Michael Halassa | Psychiatry https://michaelhalassa.com Michael Halassa's Psychiatry site Mon, 16 Mar 2026 23:09:03 +0000 en-US hourly 1 https://michaelhalassa.com/wp-content/uploads/michaelhalassa-com/2024/07/cropped-Michael-Halassa-Logo-32x32.jpg Michael Halassa | Psychiatry https://michaelhalassa.com 32 32 The Long Game of Stimulants and Psychosis https://michaelhalassa.com/stimulants-and-psyhosis/ Mon, 16 Mar 2026 23:09:03 +0000 https://michaelhalassa.com/?p=827
]]> Why I Tell My Patients “I Don’t Know” https://michaelhalassa.com/idontknow/ Mon, 16 Mar 2026 22:41:02 +0000 https://michaelhalassa.com/?p=822
]]> Substance-Induced Psychosis: When Learning Algorithms Distort Reality https://michaelhalassa.com/substanceinducedpsychosis/ Sun, 14 Sep 2025 20:15:17 +0000 https://michaelhalassa.com/?p=816 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.

]]> The Right Level of Wrong: Why Psychiatric Conditions Need an Algorithmic Explanation https://michaelhalassa.com/the-right-level-of-wrong-why-psychiatric-conditions-need-an-algorithmic-explanation/ Mon, 18 Aug 2025 19:39:33 +0000 https://michaelhalassa.com/?p=812 There’s a paradox at the heart of modern psychiatry that keeps me up at night. We have two ways of understanding mental illness that seem fundamentally at odds, yet both appear indispensable. On one side, we have the molecular story: dopamine receptors, neurotransmitter reuptake, receptor binding affinities. On the other, we have the algorithmic story: temporal difference learning, model-free versus model-based systems, computational arbitration between competing decision-making processes.

Here’s the thing: I’m convinced the molecular level is the wrong level of explanation for understanding psychiatric symptoms, while simultaneously being grateful it exists for practical treatment decisions. I should be clear that I’m not critiquing the incredible work being done in molecular psychiatry; I want to enrich it with a complementary layer of algorithmic insight. And I think this paradox reveals something profound about what we actually mean when we say we want to “explain” mental illness.

Thanks for reading! Subscribe for free to receive new posts and support my work.

The Information Problem with Molecular Psychiatry

Let me start with why telling someone their “dopaminergic system is dysregulated” is, from an explanatory standpoint… well, not so much of an explanation. I’m not saying dopamine is not involved; I’m making an argument about what constitutes useful information.

In information theory, Claude Shannon taught us that information is fundamentally about reducing uncertainty. A statement carries information only to the extent that it narrows down the space of possible outcomes. (For the mathematically inclined: information is measured as the reduction in entropy, where entropy H = -Σ P(i) log₂ P(i) across all possible outcomes i. A statement that changes the probability distribution reduces uncertainty by ΔH = H_before – H_after bits. Feel free to skip the math; the key insight is that informative statements meaningfully change what outcomes we expect.)

Consider this: glutamate is the primary excitatory neurotransmitter in roughly 80-85% of forebrain neurons. So when someone says “your condition involves glutamate dysfunction,” they’re essentially saying “your condition involves the brain.” This is like explaining a car accident by noting that “wheels were involved”; technically true, but explanatorily vacuous.

The same logic applies to dopamine; invoking “dopamine dysregulation” doesn’t specify which behaviors will emerge, under what circumstances, or why this particular person experiences their particular constellation of symptoms.

The Algorithmic Alternative: Computation Specifies Behavior

Now contrast this with an algorithmic explanation. Instead of saying “your dopamine is low,” we might say something like this:

“Your brain uses two main systems to make decisions. One system is fast and habitual; it learns what to do based on what worked before, using a process called temporal difference learning. This system updates its predictions about rewards according to the rule: V(s) ← V(s) + α[r + γV(s’) – V(s)], where V(s) is the value of your current situation, r is the reward you just got, V(s’) is the value of your new situation, α is how much you learn from each experience, and γ is how much you care about future rewards versus immediate ones. (Don’t worry about the math; the key is that this system caches simple value estimates.)

The other system is slower and more deliberate; it builds internal models of how the world works—transition models P(s’|s,a) that predict what state s’ you’ll end up in if you take action a in your current state s, and reward models R(s,a) that predict what reward you’ll get from taking action a in state s. This enables flexible planning through what’s essentially mental simulation.

The final decision combines both systems: Q_total(s,a) = ωQ_MB(s,a) + (1-ω)Q_MF(s,a), where Q represents action values, MB is model-based, MF is model-free, and ω is the weighting parameter that determines whether you rely more on cached habits or deliberate planning. (Again, skip the math if you prefer—the key is that your brain literally arbitrates between fast habits and slow deliberation.)

But here’s where dopamine gets interesting. Rather than being a simple “reward chemical,” dopamine neurons actually exhibit structured diversity. Some dopamine neurons are optimistic (α+ > α−), responding more to better-than-expected outcomes, while others are pessimistic (α+ < α−), being more sensitive to disappointments. Additionally, neurons vary in their temporal horizons; some weight immediate rewards heavily while others give more weight to delayed outcomes.

