state, providing a dynamic input to the predictive models.

Predictive Models: The Core of Treatment Decisions

The predictive models used in algorithmic psychiatry are designed to be dynamic and adaptive. As new data is integrated, the models can adjust, providing increasingly accurate predictions of treatment response. These models prioritize individual variability, ensuring that treatment plans are tailored to each patient’s unique profile of latent features.

For example, a patient with depression might have a model that predicts a higher likelihood of response to cognitive behavioral therapy (CBT) due to identified features such as cognitive distortions. Another patient with the same DSM diagnosis but different latent features might be predicted to respond better to a combination of pharmacotherapy and neurostimulation.

The Future of Psychiatric Care

Algorithmic psychiatry has the potential to revolutionize mental health care by making it more personalized, precise, and data-driven. By focusing on latent computational features and utilizing predictive models, this approach aims to optimize treatment response from the outset, reducing the need for trial-and-error prescribing and improving patient outcomes.

As the field evolves, algorithmic psychiatry could also pave the way for the early identification of psychiatric disorders, enabling preventative interventions before full-blown symptoms develop. Moreover, as more data is collected and models become more sophisticated, the hope is that we can move closer to understanding the true nature of mental illness, leading to even more effective treatments.

Conclusion

The transition from a DSM-based framework to algorithmic psychiatry represents a bold step forward in the treatment of psychiatric illnesses. By focusing on independent latent computational features and integrating diverse data sources into predictive models, this approach promises to put treatment response at the center of mental health care. As we continue to refine these models and gather more data, the future of psychiatry looks brighter, with the potential for more effective, personalized treatments that address the root causes of mental illness.