香京julia种子在线播放

The active inference framework is based on the premises that (1) perception and action of a person self-organize to minimize a quantity known as variational free energy and that (2) action selection, planning, and decision-making can be optimized by minimizing expected free energy, which quantifies the variational free energy of various actions based on expected future outcomes (Smith et al., 2022). Infants are born with self-unity that serves as a seed (ground zero) within innate complex self-organizing processes, as if they are objective perceivers and actors that are differentiated from other entities (Rochat, 2019). Such innate self-unity can serve as a seed (prior) in the active inference framework (Friston, 2018).

According to FEP, a living organism is a self-organizing system that maintains its characteristic phenotypic states and avoids surprising deviations from these expected states by generative processes that are self-organizing and self-evidencing. As the physical, biological processes of an organism embody its “best guess” about its environments, on average and over time the organism tends be attracted to a limited number of attractor states in the space of all possible states, with low entropy or spread in the probability density over the space of possible states, i.e., low variational free energy. Variational free energy is a measure of the upper bound of surprise or prediction error—the difference between the organism’s “best guess” beliefs about what caused its sensory states and what it observes (Friston, 2013; Ramstead et al., 2020; Friston et al., 2022).

Free energy principle adopts the notion of Markov blankets to define the boundary of the living system and its environments—which are partitioned as internal (systemic) states and external (environmental) states, respectively. The Markov blanket itself can be partitioned into active and sensory states, which can be differentiated as follows: active states are not influenced by external states, and sensory states are not influenced by internal states (Friston, 2013; Ramstead et al., 2018, 2020). The internal states and its Markov blanket together constitute an active inference engine that actively self-organizes to stay in the most probable expected states, i.e., the living system’s characteristic phenotypes.

Here we briefly describe the concept of Markov blanket as prescribed in FEP (Parr et al., 2022). Technically, a Markov blanket (b) is defined as follows:

This says that, statistically speaking, if b is known, then a variable μ is conditionally independent of a variable x. In other words, if knowing the values of x and μ both depends on the condition of knowing the value of b, then knowing x would give us no additional information about μ. To identify a Markov blanket in a system with conditional dependence, one can follow a rule that the blanket for a given variable comprises its parents (the variables it depends on), its children (the variables that depend on it) and, in some settings, the other parents of its children.

The FEP leverages the principle of minimizing variational free energy—the upper bound of surprise or prediction errors—to optimize the prior beliefs in the active inference engine. There are two ways to minimize variational free energy, i.e., perceptual inference and active inference. In perceptual inference, agents strive to update their prior beliefs, while in active inference agents change their environment (or their sampling of information from the environment) by selecting a plan or policy in a set of prior beliefs that would yield the least expected free energy (Peters et al., 2017). Notably, in FEP, the variational free energy is more of a function of beliefs and expectations in the internal states rather than a function of the environments hidden from the internal states (Ramstead et al., 2020). In such processes, internal and active states’ dynamics are a function of, and only of, a variational free energy bound on surprise, and the belief optimization is implicitly done in the minimization of variational and expected free energy (Friston et al., 2022).

The notion of active inference emphasizes that actions solicit a sensory outcome that informs approximate posterior beliefs about external states of the world. Such generative process in FEP renders a living organism to be participatory, or enactive in soliciting and therefore co-creating its perception of the external states, which is very different from a representationalist process by which external states generate sensory states exclusively (Friston et al., 2022). Heuristically, one may consider that an active inference engine is actively self-evidencing what the world should be (known as an enactive account), rather than passively learning to represent what the world seems to be (known as a representationalist account)—a distinction that has been elaborated in the literature (Ramstead et al., 2020). How this distinction is related to the differences in two incompatible worldviews will be clarified later.

Inspired by FEP (Friston, 2013), we suggest that a person can be formally modeled as an active inference engine in a multi-level network consisting of four nodes, namely nodes of sensory states (S), active states (A), internal states (I), and external states or events (E). This network is partitioned into an external state (E) and an active inference engine that consists of the nodes (S) and (A) at a lower level and node (I) at a higher level. See Figure 1. The internal state (I) can be conceived as an innate prior—a set of “best guess” beliefs that may guide the active inference engine’s action planning and selection. When an event in the external states (E) interacts with a person, it can only affect internal states (I) indirectly through its interaction with the Markov blanket nodes (A) and (S), such that (E) afferently causes (S) to change and (A) efferently causes (E) to change; On the other hand, the nodes (S) and (A) at the lower level interact with the person’s prior beliefs in internal states (I) at the higher level, such that (S) causes (I) to change and (I) causes (A) to change.