Let me use substance use as an example to illustrate how this might work. In substance use disorders, we might hypothesize that this sophisticated system becomes dysregulated. The arbitration process could break down, with the brain losing confidence in its ability to predict and control future outcomes, making the deliberate planning system feel unreliable.

Meanwhile, the diversity in dopamine signaling might become systematically biased: neurons that should encode optimistic long-term outcomes (career success, family relationships) could become pessimistic, while neurons encoding immediate rewards (like drug effects) might become hyperoptimistic. This would create a profound temporal bias where immediate outcomes feel unrealistically positive while delayed outcomes feel unrealistically negative. The result could be decision-making dominated by the immediate, certain relief that substances provide, while long-term goals feel abstract and unattainable.”

Notice what this explanation does that the molecular one doesn’t: it specifies behavioral patterns and connects them to computational processes. It explains why someone might simultaneously overthink decisions yet feel unmotivated to act. It makes predictions about when symptoms might be worse (high-stakes decisions requiring deliberation) versus better (routine, well-learned activities). And it incorporates the sophisticated reality that dopamine isn’t just “reward chemical”; it’s a diverse population of neurons encoding different aspects of reward prediction, temporal horizons, and uncertainty.

This is what I mean by the algorithmic level being the “right” level of explanation. It’s the level at which we can actually map between neural computation and behavioral phenotypes.

But Wait—The Molecular Level Still Matters

Here’s where the paradox gets interesting. Despite everything I just said about molecular explanations being informationally sparse, I’m not advocating we abandon them. They remain incredibly useful for a specific purpose: guiding treatment decisions.

When a clinician needs to choose between medications, knowing that one patient might respond better to a strong D2 antagonist versus a partial agonist, or that another might benefit from glutamatergic modulation, these molecular distinctions become practically vital. The neurochemical level provides a useful abstraction for pharmacological intervention, even if it’s not the right level for psychological explanation.

This is a perfect illustration of George Box’s famous maxim: “All models are wrong, but some are useful.” Molecular psychiatry is wrong as an explanation of behavior, but useful as a guide to intervention.

Why Medicine Defaults to Molecular Thinking

It’s worth understanding why molecular explanations feel so natural in medicine. In most organ systems, cellular and molecular mechanisms do provide meaningful explanations for clinical phenomena. When we explain heart failure through weakened actin-myosin interactions in cardiac muscle, or liver dysfunction through compromised cytochrome P450 enzyme systems, we’re operating at the right level of biological organization. The molecular story directly connects to the organ’s function.

This success has shaped medical training profoundly. Medical students learn to think mechanistically: identify the broken molecular pathway, understand how it disrupts normal physiology, then intervene at that level. This approach works brilliantly for most of medicine. A cardiologist can predict that ACE inhibitors will help heart failure by blocking angiotensin-converting enzyme, reducing afterload on the heart. A hepatologist knows that certain drug interactions occur because they compete for the same P450 enzymes.

But the brain is different. Unlike other organs with relatively straightforward input-output relationships, the brain’s primary function is information processing and behavioral control. The gap between molecular events and behavioral outcomes is vast, filled with multiple levels of organization: circuits, networks, algorithms, and ultimately, the psychological phenomena that patients actually experience.

The problem isn’t that molecular mechanisms don’t matter in psychiatry; it’s that they don’t directly specify the behaviors and experiences we’re trying to explain. This creates a persistent explanatory gap that algorithmic thinking can help bridge.

Augmenting Clinical Intuition

What excites me most is how the algorithmic perspective can enhance rather than replace clinical molecular thinking. When we understand that different psychiatric conditions involve different patterns of computational dysfunction, we can make more sophisticated predictions about which molecular interventions might help.

For instance, if we conceptualize certain aspects of depression as involving hyperactive model-based control combined with decreased reward sensitivity, we might predict that medications affecting both dopaminergic reward signaling AND prefrontal modulatory systems would be most effective. But here’s where the distributional dopamine story becomes crucial: if the problem involves corrupted asymmetric scaling factors in dopamine populations; neurons that should encode optimistic long-term outcomes becoming systematically pessimistic while neurons encoding immediate rewards become hyperoptimistic; then effective treatment might require rebalancing this diversity rather than simply increasing or decreasing overall dopamine function.

I’m suggesting theoretically motivated combination therapy rather than trial-and-error polypharmacy. The computational framework could guide whether a patient needs interventions that restore optimistic future-oriented dopamine signaling, recalibrate temporal discount factors across neuron populations, or strengthen the prefrontal circuits that arbitrate between competing decision systems.

The Communication Benefits

Perhaps most importantly, algorithmic explanations improve how we talk with patients about their conditions. Instead of mystifying neurochemical imbalances, we can offer explanations that connect to lived experience while maintaining scientific rigor.

When someone with OCD asks why they can’t stop checking the door, we can explain how their brain’s uncertainty-monitoring system has become hypersensitive, leading to a computational loop where the threshold for “certain enough” is never reached. When someone with ADHD struggles with procrastination, we can discuss how their brain’s reward-prediction system has difficulty maintaining motivation for distant or uncertain outcomes.

These explanations don’t just satisfy intellectual curiosity; they provide actionable insights. They help patients recognize patterns, develop coping strategies, and understand why certain therapeutic approaches might be particularly helpful for their specific computational profile.

What Makes an Explanation Meaningful?

