Let’s talk science!

Written By Joe Kleman

The Cube’s Somatic Play Therapy and Active Inference: A Conceptual Link

Introduction

Trauma recovery therapies are increasingly integrating insights from neuroscience to improve outcomes. One innovative approach is The Cube, a play-based somatic movement therapy designed to help individuals “reset” after trauma in just 10-minute sessions. The Cube combines mental stimulation, targeted physical movements, and playful interactions within a safe, compassionate setting to facilitate healing. On the other side, Active Inference is a theoretical framework developed by Karl Friston that describes how the brain operates as a prediction machine – constantly trying to minimize the mismatch between its expectations and incoming sensory inputs (i.e. prediction errors) by updating its beliefs or taking action in the world. This principle is rooted in Friston’s free energy principle, which posits that organisms survive by reducing surprise (free energy) through perception and action. In this report, we will explain The Cube’s therapeutic approach and Friston’s active inference theory, then explore how they intersect conceptually. We will examine how somatic movement and play might reduce prediction error, how The Cube’s embodied practices align with the free energy principle, how its mental and relational aspects map onto active inference, and highlight any research bridging somatic trauma therapies with active inference. Notably, even Karl Friston has recognized the creator of The Cube, Joe Kleman, as an “innovator in the field of predictive processing and Active Inference” with a deep practical understanding of these concepts – underscoring the rich conceptual links we discuss below.

The Cube: Play-Based Somatic Movement Therapy for Trauma Recovery

Figure: A participant engages in The Cube’s physical play environment. The Cube apparatus is a room-sized frame with elastic cords and colorful nodes, allowing guided movement games. Through dynamic balancing and reaching movements, the participant navigates playful challenges. This engaging, full-body activity is delivered in a compassionate, safe setting to help process trauma.

The Cube’s trauma therapy is a play-based, somatic movement intervention focused on quick but potent sessions. Each session (about 10 minutes) involves the individual interacting with a specially designed environment – essentially a “cube” space often augmented with visual or tactile elements – to perform guided movements and solve simple playful tasks. Key features of The Cube’s approach include:

  • Engaging Playful Tasks: Clients participate in game-like activities (e.g. navigating obstacles, catching objects, free-form play) that are fun and absorbing. The element of play creates a non-threatening atmosphere where learning and exploration can occur naturally.

  • Targeted Physical Movements: The therapy prompts specific body movements – reaching, balancing, shifting gaze, etc. – that are believed to help “move through and heal from trauma” by engaging the sensorimotor system. Somatic (body-based) involvement is central, acknowledging that trauma isn’t just cognitive but also “lives” in the body (e.g. in muscle tension, posture, reflexes).

  • Mental Stimulation and Puzzles: Clients also face cognitive challenges or puzzles during the play. This might include problem-solving, pattern recognition, or memory elements introduced in the game. Such mental stimulation ensures the mind is actively engaged alongside the body, potentially addressing trauma-related thought patterns while the movements address bodily tension.

  • Compassionate, Safe Engagement: The Cube is delivered in a “playful, supportive, and enriching environment”. Therapists or facilitators provide a compassionate presence, ensuring the client feels safe and understood. This safety is crucial for trauma survivors – it counteracts the hypervigilance and distrust that trauma can instill, and encourages the person to explore and take small risks in the therapeutic exercises without fear of judgment or harm.

  • Rapid Trauma “Reset”: The combination of these elements is intended to produce “immediate results” in helping patients work through distress. By the end of a short session, clients often report reduced stress or emotional release, as if a mental “reset” button was pressed. Repeated sessions aim to consolidate these gains into lasting recovery.

In practice, a client in The Cube might, for example, wear augmented-reality glasses and see virtual prompts or artistic visuals in the physical cube frame, asking them to reach for certain colored targets or respond to surprising changes in the environment. Throughout, they remain physically active and mentally engaged, while the therapist provides encouragement and ensures the challenge level stays tolerable. The experience is often described as “patient-led” and yielding immediate insights, as clients solve tasks in a judgment-free play zone and simultaneously learn about their own reactions and thought patterns. Importantly, The Cube provides not just an activity but also a framework to understand “why they might be struggling” and a “clear plan for moving forward”, tying each exercise to personal themes in the client’s life. In simpler terms, it’s a holistic tool to help mind, body, and emotions come back into alignment after being thrown off by trauma.