In this model, nodes (A) and (S) are the Markov blanket of the node (E), because (A) is a parent of (E)—because (E) depends on (A)—and (S) is a child of (E)—because (S) depends on (E). In contrary, node (I) is not the Markov blanket of (E) because (I) is neither a parent or child of (E), nor another parent of (E)’s child, (S). In this system, nodes (I) and (E) are conditionally independent of each other, under the condition of knowing the Markov blanket nodes (A) and (S). Conditioned on the Markov blanket, i.e., sensory states (S) and active states (A), the prior beliefs activated in internal states do not provide any additional information about the external state, (E), due to the conditional independence between nodes (I) and (E), as follows,

Due to the conditional independence between nodes (I) and (E), the active inference engine and its external states are in a weakly coupled (conditionally independent) state, giving rise to the apparent duality between subject (the observer) and object (the observed), because knowing the former does not provide any information about the latter, and vice versa.

We need a dyadic model of two agents that are strongly coupled to model intersubjectivity that arises from subject-subject interactions. Just like ice and water are two phases of the same H₂O molecules that behave distinctly (solid and liquid, respectively), the same active inference engine can behave very differently between the phases of weakly coupled and strongly coupled states—while an active inference engine maintains conditional independence between its internal and external states in a weakly coupled state, such conditional independence is diminished in a strongly coupled state, when its external states are no longer a unitary node (E), but rather another active inference engine, such that one engine’s active states (A) serve as a parent of the other engine’s sensory states (S), and vice versa. In the most strongly coupled state, one person’s active states will become total environmental inputs for the other person’s sensory states, and vice versa.

Assuming this strongly coupled state in parent-child relationship, we have published a dyadic model to account for the inverse relationship between stress and intersubjectivity (Ho et al., 2020), wherein we studied how parental intersubjectivity is embodied and enacted in the brain. In this dyadic active inference model, when two active engines (two persons, say mother as Person 1 and child as Person 2) are strongly coupled, such that mother’s active state (A_M) causes child’s sensory state (S_C) and child’s active state (A_C) causes mother’s sensory state (S_M), one’s A and S become progressively similar to the other’s (A) and (S), respectively, over time, and prior beliefs in their internal states (Is) are hence attuned. Under such strong coupling, the two persons are actively inferring the other’s intentions in their prior beliefs hidden behind the sensory and active states and the variational free energy in the dyad is yoked as well. Thus, the strong coupling state mandates that one person’s prior beliefs about the other’s prior beliefs cannot reach the minimal variational free energy unless the other’s prior beliefs of the one’s prior beliefs also reach the minimal variational free energy—In other words, a higher level of intersubjectivity is attained if and only if the variational free energy in this strongly coupled state is minimized collectively. See Figure 2.

We used PubMed database to search for randomized controlled trials (RCTs) reported in English in the last 5 years prior to January 14, 2021, using the following keywords: “Parenting intervention,” “RCT” or “randomized controlled trial,” and “PSI.” We found 52 studies that met inclusion criteria and screened out 17 of them due to the following reasons: (1) the lack of PSI total score as an outcome variable, (2) the absence of comparisons between a intervention condition and a control/baseline condition, or (3) the presence of a medical condition, e.g., traumatic brain injury in the child, that may originate from and/or result in complications in the social environments beyond the parent-child dyads. The list of the final 35 studies reviewed and the coding results for each study are presented in Table 1. These studies were coded by two authors SSH and MG (hereafter Raters 1 and 2, respectively) independently, according to the following binary coding scheme.

The Components 1–3 in the coding scheme corresponded to the three components of the dyadic active inference model, which we developed and presented above to address the problems of (1) deficient relational benevolence, (2) under-coupling, and (3) over-mentalizing, respectively. Notably, the major distinction between Component 1 and other Components is that Component 1 serves as a mindset, a frame centered in the dyad, not a single person, with an emphasis on unconditional positive regards, e.g., love and warmth, of the relation, while Components 2 and 3 are more specifically contingent upon specific situations and skill oriented. The major distinction between Components 2 and 3 is that Component 2 should be focused on child-oriented observations with a data-driven approach, not inward observations of parents’ internal working model of the child. Conversely, Component 3 should be focused on parental inward-observations of parental thoughts and emotions when they are used to mentalize the child.