This brings us to a deeper philosophical question: what should we demand from a psychiatric explanation? I propose three criteria:
  1. Behavioral specificity: It should predict or account for specific patterns of behavior, not just invoke general dysfunction.
  2. Mechanistic precision: It should specify how inputs transform into outputs through identifiable computational processes.
  3. Experiential relevance: It should connect to the patient’s subjective experience in a way that feels both accurate and helpful.

Molecular explanations typically fail the first and third criteria, while succeeding at aspects of the second. Algorithmic explanations can potentially satisfy all three, while still remaining grounded in neuroscience.

The Path Forward

I’m not arguing for the wholesale abandonment of molecular psychiatry. Rather, I’m suggesting we need a more sophisticated understanding of which level of explanation serves which purpose. Use molecular models to guide pharmacological decisions. Use algorithmic models to understand, predict, and explain behavior. And use the integration of both levels to develop more precise, personalized approaches to treatment.

The future of psychiatric explanation lies not in choosing between levels of analysis, but in understanding how they complement each other. This shift could guide trial design, biomarker selection, and even AI-driven decision tools. For example, clinical trials could stratify participants by computational profiles rather than broad DSM categories, testing whether specific algorithmic phenotypes predict response to targeted treatments. The brain is simultaneously a chemical system, a computational device, and a meaning-making apparatus. Our explanations should honor this complexity while remaining useful for the people who need them most—our patients.

After all, the goal goes beyond being scientifically correct. We need to provide explanations that actually explain; that reduce uncertainty, specify mechanisms, and empower people to understand their own minds. Finding the right level of “wrong”; useful models that aren’t ultimate truths; that’s the heart of progress in psychiatry. That’s information worth having.

]]> Reflections on the 2025 Psychosis Innovation Summit in Boston https://michaelhalassa.com/reflections-on-the-2025-psychosis-innovation-summit-in-boston/ Wed, 16 Jul 2025 21:27:22 +0000 https://michaelhalassa.com/?p=794 I had the privilege of attending and in part organizing the inaugural Innovation in Psychosis Therapeutics Summit in Boston (June 9-11, 2025). This intimate gathering brought together scientists, clinicians and biotech leaders who share a common goal: to finally push psychiatric drug discovery into the 21st century by moving beyond existing preclinical models, expand frameworks beyond dopamine and aspire towards biomarkers embracing the digital revolution.

Thanks for reading! Subscribe for free to receive new posts and support my work.

Day 1: Systems Neuroscience Enters the Room

The summit opened with workshops, one of which I co-organized with Mikhail Kalinichev (Neurosterix Therapeutics, Geneva) and Rouba Kozak (Foundation for the National Institutes of Health). Mikhail brings deep experience from his years at GSK, Lundbeck, and Ipsen, having helped drive clinical development in schizophrenia-related cognitive impairment. Rouba previously led programs developing many neuro/psych relevant compounds with a focus on precision medicine.

Our workshop centered on how systems neuroscience and computational models of circuit dysfunction—what I’ve called “algorithmic circuit psychiatry”—can refine target selection, inform biomarker development, and guide trial design. This perspective has been underrepresented in drug development but is increasingly seen as essential for psychiatric disorders, where symptom clusters often reflect underlying circuit-level failures. See my earlier post here

A central question emerged: Should biomarker development occur during phase 1-3 clinical trials, or wait until after drug approval? The tension reflects the translational challenge in psychiatric drug development. Without predictive biomarkers, clinical trials must balance stringent selection criteria against practical timelines, often leaving drug developers flying blind.

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2Fdb19fe09 eb18 4eb4 8b72

Left to right: Michael Halassa, Mikhail Kalinichev, Rouba Kozak

 

Following our session, Paulo Lizano from BIDMC led a compelling workshop on integrating patient-centered outcomes, such as quality of life measures, directly into clinical trial endpoints. As the field increasingly recognizes that pure symptom scales may not fully capture meaningful change, incorporating functional and experiential outcomes feels both scientifically valid and deeply patient-centered.

Day 2: The Rise of Muscarinic Agents

By the second day, one theme dominated: muscarinic receptor agonism has become the most exciting frontier in schizophrenia drug development.

The KarXT/Cobenfy Story

Steve Paul delivered a masterful keynote tracing the long and improbable road to Cobenfy (xanomeline-trospium), the most important advance in psychosis pharmacotherapy since clozapine. Originally developed as a derivative of arecoline (from betel nut) for Alzheimer’s, xanomeline’s repurposing for psychosis required remarkable translational persistence.

Key highlights from Paul’s talk:

  • Exceptional CNS penetration: A 10:1 brain-to-plasma ratio, highly atypical for psychiatric agents.
  • Side effect complexity: Pro- and anticholinergic effects drive the need for more selective muscarinic agents.
  • Domain-specific effects: Cognitive gains, especially in working memory and executive function, often decouple from standard PANSS reductions.

Paul emphasized that head-to-head trials against atypical antipsychotics remain a critical next step.

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2F4608e43a a468 4a5b 992a

Steve Paul being introduced by Murali Gopalakrishnan

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2F4ea6e92d 2b46 4032 9c42

Steve Paul gave an inspiring and forward-looking talk

From Algorithm to $14 Billion

Andrew Miller, one of Karuna’s original R&D leads, described how 7,410 compound combinations were screened through an algorithm-driven selection process, ultimately resulting in the xanomeline-trospium pairing. Starting with just $4,000 (and critical support from the Wellcome Trust), Karuna’s journey culminated in its $14 billion acquisition by Bristol Myers Squibb—a biotech success story that underscores how unconventional mechanisms can lead to transformative breakthroughs.