Crucially, The Cube’s design is informed by modern neuroscience – particularly the concept of the predictive brain. The creator, Joe Kleman, built The Cube around principles of predictive processing, which is closely related to Friston’s active inference theory. In a technical description, The Cube “titrates prediction error — introducing small, manageable mismatches between expectations and reality” during the play exercises. By carefully controlling the degree of surprise a client encounters, the system engages the brain’s natural learning processes without overwhelming the client. For example, a task may subtly violate the client’s expectation (perhaps a ball doesn’t move as expected or a visual element changes unpredictably) just enough to get their brain’s attention. These surprises activate introspective neural networks – “engaging the brain’s Default Mode Network (DMN), which processes self-referential thinking and emotional memory”. In essence, The Cube leverages the brain’s own predictive modeling system to help the person reflect on their reactions and gently revise the mental models that were shaped by traumatic experiences. This innovative blending of play therapy with predictive processing is what caught the attention of Karl Friston and others in the neuroscience community.

Active Inference and the Free Energy Principle: The Brain as a Prediction Machine

Active Inference is a theory from computational neuroscience that provides a unifying explanation for perception, action, and learning in the brain. At its core is the idea that the brain is not a passive receiver of sensory information, but an active predictor. According to this framework (and the related predictive processing or predictive coding theories), the brain continuously generates internal models of the world and of the body, and uses these models to predict what sensory inputs it should be receiving at any moment. For example, your brain has a model that predicts the feel of the floor under your feet when you stand, or the sound your phone makes when it rings. When the actual sensory input matches the prediction, everything feels normal and the world is as expected. When there’s a mismatch – say you expect a friend to greet you but they ignore you, or you expect the ground to be solid but it suddenly shakes – the brain registers a prediction error. This prediction error is essentially surprise or uncertainty that needs to be resolved.

Active inference refers to the brain’s strategy for minimizing these prediction errors (or equivalently, minimizing “free energy” in Friston’s terms). There are two broad ways the brain can reduce a prediction error:

  1. Update the Internal Model (Perceptual Inference): The brain can revise its beliefs or predictions to better fit the incoming data. This is like saying “I predicted wrong, let me adjust my expectations.” For instance, if you expected your friend to greet you and they didn’t, you might update your model (perhaps thinking “maybe they didn’t see me” or “sometimes people are preoccupied”). In neural terms, this involves adjusting synaptic weights or neural firing patterns to encode the new information. This is essentially learning – each time a prediction error occurs, it’s an opportunity for the brain to learn and refine its model of the world.

  2. Act to Change the Sensory Input (Active Inference in action): The brain can also take action to make the world align with its predictions, thereby eliminating the error. Using the same example, if you expected a greeting, you might actively wave or say hello first – changing the interaction so that your prediction of engagement is fulfilled. In other words, the organism can “selectively seek and sample sensory information to fulfill its predictions”. This is a hallmark of active inference: actions are driven by the desire to confirm our brain’s predictions or to bring sensory input back in line with what we expected. Friston famously gave an example of a baby “fulfilling” its prediction of a sweet taste by crying for milk and then sucking on a pacifier, effectively using action to satisfy its predicted sensation. In summary, under active inference, behavior itself is in service of reducing prediction errors.

This constant looping of prediction -> comparison with input -> error -> adjustment (by learning or acting) is thought to be how the brain works at every level, from basic reflexes to complex planning. The free energy principle is a broad theoretical claim that living systems maintain their order and resist entropy by minimizing a quantity called free energy, which in these models is equivalent to “surprise” or prediction error. In simple terms, an organism survives by avoiding big surprises – it tries to keep its sensory input within expected bounds (e.g., not too hot, not too cold, food arrives when hungry, etc.). If something unexpected happens, the organism either changes its understanding or changes its situation to remove the unexpected element. This principle can be applied to physiological regulation (like sweating to reduce the surprise of overheating) as well as to perception and behavior.