The outcome variable, the Tx effect on PSI, was coded according to the following rule: If there was a statistical significant difference in PSI (total score) between the intervention (Tx) and Control groups, as a significant Group main effect or a Time-by-Group interaction effect, or a within-subject difference from a baseline, such that the PSI total score was lower in the Tx than the control condition, then the Tx effects of PSI was coded as “1” (positive effect), otherwise as “0” (negative effect).

As already described in the beginning, living organisms are impermanently, inter-dependently self-organizing through ceaseless interactions with their environments. It is emphasized in FEP that living organisms as active inference engines are self-evidencing in its environments by using their own actions to solicit their sensory inputs from the world. In a way, active inference is consistent with the notion of participatory universe—the outcome of measuring a quantum system depends on the apparatus chosen to perform the act of measuring—coined by John A. Wheeler, one of the greatest Physicists and Philosophers in our time.

Most physicists would agree that, ontologically, the universe is fundamentally relational, and, epistemologically, to be observable is to be interactable in physics (Rovelli, 2021). In the standard model of physics, fundamental particles are nothing but products of interactions of even more fundamental quantum fields (Peskin and Schroeder, 1995). The relational nature of the participatory universe has been rigorously demonstrated many times in a family of so-called delayed-choice experiments. This kind of experiments aims to demonstrate the dual nature of a quantum system, e.g., a single photon that can behave either like a particle or like a wave. In this kind of experiments, whether a single photon behaves particle-like or wave-like depends on whether a particle detector or wave detector is chosen to measure the photon’s behavior. Because the act of choosing either one of the detectors to observe the single photon occurred after the very photon has completed its behavior, this kind of experiments are thus called “delayed-choice.” After decades of rigorous delayed-choice experiments, most physicists would agree that:

The notion of participatory universe suggests that all information is relationally dependent upon the existence of observations or observers whereas the existence of observations or observers is relationally dependent upon the ingredients of the universe. The observer here may be a living system, which is modeled as an active inference engine, or simply a quantum-system measuring apparatus that solicits and thus co-creates the outcomes of observing an incoming event.

To summarize the relational worldview underlying active inference and participatory universe, we resort to the notion of “Dependent Origination” that can be abstractly expressed in the following equation:

Colloquially, Eq. 1 should read “Effect is an interactive product of Cause by Condition.”

Both Cause and Condition are factors participating in an interaction that produces Effect. Among these factors, some are called “Cause,” if they maintain certain systemic continuity with the “Effect”; others are called “Condition,” if they lack apparent systemic continuity in relation to either Cause or Effect. The term systemic continuity is used here to refer to the relation between Cause and Effect—that they are continuous but successive phenotypes of the same system.

For example, fruit is an interactive product of its seed and other factors such as soil, bacteria, water, sunlight, farmer, etc. In this case, the fruit is Effect, the seed is Cause, and other factors are Conditions. There is systemic continuity between the seed (Cause) and fruit (Effect) because the seed and fruit belong to the same system defined by the same genes that they carry (systemic), but they never co-exist simultaneously in the temporal succession of seed and fruit (continuity). In contrast, while the other factors (soil, bacteria, water, sunlight, farmer, etc.) are necessary to produce the fruit, they are designated as Conditions because they lack the systemic continuity with either the seed or the fruit.

The interactive product, designated by the sign “×” in Eq. 1, renders Eq. 1 as a non-linear formula. Mathematically, the non-linearity mandates that Effect is neither a linear transformation of Cause, nor Condition, nor a linear combination thereof.

Interestingly, Eq. 1 can serve as a mathematical expression of the ultimate nature of reality, namely “Emptiness,” that has been established by Arya Nagarjuna, the founder of Madhyamaka School of Buddhist Philosophy, in the following reasoning:

“Neither from itself,

Nor from another,

Nor from both,

Nor without a cause,

Does anything whatever, anywhere arise.” (Nagarjuna, 1995) Ch. 1 V. 1

Arya Nagarjuna’s reasoning on Emptiness can be translated in terms of Eq. 1, as follows:

“Neither from self” means that the Effect is not a linear transformation of its Cause—, e.g., although the fruit and seed carry the same genes, the fruit is not identical to the seed or a scaled-up version of the seed.