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2F0a34d262 0b2e 4677 9c46

Andrew Miller gave a superb talk about the development of KarXT/Cobenfy

Emerging Themes: Biomarkers, Stratification, and Digital Tools

Throughout the summit, multiple sessions drilled into the next central challenge: how to stratify patients, measure meaningful improvement, and de-risk clinical development earlier with predictive biomarkers.

Biomarkers: The Missing Link—and what’s up with Emraclidine?

The panel featuring Nick Brandon (Neumora), Hadile Ounallah-Saad (Clexio), Larry Park (AbbVie), and Rob Goldman (MapLight Therapeutics) emphasized that lack of biomarkers has plagued psychiatric drug development for decades. AbbVie made clear they are not abandoning Emraclidine despite Phase II setbacks. They attribute the setback primarily to suboptimal trial design rather than flaws in the compound itself—a refreshingly sophisticated view that acknowledges how heterogeneity, endpoint selection, and sample stratification can easily derail psychiatric trials.

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc295f932 a279 470b 9b63

Right to Left: Nick Brandon (Neumora), Hadile Ounallah-Saad (Clexio), Larry Park (AbbVie), and Rob Goldman (MapLight Therapeutics)

I was particularly happy with the opportunity to discuss early clinical real-world experience with Cobenfy. Having the people who developed this compound in the same room was surreal!

Digital Interventions for Negative Symptoms

Click Therapeutics showcased their work on digital interventions targeting experiential negative symptoms. Their augmented reality glasses, designed to enhance social salience during simulated interactions, represent a highly innovative, non-pharmacological circuit intervention.

A person standing in front of a screen AI-generated content may be incorrect.

The wonderful click team

I had particularly rewarding conversations with Click’s CMO, Shaheen Lakhan, about converging digital therapeutics with systems neuroscience. We’re now collaborating on a forthcoming article about the future of inpatient psychiatry, combining digital augmentation with circuit-level mechanistic frameworks. Stay tuned!

Novel Biology: Retroviruses & Neuroplastogens

  • Jonathan Javitt presented provocative data implicating endogenous retroviral elements (HERV-W-ENV) in schizophrenia pathophysiology—a reminder that viral mechanisms may still yield unexpected insights.
  • Rajiv Agrawal (Deluxe Therapeutics) introduced DLX-2270, a neuroplastogen targeting synaptic vesicle protein 2A density. While neuroplasticity-based interventions (e.g., ketamine in depression) have gained traction, their application in psychosis remains early-stage.

Long Acting Injectables (LAIs) and Early Intervention

Hannah Brown delivered an excellent talk on the value of long-acting injectables (LAIs) in first-episode psychosis, referencing mirror-image study designs like PRELAPSE. Reconnecting with Hannah—whom I trained alongside in residency—was a highlight, and her combination of methodological rigor and clinical pragmatism was outstanding.

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2F18e58832 9122 4d6f 9a38

Hannah Brown did a masterful job at emphasizing the importance of LAIs

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff85c94ef 52b9 495f 8145

Was great to connect after 12 years!

The Takeaway: A Field at Inflection Point

The closing day featured Hadile Ounallah-Saad’s exceptional talk on developing an M1/M4 agonist (CLE-905) with different properties than Xanomeline and one that may not need a peripheral antagonist like trospium! If this ends up panning out in human trials, it will be game-changing and will give us a shot at a muscarinic LAI! Go Clexio!

A person standing at a podium in front of a large screen AI-generated content may be incorrect.

Hadile Ounallah-Saad giving a talk about Clexio’s lead compound CLE-905

 

My friend Zhong Zhong, CMO for OVID therapeutics gave a superb talk on KCC2 activators as a potential class of antipsychotic medications. My lab is involved in the preclinical side of this endeavor so we’re all hopeful this will end up having some clinical utility!

https%3A%2F%2Fsubstack post media.s3.amazonaws.com%2Fpublic%2Fimages%2Fea8dab3e 1db5 4121 8200

Zhong talking about KCC2 modulators as disease modifying agents in schizophrenia

Throughout the summit, there was palpable energy. After decades of stagnation, the field appears to have finally cracked open. For too long, psychiatric drug development has been stalled by perceived biological intractability and late-stage trial failures where clinical efficacy predicted pre-clinically failed to materialize. But this summit showcased the tools now available to change that trajectory:

  • Systems neuroscience informing target selection
  • Circuit models guiding conceptual framing
  • Sophisticated biomarkers enabling patient stratification
  • Digital augmentation expanding therapeutic options
  • Novel molecular targets beyond dopamine

The muscarinic wave, exemplified by xanomeline, validates that genuinely novel mechanisms can succeed in psychiatry, while also reinforcing that circuit-level understanding must drive future drug development.

For those of us working in systems and circuits, it seems obvious that this is the type of neuroscience that pharma needs. This is but one of several facets of the algorithm circuit framework we have been working on for the last several years.

I left Boston energized and cautiously optimistic that we are witnessing the dawn of psychiatry’s own precision medicine revolution. The science is finally catching up to the clinical complexity. Now we need the discipline, patience, and collaborative ambition to build on this momentum.