Active inference, therefore, provides a powerful lens to understand behavior and also maladaptive states. In the context of trauma, predictive processing research suggests that traumatic experiences can distort the brain’s internal models. For instance, after traumatic stress, a person’s prior beliefs (deep expectations) might become biased toward expecting danger everywhere. Their brain might continually predict threat or betrayal, even in safe environments, because that model was “learned” during trauma. This can lead to constant prediction errors – the world is actually safer than they expect, but they cannot easily update that belief. It can also lead to pathological active inference behaviors – for example, a trauma survivor might engage in hypervigilant scanning of the environment or avoid social interaction, in effect acting as if threats are present, which sadly may reinforce their sense of danger (since they never receive disconfirming evidence). From an active inference standpoint, PTSD and related conditions can be seen as the brain being “stuck” in a maladaptive predictive state: the person’s model of the world (and of self) generates lots of prediction errors (because it’s out of alignment with current reality), yet the model isn’t getting properly updated due to various factors like overwhelming fear, extremely “strong” prior beliefs, or lack of safe experiences to learn from. This perspective is useful because it points to a key goal for therapy: recalibrate the predictive model. In other words, help the person update their deep expectations about safety, self, and others so that their brain’s predictions match the real, hopefully safer, world more closely – thereby lowering chronic prediction error and freeing them from constant surprise and stress.

Somatic Movement and Play: Reducing Prediction Errors Through Experience

One of the most direct intersections between The Cube’s method and active inference theory is the use of somatic movement and play to modulate prediction error. The Cube deliberately introduces manageable surprises in a playful, bodily way, which is essentially a form of controlled prediction error. The rationale behind this can be illuminated by looking at what play means in a predictive processing context. According to a recent cognitive theory of play, play behavior is actually the brain’s way of seeking out surprises in order to master them. Andersen et al. (2023) describe that in play, an agent “deliberately seeks out or creates surprising situations that gravitate toward sweet-spots of relative complexity with the goal of resolving surprise.” Moreover, play feels fun or rewarding partly because “the agent is reducing significant levels of prediction error (i.e., surprise) faster than expected” during play. In other words, play is enjoyable because it lets us generate a bit of uncertainty and then promptly conquer it, giving the brain a satisfying drop in free energy. This insight beautifully maps onto what The Cube does for trauma patients: it provides a safe playground for the brain to encounter small surprises and resolve them through movement and adaptation.

In a Cube session, when a client performs a physical task that is slightly unpredictable (say a ball comes from an unexpected direction or a routine is altered), their brain registers a modest prediction error – something happened that wasn’t entirely anticipated. Because the context is playful and non-threatening, this error is not perceived as a danger but as a challenge. The client can act (move their body, adjust their strategy) to respond to the surprise and succeed in the task. In doing so, they update their internal model a little bit – perhaps learning that a sudden motion in their peripheral vision isn’t always a sign of threat, or that they can regain balance after a slip. The Cube “subtly challenges predictive models” like this to encourage reflection and learning “without overwhelming the system”, allowing individuals to “reassess past experiences and emotional narratives” in light of new, safe outcomes. Essentially, each mini surprise followed by a successful action is a corrective micro-experience for the brain’s predictions. Over time, these add up to a significant revision in the person’s world-model – one that expects a bit less catastrophe and a bit more control or goodness.