“Nor from other” means that the Effect is not a linear transformation of its Condition—, e.g., the fruit is not identical to soil, bacteria, water, sunlight, farmer, etc., that do not even share the same genes with the fruit.

“Nor from both” means that the Effect is not a linear combination of the Cause and Condition—, e.g., the fruit is not just the sum of the seed, soil, bacteria, water, sunlight, farmer, etc. that do not have any interactions among them.

“Nor without a cause” means that the Effect is not something other than an interactive product of Cause by Condition—, e.g., the fruit does not come to exist without being the effect of the interactions among the seed, soil, bacteria, water, and other conditions.

“Does anything whatever, anywhere arise” means that nothing can be observed without following the ultimate nature of the participatory universe.

In short, Eq. 1 is not only an abstract expression of “Dependent Origination,” but also an axiomatic translation of Arya Nagarjuna’s reasoning on “Emptiness.” The resulting functional equivalence between “Emptiness” and “Dependent Origination” is eloquently reflected in the pith of Buddhist wisdom, as Je Tsongkhapa (1357–1419) stated in his masterpiece “In Praise of Dependent Origination” (Lobsang Gyatso and Woodhouse, 2011) that:

—Je Tsongkhapa, translated by Geshe Thupten Jinpa.

The consistency between the Buddhist wisdom and John A. Wheeler’s notion of participatory universe becomes evident when we summarize experimental evidence demonstrated in those delayed-choice experiments (Ma et al., 2016) as a special case of Eq. 1, that Effect of measuring a quantum system (e.g., a single photon) is an interactive product of Cause by Condition—wherein Cause is the to-be-measured single photon and Condition is the apparatus used to detect the photon’s particle-like or wave-like behavior. When Conditions favor the single photon’s wave-like or particle-like behavior, the photon will appear to behave like a wave or a particle, respectively, after it interacts with Conditions.

Altogether, the relational worldview in the current context specifically refers to the notion that Effect is as an interactive product of Cause by Condition, which can be equivalent to the notions of Dependent Origination and the ultimate nature of reality, Emptiness, in Buddhist Philosophy as well as the notion of Participatory Universe in Physics.

In the following sections, we will use similar abstract expressions to describe how intersubjectivity and active inference framework can be understood as additional special cases of this relational worldview.

(Fuchs, 2017, p. 27)

Intersubjectivity—the awareness of self and other’s intentions and feelings—is relational, because the effect of awareness in intersubjectivity depends on the interactive coupling between the participants. Here we apply the abstract expression of Dependent Origination to the nature of awareness and intersubjectivity.

According to the Buddhist science of mind, the nature of awareness is fundamentally relational, described as follows:

“The nature of cognition is stated to be awareness, and the nature of consciousness is said to be clear (or luminous) and aware. ‘Clear’ here expresses the essential nature of consciousness, and ‘aware’ expresses its function. ‘Clear’ also indicates: (1) that consciousness is beyond the nature of matter, which is characterized as tangible and obstructive, so it is clear in nature; (2) that just as reflections appear in a mirror, any internal or external object whatsoever—good or bad, pleasant or unpleasant—can appear in consciousness, so consciousness is luminous in that it illuminates objects; and (3) that the essential nature of consciousness is not contaminated by the stains of mental afflictions such as attachment, so its nature is clear or luminous.” (Tenzin Gyatso, 2020, p. 41).

The relational nature of the awareness is often likened to a clear lampshade or mirror metaphorically, as discussed in our previous work (Ho et al., 2021). In the former, a clear lampshade is colorless (clear) and any object that the mind perceives is like a light bulb in the clear lampshade, which can color the lampshade with its light, e.g., the lampshade’s color becomes blue when a light bulb emits blue light. However, just as the light bulb can never stain the lampshade, the object perceived by the mind can never stain the mind. Thus, the mind (lampshade) returns to its colorless clarity as soon as the object (light bulb) is turned off. In the latter, awareness is also likened to a mirror, as it reflects the object in front of the mirror, but the mirror is not the object nor the image in the mirror. In other words, the image in the mirror is an interactive product of the mirror and the object in front of the mirror.