A person and person smiling for a picture AI-generated content may be incorrect.

This piece reflects my experience co-organizing the Systems Neuroscience workshop at the 2025 Innovation in Psychosis Therapeutics Summit and observations from three days of presentations, panels, and discussions with leading researchers, clinicians, and industry executives.

]]> Systems Neuroscience as a Foundation for Psychiatric Drug Discovery https://michaelhalassa.com/systems-neuroscience-as-a-foundation-for-psychiatric-drug-discovery/ Fri, 11 Jul 2025 04:55:38 +0000 https://michaelhalassa.com/?p=782 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.

]]> The Brain’s Confidence Problem: New Insights into Schizophrenia from an Unexpected Source https://michaelhalassa.com/the-brains-confidence-problem-new-insights-into-schizophrenia-from-an-unexpected-source/ Thu, 10 Jul 2025 07:49:52 +0000 https://michaelhalassa.com/?p=777 As a psychiatrist who treats patients with schizophrenia, I’ve long been struck by a fundamental puzzle: why do individuals with psychosis hold onto beliefs with such unwavering certainty, even when presented with compelling contradictory evidence? The answer, it turns out, may lie in a marble-sized brain region most people have never heard of—and the revelation comes from an entirely unexpected source.

When Tremor Treatment Accidentally Illuminates Psychosis

A groundbreaking study by Mackenzie et al. (2025, bioRxiv) has provided some of the strongest evidence yet for a circuit-level understanding of belief formation and revision. The researchers weren’t studying schizophrenia at all—they were investigating patients receiving focused ultrasound treatment for essential tremor. But when post-surgical brain swelling accidentally affected the mediodorsal (MD) thalamus, something remarkable happened: patients developed a specific pattern of overconfident decision-making that mirrors core features of delusional thinking.

Using a sophisticated behavioral task that probes how people balance exploiting known information versus exploring new possibilities, the researchers found that MD disruption led to a precise computational deficit: patients lost their capacity for adaptive doubt. They became overly confident in their existing beliefs and stopped seeking information that might challenge those beliefs—the very cognitive pattern we see in psychotic disorders.

The Neurobiology of Certainty Gone Wrong

This finding connects directly to my clinical experience treating patients with schizophrenia. In my practice, I’ve observed that the challenge isn’t simply that patients hold false beliefs—it’s that they hold beliefs with pathological certainty. The traditional psychiatric focus on the content of delusions may be missing the more fundamental issue: a breakdown in confidence calibration.

The MD thalamus appears to act as a critical “confidence regulator” in the brain’s decision-making networks. When functioning normally, it helps determine how much we should trust our own predictions versus remaining open to new information. This circuit-level understanding aligns with emerging theoretical frameworks about how the brain coordinates distributed computations for flexible cognition (Scott et al., 2024).

Consider the implications: if the thalamus normally helps us maintain appropriate uncertainty about our beliefs, then thalamic dysfunction could explain why patients with schizophrenia often exhibit such rigid certainty in their delusional beliefs. They haven’t simply acquired false information—they’ve lost the neural capacity to doubt what they think they know.

From Confidence to Delusions: A Circuit-Based Understanding

The Mackenzie study reveals something crucial about the computational nature of belief updating. When MD-prefrontal circuits were disrupted, patients didn’t simply become perseverative or confused. Instead, they showed a specific pattern:

  • Increased reward sensitivity: Greater influence of learned values on choices
  • Eliminated exploration bonus: Loss of information-seeking behavior
  • Overexploitation: Excessive reliance on existing knowledge
  • Reduced directed exploration: Failure to investigate uncertain but potentially informative options

This behavioral signature maps remarkably well onto what we observe clinically in psychotic disorders. Patients with delusions often show:

  • Pathological certainty in false beliefs despite contradictory evidence
  • Reduced information-seeking that might challenge their beliefs
  • Overreliance on internal models rather than external feedback
  • Failure to update beliefs when environmental contingencies change

The convergence is striking and suggests we may be looking at the same underlying computational dysfunction from different angles—one measured in the laboratory, the other observed in the clinic.

The Promise of Circuit-Based Psychiatry

This research opens exciting possibilities for precision approaches to treating schizophrenia. Rather than the broad neurochemical interventions we currently rely on, we might be able to target specific computational dysfunctions in thalamocortical circuits.

The study’s anatomical precision is particularly encouraging. The behavioral effects correlated specifically with disruption of the lateral (parvocellular) MD, which connects primarily to dorsolateral prefrontal cortex and frontal pole—regions critical for cognitive flexibility and belief updating. This anatomical specificity suggests that focused neuromodulation approaches could potentially restore more adaptive confidence calibration without affecting other brain functions.

Clinical Implications for Treatment

In my practice, I’ve been developing approaches that integrate computational insights with traditional psychiatric care. The MD thalamus findings suggest several potential therapeutic directions:

  1. Targeted Neuromodulation: Technologies like focused ultrasound or deep brain stimulation could potentially modulate MD activity to restore appropriate exploration-exploitation balance.
  2. Confidence Calibration Training: Cognitive interventions could be designed specifically to help patients develop more accurate metacognitive awareness of their own uncertainty.
  3. Precision Diagnostics: Computational tasks like the restless bandit could help identify specific cognitive profiles and guide personalized treatment approaches.
  4. Early Intervention: Understanding confidence miscalibration as a core deficit could lead to earlier detection and intervention before full psychotic episodes develop.