Somatic movement (movement of the body) is key here because active inference emphasizes that we learn by doing. The brain’s predictions are not just abstract ideas; they are often embodied expectations about what our muscles will feel, how our heart will beat, what our eyes will see, etc. By moving, especially in novel ways, a person generates a rich stream of sensory feedback – proprioceptive (position of limbs), vestibular (balance), tactile, visual, and even interoceptive (inner body signals like heart rate changes). All this sensory information, when it slightly violates expectations, forces the brain to adjust. In trauma, there may be certain bodily movements or sensations that the person unconsciously avoids because they predict these will lead to panic or pain (for example, opening up one’s posture might make one feel vulnerable, or being touched might trigger memories). Through playful movement, The Cube coaxes the person to experience these sensations in a new, safe context, giving the brain new evidence that contradicts the old, feared prediction. This is comparable to the exposure principle in PTSD therapy: indeed, all effective trauma therapies have some form of controlled exposure at their core, which “allows for reshaping the patient’s perceptual and active inference (including interoceptive) responses by changing the external and internal context of traumatic imagery”. Traditional exposure therapy might have a patient deliberately recall traumatic memories or face feared situations in a safe environment to update their expectations. The Cube does something analogous, but through an embodied, playful paradigm – it exposes the client to sensory and situational surprises in a tolerable dose, so that they can unlearn the idea that every surprise or deviation spells disaster.

Concrete example: imagine a trauma survivor who flinches (an involuntary defensive movement) whenever something approaches them quickly, due to a history of sudden violence. In the Cube, perhaps they play a game of catch with soft balls or virtual objects that appear in peripheral vision. At first, their prediction (object coming = harm) might cause them to startle. But repeatedly, the object turns out to be harmless and even “catchable” in a fun way. Gradually, their brain updates the prediction – an object coming toward me is not necessarily a threat; I can interpret it differently and respond calmly. The prediction error (object came, no harm followed, contradicting the prior belief) is resolved by either adjusting the belief (“not all fast movements are dangerous”) or by the person’s action (catching or dodging in a controlled way that gives a sense of control rather than helplessness). In active inference terms, the person is sampling new sensorimotor evidence that re-trains their generative model towards safety. Over time, this reduces the chronic prediction error that was present in everyday life when their brain kept predicting danger inappropriately. Somatic engagement thus directly aids in error reduction: the body becomes an instrument for the mind to discover which predictions are wrong and to correct them, all in the moment.

Furthermore, because The Cube’s challenges are framed as play, they inherently have what Andersen et al. called the “sweet-spot” of complexity: not so hard as to be overwhelming (which would spike anxiety or error too high), but not so easy as to be boring (which yields no learning). This balance maintains a state of focused attention and relative enjoyment. A playful mood also likely increases dopamine and positive affect, which can signal the brain that it’s safe to update beliefs (since high stress tends to lock down our existing predictions as a defense). Thus, play reduces prediction error both in the immediate sense (by providing a mechanism to resolve surprises) and in the long-term sense (by teaching the brain new, more accurate predictions that lead to fewer errors in daily life). As one theorist put it, “playful agents may create an environment tailored to the generation and further resolution of surprise and uncertainty”, effectively using play to train their predictive models. The Cube can be seen as precisely such an environment, purpose-built for a trauma survivor to re-train their brain’s predictions about the world and their own capacities.

Embodied Practices and the Free Energy Principle: Alignment in Healing

Active inference and the free energy principle emphasize that mind-body, and environment form an integrated system for minimizing surprise. An embodied therapeutic practice like The Cube aligns naturally with this view, because it doesn’t treat the mind in isolation – it engages the whole organism in the healing process. Under the free energy principle, a person will attempt to maintain physiological and psychological homeostasis by avoiding unexpected states. Trauma throws off this equilibrium: for example, a traumatized individual often has an overactive adrenaline response (unexpected surges of fight-or-flight signals) and an expectation of danger that keeps being unmet or negatively met (they expect danger, they either find none and remain on edge, or they interpret ambiguous cues as threats, perpetuating a cycle of stress). The result is a body in continual flux, far from the low free-energy steady state of feeling safe and regulated.