Here we describe the mirror-like nature of the awareness in terms of the abstract expression of Dependent Origination. For an object, A, let a subject’s awareness of A be “A-ness,” which is called the “qualia” of perceiving A. As Effect is an interactive product of Cause by Condition, “A-ness” is an interactive product of subject (Cause) and objects (Conditions), which include the object A and other environmental conditions, e.g., the subject’s visual system and other physical environments. Note that the subject, not the object A, is designated as Cause because the qualia as Effect is a subjective experience that has the systemic continuity with the subjectivity of the subject, whereas the object A does not have such systemic continuity. Therefore, in the realm of awareness, Eq. 1 can be re-expressed as follows:

Using the mirror metaphor of the mind, the qualia “A-ness” is like the image in the mirror (Effect), the subject’s mind is like the mirror (Cause), and the object A is like an object placed in front of a mirror (Condition).

In parallel, the object A should be changed after the subject-object interaction too. There should be a counterpart to Effect_[Qualia], which can be re-expressed as follows:

wherein Effect_{[Object A’]} denotes the post-interaction object A, with the object A as its Cause and the subjects of awareness and environmental objects (e.g., the brain system and lights in the room) as its Conditions. Due to the subject x object interaction, such Effect_{[Object A’]} is effectively infusing the object with certain “mental energy” in a process called “cathexis.” We will further discuss the concept of cathexis in the context of active inference framework below.

As the Ubuntu proverb says, “I am because you are,” there is no independent subject “I” that can be designated without also simultaneously designating the objects of “other” in relation to the “I.” The intersubjective awareness is a relational awareness of self and other. Trevarthen made a distinction between primary and secondary intersubjectivity (Trevarthen and Aitken, 2001). The former refers to “an infant’s active and immediately responsive conscious appreciation of the adult’s communicative intentions” and the latter refers to “the integration in the new form of cooperative person-person-object awareness” that combines object awareness (e.g., doing with things) and person awareness (e.g., communicating with persons).

The primary intersubjectivity is considered innate and mainly characterized by embodied, affective, and intuitive forms of relationships, preceding communications mediated by symbolic and verbal processes (Trevarthen and Aitken, 2001). A key attribute underlying primary intersubjectivity is spontaneous mimicry or voluntary imitation of others’ facial expressions or manual gestures. Infants show spontaneous facial mimicry soon after birth (Meltzoff and Moore, 1977). When primary intersubjectivity is successfully executed and maintained, it leads to establishing the sense of safety and assurance that everything is all right at that moment.

Secondary intersubjectivity, which emerges around 9 months of age, incorporates objects into the mother–infant interactions, forming a person-person-object triadic relationship (Trevarthen and Hubley, 1978). Jointly attending to an object of shared interest is seen as a critical step toward a mutual incorporation of the other’s perspective into shared experiences. The emergence of secondary intersubjectivity points to a major developmental milestone where shared joint activities between mother and infant can create a significantly more advanced level of interactive intersubjectivity that can support the development of higher cognitive capacity for language, reflection, and perspective-taking.

Here we focus on how to understand primary intersubjectivity in terms of the abstract expression of the relational worldview. Let Eqs 2a, 2b be applied to dyadic processes wherein two subjects, Person 1 (P1) and Person 2 (P2), are strongly coupled in their person-person interactions.

For Person 1,

For Person 2,

The notations, Effects, on the left side of the equations are: Effect_{[“P1×P2–ness” in P1]} denotes P1’s qualia about P1 × P2 dyadic interactions, i.e., “P1 × P2-ness,” and Effect_{[“P2 × P1–ness” in P2]} denotes P2’s qualia about P2 × P1 dyadic interactions, i.e., “P2 × P1-ness.” The notations on the right side of the equations are: Cause_[P1] or Cause_[P2] denotes P1 or P2’s mirror-like awareness as the subjects; Condition_{[P1 × P2]} or Condition_{[P2 × P1]} refers to the conditions that interact with the mirror-like awareness, which can be any objects or behaviors of the dyadic system, e.g., the dyad’s brains, bodies, verbal or physical behaviors, during the dyadic interactions.