Beyond Schizophrenia: A New Framework for Mental Health

The implications extend beyond schizophrenia to other conditions where belief updating goes awry:

  • Depression: May involve underconfidence leading to learned helplessness
  • Anxiety disorders: Could reflect miscalibrated threat assessments
  • Substance use disorders: Might involve overconfidence in drug-related beliefs
  • Obsessive-compulsive disorder: May reflect inability to achieve confidence in safety

This represents a fundamental shift from thinking about psychiatric symptoms as categorical disease states toward understanding them as specific computational dysfunctions in learning and decision-making algorithms.

The Clinical Reality: From Lab to Bedside

As someone who treats patients with schizophrenia daily, I’m acutely aware of the challenges in translating neuroscience findings into clinical practice. However, this study is particularly compelling because it provides causal evidence in humans—not just correlational findings from neuroimaging studies.

The patients in the Mackenzie study didn’t lose their ability to learn or make decisions entirely. They maintained overall task performance while showing specific deficits in belief updating and uncertainty management. This selectivity suggests that interventions targeting MD-prefrontal circuits might improve cognitive flexibility without causing global cognitive impairment.

Looking Forward: A Personal Perspective

For me, this research represents something I’ve been working toward throughout my career: a true bridge between basic neuroscience and clinical psychiatry. The fact that these insights emerged from a completely different clinical context—tremor treatment—underscores how interconnected our understanding of brain function really is.

In my clinical work, I’ve seen how traditional approaches to schizophrenia, while helpful, often fall short of fully restoring cognitive flexibility and adaptive functioning. Understanding the neural basis of confidence calibration offers hope for more targeted, effective interventions.

The convergence between this human lesion study and years of animal research on thalamic function (including work from our lab and others) gives me confidence that we’re identifying fundamental principles of brain organization rather than isolated curiosities. When different methodologies and species point toward the same underlying mechanisms, it usually means we’re onto something important.

The Road Ahead

Several critical questions remain:

  1. Reversibility: Can confidence calibration deficits be restored through targeted interventions?
  2. Early detection: Could computational tasks identify at-risk individuals before psychotic episodes?
  3. Personalized medicine: How can we match specific circuit dysfunctions to optimal treatments?
  4. Combination approaches: How might neuromodulation combine with cognitive and pharmacological interventions?

As we move forward, the goal isn’t to replace current treatments but to enhance them with circuit-based insights. The patients I treat deserve approaches grounded in rigorous understanding of how their brains actually work, not just symptomatic management.

Conclusion: When Certainty Becomes the Enemy

The patients in the Mackenzie study teach us something profound about the nature of adaptive cognition: the capacity to doubt ourselves, when doubt is warranted, may be one of our most important mental faculties. When that capacity is lost—whether through thalamic dysfunction, psychiatric illness, or other causes—we become trapped by our own certainty.

This has broader implications beyond psychiatry. In an era of polarization and “alternative facts,” understanding the neural basis of belief formation and revision is more important than ever. The same circuits that go awry in schizophrenia may also be involved in more everyday forms of rigid thinking and confirmation bias.

For my patients with schizophrenia, this research offers something precious: hope for treatments based on understanding rather than trial and error. Instead of simply suppressing symptoms with broad-acting medications, we may soon be able to restore the specific cognitive functions—like appropriate confidence calibration—that enable adaptive functioning in a complex, uncertain world.

The brain’s confidence problem is solvable. And that gives me confidence that better treatments are within reach.

This research builds on extensive work linking thalamic circuits to cognitive flexibility and psychiatric disorders, offering new insights into the computational basis of belief updating and its therapeutic implications.

References:

  • Mackenzie, G., et al. (2025). Focused ultrasound neuromodulation of mediodorsal thalamus disrupts decision flexibility during reward learning. bioRxiv.
  • Scott, D.N., Mukherjee, A., Nassar, M.R., & Halassa, M.M. (2024). Thalamocortical architectures for flexible cognition and efficient learning. Trends in Cognitive Sciences, 28(7), 639-652.
]]> What Cobenfy Reveals About the Future of Psychiatry https://michaelhalassa.com/cobenfy-future-psychiatric-drug-development/ Wed, 14 May 2025 16:16:37 +0000 https://michaelhalassa.com/?p=769 What Cobenfy Teaches Us About the Future of Psychiatric Drug Development

The approval of Cobenfy (xanomeline-trospium) marks more than just the arrival of a new antipsychotic. It’s a signal—subtle but decisive—that the center of gravity in psychiatry may be shifting. For the first time in decades, we have a schizophrenia treatment that doesn’t target dopamine. That’s a headline, yes. But the real story is what it reveals about where psychiatric drug development may be headed next.

Beyond Dopamine: A New Kind of Intervention

Cobenfy works through muscarinic receptors—part of the cholinergic system—offering a pathway into domains like cognitive function and motivational engagement that have historically been resistant to treatment. In our own inpatient setting, we’ve seen cases where patients previously unresponsive to dopamine-targeting drugs showed marked improvement with Cobenfy, particularly in domains long dismissed as untreatable “negative symptoms.”