Embodied therapies aim to restore a healthy equilibrium by working through the body to influence the mind’s predictions. The Cube’s exercises like “shifting between peripheral and central vision” and alternating between broad awareness and focused tasks are a good illustration. This practice forces the brain to fluidly transition between different modes of processing – one more exploratory and diffuse, the other more goal-directed and precise. In predictive processing terms, it’s like training the brain to flexibly allocate attention and interpret signals, rather than getting stuck in a rigid narrow focus (often trauma = hyper-focus on threat) or an overly diffuse, unfocused state (trauma = dissociation). By engaging both hemispheres and both attentional modes, The Cube “supports cognitive flexibility, emotional regulation, and a deeper integration of perception and movement”. This integrated perception-action training is exactly what one would design if following the free energy principle: you are teaching the system (the person) how to smoothly toggle between acting on the world and reflecting internally, so that they can always minimize surprise whether it requires an external action or an internal reappraisal.

Another embodied aspect is The Cube’s use of free association with visual and tactile elements to rebuild meaningful connections in the brain. Trauma can cause dissociation and a loss of meaning in one’s perceptions – people often report feeling detached, or that the world seems unreal or devoid of the personal significance it once had. In predictive brain terms, this can be seen as a failure of the brain to integrate sensory inputs with the appropriate context or emotional significance (a kind of prediction error where things look the same as before the trauma but feel different, because the internal model has shifted to a defensive, numb state). The Cube addresses this by having participants interact with aesthetic or symbolic elements (like art, nature scenes, or abstract shapes) and freely associate thoughts or feelings with them. This is done while the person is in a calm, playful state, so the associations are allowed to surface naturally. By doing so, “participants rebuild connections between perception and personal meaning”, essentially re-linking the sensory world with positive or normal significance. The process “leverages predictive processing to update internal models, allowing the brain to reestablish meaningful interpretations of its surroundings”. In active inference terms, the therapy is providing new sensory data (the visual elements) and encouraging the brain to actively reinterpret them in a personal way, thus updating the high-level priors about the world’s meaning. When successful, this helps the individual “move beyond dissociation” and recover a cohesive sense of reality and self. Essentially, the embodied practice of physically engaging with objects and contexts – touching them, observing them, moving around them – under guidance of free association acts as a bridge for the brain to infuse sensation with renewed context, lowering the free energy associated with a previously fragmented or meaningless world.

The Cube’s alignment with the free energy principle is also evident in how it encourages active rather than passive participation. In an active inference view, agency is crucial: the person must be an agent in reducing their own prediction errors, not just a passive recipient of new information. The somatic movements in The Cube give the person a sense of agency over their body (they literally move to solve problems) which can counteract the helplessness often learned in trauma. Research in active inference frameworks of therapy suggests that offering new bodily experiences can help “weaken prior beliefs” that were rigid (for instance, “I have no control” or “my body is broken”) and allow new, healthier priors to form. One paper by Friston and colleagues on touch therapy notes that providing novel interoceptive/proprioceptive inputs (like therapeutic touch, or by extension movement) along with guidance can promote a reinterpretation of those signals, effectively “rewiring” the generative model underlying chronic pain or anxiety. The Cube’s embodied play likely works on a similar principle: unusual but pleasant bodily sensations (e.g., the exhilaration of a quick spin, the pressure of a stretchy cord, the coordination of eye and hand) coupled with a safe context give the brain data to revise old assumptions (“when my heart races, it’s not always panic – it could be excitement from a game”; “when something approaches me, it might be a ball to catch, not a fist to fear”). Bit by bit, these new priors that include safety and competence replace the old priors rooted in trauma.

In summary, The Cube’s embodied practices – from coordinated movements to sensory engagement – are fully in line with the active inference notion that action-perception loops are fundamental to reducing free energy. By actively moving and perceiving in a controlled setting, the individual is doing exactly what the free energy principle says a self-organizing system should: acting to bring sensations in line with predictions, and updating predictions to better match reality, until a harmonious state (lower free energy state) is achieved. The “harmony” in therapeutic terms is a state of regained balance, where the client’s body is relaxed and their mind no longer constantly anticipates danger. Through embodied play, the client has effectively learned their way back to a regulated state.

Mental Stimulation and Compassionate Engagement: Updating Beliefs in a Safe Context

Beyond movement and play, The Cube also incorporates deliberate mental and social elements that can be mapped onto active inference concepts. The mental stimulation provided by puzzles or cognitive challenges in the Cube serves to engage higher-level predictions and beliefs. While the body is learning through movement, the mind is also being exercised – for example, a game scenario might involve moral reasoning or perspective-taking, which engages the person’s social cognition. The Cube’s developers highlight how it “reinforces theory of mind, moral reasoning, and empathy” by creating dynamic social environments (sometimes via augmented reality) where players must interpret others’ perspectives and emotions. This means the client is practicing updating their social predictions: they might have to guess what a character (or another player) will do, and adapt when it turns out differently. In doing so, they confront any rigid social beliefs (e.g., “people will always hurt me” or “no one understands me”) in a gentle way. The scenarios“challenge assumptions about others’ perspectives” and require “real-time adaptation in social cognition”, leading participants to adjust preconceived notions and foster a more flexible understanding of others. In active inference terms, this is an exercise in updating the generative model in the social domain – revising those high-level priors about relationships that trauma often distorts (such as mistrust or expectation of harm). By navigating make-believe social challenges and seeing them resolve safely or even humorously, the person’s brain gets to unlearn some of the hypervigilant social predictions and replace them with more nuanced, hopeful ones.

Now, a critical ingredient that makes all this possible is the compassionate engagement aspect. No matter how clever the activities, if a trauma survivor does not feel safe and supported, their brain will likely cling to its old predictions (e.g., “I’m in danger” or “I will be judged”) with high precision. The Cube emphasizes a compassionate, play-based approach – essentially providing a secure relational context for the experiment of play. From a predictive processing perspective, a strong therapeutic alliance provides an alternative prediction for the client: “I am safe here and cared for.” Over repeated interactions, this relational safety becomes a stable prior belief that can counteract the trauma-derived prior “no place is safe”. In fact, theorists have suggested that common psychotherapy techniques work by establishing exactly this kind of reliable, low-surprise relationship. For example, one analysis explains that “through a predictive processing lens, active listening strengthens the prediction of a supportive relationship and safe environment” for the client. In other words, when a therapist consistently listens and accepts the client, the client’s brain starts to predict (expect) that “I will be supported and not harmed in this interaction.” This new prediction, once trusted, can drastically reduce anxiety (since there is less uncertainty about being attacked or shamed) and allow the client to start exploring new behaviors and thoughts. The Cube’s compassionate play context likely serves the same function: the facilitator’s warmth, the playful non-judgmental vibe, and even the design of the game (which might celebrate mistakes as part of learning) all signal to the client’s brain that errors are okay and support is available. This lowers the perceived risk of updating – the person can tolerate more surprise and novel experience because a safe “buffer” is in place. With lower fear, the brain can afford to loosen the grip of old priors and consider new information (something active inference theory often frames in terms of adjusting the precision of prediction errors – safety increases our willingness to be uncertain for a moment, because we trust things will be fine).

Additionally, compassionate engagement can itself be seen as a two-way active inference process between therapist and client. The therapist is also an agent with their own brain, trying to anticipate the client’s needs and signals, and adjusting their approach (tone of voice, posture, guidance) in real-time. This synchronized dance could be considered a form of interactive active inference or an enactive coupling: both people are actively minimizing prediction errors in the relationship (the client tries to behave to keep the therapist’s support, the therapist monitors and soothes the client’s distress). When done well, this creates an attuned relationship where the client experiences new relational outcomes – for instance, expressing upset and seeing the therapist respond kindly, which might be a novel experience if in the past expressing upset led to punishment or abandonment. Each such encounter is a little violation of the client’s pessimistic social predictions and a positive update to their model of how humans can interact.