But the most important question isn’t does it work—it’s why. That’s where the model breaks open.

From Observation to Mechanism: Backtranslation Begins

When a drug like Cobenfy produces unanticipated gains, it creates a unique opportunity for backtranslation. In my lab, we’ve turned to animal models to explore how muscarinic modulation affects frontal cortical circuits responsible for sustained engagement, rule-based behavior, and decision-making. These are the very domains patients often describe as being “offline” in psychosis—not overt hallucinations or delusions, but the feeling of not being plugged in to the world’s expectations.

To get at this, we’re using single-cell resolution recordings from the frontal cortex to ask how muscarinic agents reshape cognitive control. But our aim isn’t just cortical. Informed by our DBS work and broader computational frameworks, we’re examining the role of thalamocortical circuits, especially the mediodorsal (MD) thalamus, in mediating engagement and belief updating. The thalamus, far from being a passive relay, appears to act as a contextual inference engine—a kind of hidden state generator that supports flexible cognition.

If Cobenfy has a circuit-level signature, we suspect it may involve unlocking thalamocortical loops that promote alignment with the task at hand—a kind of neurobiological re-engagement with reality.

Cobenfy as a Test Case for Algorithmic Psychiatry

These observations speak to a larger need: psychiatry can no longer afford to treat molecules and symptoms as though they exist in isolation. We need a framework that links interventions at the molecular level to computational operations of the brain, and from there to clinical manifestations.

This is the project of algorithmic psychiatry—a model I explore in a recent co-authored perspective in Cell Reports Medicine. There, we argue that psychiatric disorders may be best understood as failures of computation, not chemical imbalances: disruptions in belief updating, loss of control over long-range planning, breakdowns in state estimation. Molecules like Cobenfy are useful not just because they help some patients, but because they force us to refine our understanding of which computational processes have been rebalanced.

The Path Ahead: From Circuits to Treatments

Cobenfy is not a miracle drug—but it is a conceptual milestone. It breaks the mold and gives us a reason to dig deeper into the neural systems it affects. More importantly, it gives us a reason to restructure how we develop and evaluate psychiatric treatments—starting from circuits and working outward.

If dopamine was the story of the 20th century in psychiatry, muscarinic modulation—and the circuit-level insight it demands—may be one part of the 21st

]]> Rewiring Psychosis: How Neuromodulation Is Shifting the Schizophrenia Treatment Paradigm https://michaelhalassa.com/rewiring-psychosis-how-neuromodulation-is-shifting-the-schizophrenia-treatment-paradigm/ Wed, 14 May 2025 16:06:51 +0000 https://michaelhalassa.com/?p=766

Schizophrenia, Circuits, and the Case for Neuromodulation

In schizophrenia, the gap between what we know and what we can treat is wide. While dopamine-blocking drugs help with positive symptoms like delusions and hallucinations, they often leave patients stuck—disengaged, cognitively flattened, and unable to rejoin their lives. These are not just medication side effects. They reflect a deeper circuit-level pathology we’ve yet to fully grasp, let alone treat.

That’s why a recent study from Johns Hopkins University on deep brain stimulation (DBS) in treatment-resistant schizophrenia (https://www.medrxiv.org/content/10.1101/2025.04.09.25325419v1) is important.

From the Basal Ganglia to the Mediodorsal Thalamus

Most neurostimulation studies in schizophrenia have chased symptoms by targeting structures downstream: the nucleus accumbens, anterior cingulate, habenula. The Hopkins group tried something different. They went upstream.

Their target was the substantia nigra pars reticulata (SNpr)—a GABAergic output nucleus of the basal ganglia—and their hypothesis was bold: by stimulating SNpr, they could indirectly normalize activity in the mediodorsal thalamus (MD), which then projects to prefrontal areas involved in internal narrative, belief evaluation, and cognitive flexibility.

This is the same MD thalamus our lab has studied for over a decade—an area that helps prefrontal cortex update inferences based on context. It’s a structure we believe sits at the heart of psychosis: when it fails, belief updating can go haywire, creating fixed delusions or disordered thought.

In the Hopkins study, the researchers observed that relief from auditory verbal hallucinations (AVH) correlated with stimulation sites that were structurally connected to the angular gyrus, precuneus, and supramarginal gyrus, and functionally connected to the MD thalamus, orbitofrontal cortex, and dorsolateral PFC. In other words: the circuit matters. It’s not about shutting down one region. It’s about recalibrating a distributed network.

Why This Matters

The implications are profound. First, it reorients DBS in schizophrenia away from crude anatomical targets and toward network-informed precision. Second, it reinforces the idea that symptoms like hallucinations emerge not from hyperactive “voice centers,” but from failures in higher-order control—failures of gating, updating, and internal monitoring.

It’s also a reminder that the basal ganglia-thalamocortical loop, long studied in motor and motivational systems, may be just as central to psychosis. And importantly, it opens the door to interventions that don’t rely on blocking dopamine—interventions that might actually re-enable learning in brains locked into maladaptive internal narratives.

Non-Invasive Options: The Road Ahead

Of course, DBS isn’t scalable for most patients. But the logic behind it—targeting circuits, not just symptoms—extends to non-invasive tools.

Transcranial Magnetic Stimulation (TMS) has been used to target dorsolateral PFC in negative symptom treatment, with modest but real results. Theta-burst TMS and deep TMS aim to access deeper or more complex circuits, but their effects remain variable.