In the Cube, the presence of play likely amplifies the compassion effect because play naturally tends to be “unconditional” and process-oriented rather than outcome-oriented. There’s a sense of permission to fail in play. The client can try something and laugh if it doesn’t work, rather than feel ashamed. This aligns with trauma-informed care principles and also with keeping prediction errors at a tolerable level – a mistake is not met with criticism but maybe with humor or encouragement, so the error doesn’t register as a severe negative surprise but rather a mild, even enjoyable surprise. Thus, the mental stimulation and problem-solving remain in the realm of optimistic exploration rather than triggering feelings of defeat. From a cognitive standpoint, this may encourage the client to engage the prefrontal cortex in new ways (“What if I try this approach? Maybe that will work.”) instead of being locked in amygdala-driven fear responses. In predictive coding language, one might say the precision weighting of errors is managed skillfully: the client learns which surprises are safe and worth investigating, and which are truly threats. Over time, this can calibrate their responses outside of therapy, too – not every unexpected event will send them into panic, some will even be met with curiosity or playfulness.

Bridging the Cube and Active Inference: Emerging Insights and Research

The conceptual connections between The Cube’s somatic play therapy and active inference are not just theoretical musings – they reflect a broader movement in psychology and neuroscience to understand and design therapies in light of predictive processing. Researchers have begun explicitly linking trauma and other psychopathologies to predictive processing models. For instance, a 2024 review by Putica and Agathos reframes complex PTSD as fundamentally a disorder of disrupted predictive processing, where “prolonged trauma disrupts the brain's predictive processing” and leads to symptoms like negative self-concept, emotional dysregulation, and interpersonal problems via maladaptive prediction mechanisms. They highlight how maladaptive priors (deep expectations) formed from trauma can bias perception and interoception – e.g. the person’s brain expects negative outcomes or interprets bodily arousal as danger – resulting in a persistent misalignment between what is predicted and what is actually happening. The logical implication is that therapy must target those priors and prediction errors. The Cube is a concrete example of such an intervention: it targets the prediction errors in multiple domains (sensory, cognitive, social) and attempts to recalibrate them in a positive direction.

Moreover, active inference as a framework has inspired new ways of thinking about therapy in general. Some authors are working on integrative models of psychotherapy through the lens of predictive processing, suggesting that many common factors in therapy (like the therapeutic alliance, narrative reframing, exposure techniques) can be understood as helping the client minimize certain prediction errors or adjust the precision of their beliefs. In one such view, “psychotherapy is effective mainly because of common factors” and predictive processing has the scope to provide a unified explanation for those factors. The Cube’s approach embodies several of those factors – a strong therapeutic alliance (compassionate engagement), experiential learning (exposure through play), cognitive reframing (making sense of experiences via free association and insight), and emotional processing (safe expression of feelings during play). By mapping these onto active inference, we can appreciate that The Cube not only uses these factors but does so in a way that is mechanisticallytargeted at updating the brain’s model (something that predictive processing frameworks articulate well). For example, if we consider emotional processing, traditional therapy might have a client verbally recount a trauma and then notice that they survived (updating the model from “I’m eternally trapped in that moment” to “it’s over, I’m safe now”). In the Cube, an analogous process might occur nonverbally: the client physically enacts struggle and success in a metaphorical way, perhaps unconsciously revisiting themes of their trauma but experiencing a different outcome (control, success, support). This too updates the model, just through a more implicit, body-centered route.

There is also growing research interest in embodied or enactive therapies that resonates with active inference principles. A paper by Kim et al. (2022) – co-authored by Karl Friston – proposed an “active inference account of touch and verbal communication in therapy.” They argue that issues like chronic pain or emotional trauma involve “distorted and exaggerated patterns of interoceptive and proprioceptive inference”, and that to heal, therapy should introduce new bodily experiences while guiding the interpretation of those experiences. The process they outline is remarkably similar to what happens in The Cube: first, weaken the old priors that maintain the disorder (e.g., the certainty that any bodily sensation is dangerous) by providing disconfirming evidence in a safe context, and second, install new priors through guided experiences (e.g., experiencing one’s body as strong, or movement as joyful rather than harmful). The use of manual touch in their example could be paralleled with The Cube’s use of full-body movement and contact with objects – both are modalities of delivering corrective sensory signals. Their work suggests that even seemingly physical interventions (like an osteopathic adjustment or, say, a balancing exercise in The Cube) can have downstream effects on the mind by updating how the brain predicts bodily states. This is a clear instance of active inference theory informing practice.

Another relevant line of thought is the role of play in development and healing. Beyond the Andersen et al. cognitive theory of play, others have noted that playfulness might contribute to resilience and meaning-making. A 2022 paper on “playfulness and the meaningful life: an active inference perspective” (as hinted by search references) likely ties how being playful can help one deal with life’s uncertainties more gracefully. In therapy for trauma, reclaiming a sense of play can itself be a healing milestone – it indicates a shift from a defensive, rigidity (high free energy, high prediction error state) to a flexible, exploratory stance (lower free energy, better prediction tuning). It’s noteworthy that trauma often robs individuals of their playfulness, making them feel old beyond their years or constantly on guard. So reintroducing play in a safe way can be seen as restoring a fundamental adaptive capacity of the brain. Under active inference, a playful brain is one that is confident enough in its overall model of safety that it can deliberately entertain a bit of surprise for fun. Thus, conceptually, when a trauma survivor finds themselves able to play in The Cube – laughing, experimenting, and not panicking – it’s a sign that their latent free energy has decreased and their predictive model now expects that not every unknown is lethal. This aligns with the free energy principle’s vision of a healthy organism: adaptive, curious, and equilibrated with its environment.

To sum up the research and conceptual bridge: The Cube’s somatic play therapy exemplifies how we can apply active inference in the real world to aid trauma recovery. It shows that by engaging the body (action) and the mind (perception) together in a controlled, compassionate setting, we can facilitate the brain’s natural tendency to minimize prediction errors. Each element of The Cube can be seen through this lens – from titrating surprises to encourage model updates, to using embodied actions to directly counteract traumatic predictions, to leveraging the therapeutic relationship to set positive expectations. This intersection between a cutting-edge therapeutic tool and a leading neuroscience theory is encouraging. It not only provides a scientific rationale for why The Cube may be effective, but it also offers active inference theorists a practical validation of their ideas. As Karl Friston remarked about The Cube’s creator, this work is a case where deep theoretical knowledge is driving “real, pragmatic change” in mental health treatment. Going forward, we may see more therapies built on these principles, and more research investigating how precisely these embodied, play-oriented, prediction-focused interventions can rewrite the brain’s predictions to heal trauma. In bridging The Cube and active inference, we witness a promising convergence of somatic psychology, playful engagement, and computational brain science – all working together to help individuals regain safety and wholeness after trauma.

 

 

Conclusion

Both the Cube’s play-based somatic therapy and Friston’s active inference framework converge on a simple yet profound idea: healing comes from reducing the gap between what our brain expects and what is actually happening. The Cube achieves this through compassionate play and movement – allowing a traumatized person to safely experience surprises and learn from them – while active inference provides the conceptual language for why this works – the brain is updating itself, minimizing prediction errors and free energy. We have seen how somatic movement and play in The Cube provide corrective experiences that chip away at maladaptive predictions, how its embodied practices resonate with the free energy principle by engaging the action-perception cycle for stability, and how the therapy’s mental and social support components map to active inference by creating a safe space for belief updating. Emerging research and commentary are beginning to connect these dots, framing trauma recovery in terms of predictive processing and highlighting interventions that effectively update the brain’s model of the world. The Cube stands as a practical embodiment of these concepts – essentially an active inference playground for the wounded mind to rebuild itself. Conceptually, it illustrates that when therapy leverages the body, embraces play, and operates with compassion, it aligns with our brain’s fundamental computational strategy of active inference. By minimizing dangerous prediction errors and maximizing safe, new learning, such an approach can free a person from the predictive prison of trauma. In effect, The Cube and approaches like it allow the brain to do what it does best – adapt – and in doing so, guide the individual toward a state of regained safety, connection, and well-being, which is the ultimate low free-energy state that both neuroscience and trauma therapy strive for.

Sources:

Joe Kleman
Next

No cube?