More recently, focused ultrasound (tFUS) has emerged as a way to non-invasively stimulate deep brain structures—including the thalamus. Preclinical studies suggest tFUS can alter prefrontal-striatal-thalamic dynamics in animal models of schizophrenia. In principle, it could access hubs like the MD thalamus or even modulate the same nigrothalamic loops targeted by DBS, without surgery.

From Circuits to Computation

All of this underscores a broader shift: away from neurotransmitters as endpoints, and toward circuit-level computation. The brain isn’t just a set of chemical gradients. It’s a machine that builds models of the world and updates them based on surprise.

Psychosis, then, may not reflect excessive dopamine as much as faulty inference: a thalamus that fails to signal uncertainty, a cortex that overcommits to priors, a basal ganglia that no longer weighs competing predictions.

Neuromodulation—whether invasive or not—offers a way to restore the computational dialogue between regions, to re-enable plasticity rather than dampen symptoms. And it pushes psychiatry toward a future where interventions are judged not just by where they act, but how they change the brain’s ability to learn.

]]> Why Mental Health Treatments Need Algorithmic “Flight Simulators https://michaelhalassa.com/why-mental-health-treatments-need-algorithmic-flight-simulators/ Tue, 06 May 2025 17:29:43 +0000 https://michaelhalassa.com/?p=764 Why Mental Health Treatments Need Algorithmic “Flight Simulators”

I would like to highlight a recent perspective article that my colleagues and I have recently published: https://www.cell.com/cell-reports-medicine/fulltext/S2666-3791(25)00167-3

Despite decades of research and billions in global spending, mental health treatment remains an incomplete puzzle. Nearly one-third of people with schizophrenia remain treatment-resistant, and fewer than 15% experience functional recovery. The lack of progress in psychiatric treatment underscores a pressing need for innovative solutions. While new therapies like xanomeline/trospium—a recently FDA-approved drug targeting muscarinic receptors rather than dopamine receptors—offer fresh hope, they only highlight a more profound issue in the field. Simply put: We still lack a framework for predicting how molecular interventions impact cognition and behavior.

In traditional psychiatry, treatments such as pills, therapy, and brain stimulation are often designed in isolation, targeting one specific aspect of the brain’s complex network. For example, a schizophrenia drug may target dopamine receptors, but it doesn’t predict how its effects will cascade through neural circuits to influence decision-making, emotion regulation, and belief systems. Similarly, cognitive therapies are based on the idea of flexible thinking but fail to account for how underlying molecular deficits may limit their effectiveness.

This disconnect is psychiatry’s fundamental challenge: treatments are designed for isolated components, while their effects unfold across all levels of the brain’s functioning, from molecular mechanisms to cognitive behaviors. This is where algorithmic psychiatry steps in, offering a computational “flight simulator” that can model how perturbations at any level—molecular, circuit, or cognitive—affect the entire system.

What is Algorithmic Psychiatry?

Algorithmic psychiatry combines data from behavioral tasks and neural signals (from EEG, fMRI, or intracranial recordings) to model the brain’s internal processes. These models focus on hidden variables, such as how the brain updates beliefs and expectations, as well as its ability to adjust predictions in response to new information. These variables are what drive psychiatric symptoms, and from a treatment perspective, recalibrating them is the key to success.

For example, imagine a treatment designed to enhance the brain’s ability to predict sensory inputs more accurately, helping to reduce hallucinations in schizophrenia. Another approach might focus on lowering overconfidence in rigid memories, which would improve cognitive flexibility and reduce symptoms of rigidity and paranoia. Together, these interventions aim to re-wire the brain’s internal algorithms, addressing the root causes of symptoms rather than just masking them.

The Promise of Multi-Level Interventions

In algorithmic psychiatry, success is not just about reducing symptoms—it’s about recalibrating the brain’s internal computational processes. It involves not just one intervention, but a combination of approaches that work across different levels of brain function. For instance, pairing a drug that enhances a neurochemical feature with targeted neurostimulation can enhance specific circuits, thereby boosting the drug’s effects. When combined with behavioral therapies that are timed appropriately, this multi-level approach has the potential to rewire the brain for recovery.

This approach goes beyond symptom control—it focuses on designing treatments that consider the entire biological, cognitive, and neural network. The idea is that by interacting with the brain’s “software” (its internal computations) and improving its “hardware” (its neurochemical and neural networks), we can create truly transformative treatments for mental health.

The Road Ahead for Algorithmic Psychiatry

The goal of algorithmic psychiatry is to create a precision psychiatry model, where treatments are individualized based on how each patient’s brain is wired. This model offers new hope for those suffering from chronic and treatment-resistant conditions like schizophrenia. Instead of simply targeting symptoms with broad drugs, this approach focuses on understanding and recalibrating the brain’s underlying computations.

While this vision of “flight simulator” models is still evolving, the potential for better-targeted treatments is already within reach. With advancements in computational neuroscience, machine learning, and neurostimulation, we are beginning to see the first real glimpses of how these multi-level, algorithmic treatments could revolutionize mental health care.

As we continue to refine these models and develop new technologies, we move closer to the promise of a psychiatry that is not just based on treating symptoms, but on re-wiring the brain for a more functional and flexible future.

]]>