As mentioned above, the Buddhist notion of mirror-like awareness suggests that “the essential nature of consciousness is not contaminated by the stains of mental afflictions such as attachment, so its nature is clear or luminous.” (Tenzin Gyatso, 2020, p. 41). Eqs 3a, 3b do not guarantee that all qualia of “P1 × P2-ness,” or “P2 × P1-ness,” are equal in its level of intersubjectivity. Instead, the capacity of P1 or P2’s intersubjectivity (Effect_{[“P1 × P2–ness” in P1]} or Effect_{[“P2 × P1–ness” in P2]}) depends on the quality of the conditions (Condition_{[P1 × P2]} or Condition_{[P2 × P1]}) that one’s mirror-like awareness (Cause_[P1] or Cause_[P2]) interacts with. In other words, Eqs 3a, 3b can be applied to any level of intersubjectivity, whether it is an optimal level of intersubjectivity (as depicted in Figure 2) or a sub-optimal level of impaired intersubjectivity (as depicted in Figure 3).

The relational worldview—Effect is an interactive product of Cause by Condition—can be applied to understand the active inference framework. In the active inference framework, actions in active inference co-create the perceptions with the incoming event, which is analogous to the facts that actions of measuring co-create the effects of the measurement of the quantum systems in delayed-choice experiments in Physics. When a person or agent is modeled as an active inference engine, the engine serves as the subject that interacts with an object in the external world. As an example, here we apply the abstract expression of Dependent Origination to the active inference process in the four-node network (Figure 1).

Let nodes (A), (S), and (I) of an active inference engine be the Cause and node (E) in the external state be objects or events be the Condition that interacts with the Cause in Eq. 1. Now we have a pair of expressions as follows:

For the effect on the active inference engine,

wherein Effect_{[nodes (A’), (S’), (I’)]} designates the active inference engine after the Cause by Condition interaction, which involves iterations of perceptual and active inferences to optimize the prior beliefs to minimize the variational free energy in FEP. The systemic continuity between the Cause and Effect in Eq 4a is consistent with the notion that active inference engines are self-evidencing (Friston, 2018; Friston et al., 2022).

For the effect on the external state,

wherein Effect_{[node (E’)]} designates the external state after the Cause by Condition interaction. Notably, the subject-by-object interaction effect on the object is denotated as the “cathexis” here, because it potentially entangles the object with certain orientation or propensity of the subject. For example, Edward Tolmen construed the process of cathexis as the learned tendency to associate certain objects with certain drives, which is one of the major determinants of choice behaviors, e.g., why meat lovers tend to satisfy their hunger with meat (positive cathexis) rather than non-meat products (negative cathexis; Tolman, 1945).

In the neuroscience literature, the cathexis effect is usually conceptualized in terms of the construct of incentive value—a positive or negative cathexis of an object is construed as a positive or negative incentive value of the object, respectively, (Dickinson and Balleine, 1994; Balleine and Dickinson, 1998). For example, a rat has tasted a very salty liquid solution and then does not consume the same solution anymore. In this case, the very salty solution has a negative incentive value for the rat.

The question is, after the negative incentive value of the solution has been established, what would the rat do when it is put in a salt-deprived state and then encounters the same salty solution again? This is the question ingeniously answered in an animal study, which showed that when the rat re-encountered the very salty solution after being put in a salt-deprived state surgically, it immediately ran toward and consumed the very salty solution appetitively (Robinson and Berridge, 2013).

As explained above, the non-linearity in those abstract expressions of Dependent Origination, including Eq. 4b, mandates the following three rules mathematically: The effect of Cause-by-Condition interaction, node (E’), is not identical to (1) any linear transformation of node (E), (2) nor a linear transformation of nodes (A), (S), (I), and (3) nor a linear combination of all four nodes. Hence, the incentive value of an object, post-cathexis, is not a property of the object, nor an internal representation encoded in the subject, nor a linear combination of the external object and internal representation.

The above axiomatic reasoning based on the non-linearity of Eq. 4b is corroborated by the experiments conducted by Robinson and Berridge (2013). As their study demonstrated an instant flip of the incentive value of the salty solution, from something negative to something positive, without any new learning, it is strongly suggested the following: The incentive value is not a property of the object, because the salty solution is not manipulated at all in their study. Nor the incentive value is encoded as the subject’s internal representation. The reason for this refutation is that, had this notion been true, the rat would have avoided the salty solution even when it was salt-deprived after surgery, because (a) presumably the internal representation of the salty solution in the brain was not altered by the experimental manipulations and (b) the rat did not have any new learning trials to encode a new internal representation of the solution’s incentive value in the novel salt-deprived state, as a representationalist account of the cathexis of incentive value would predict (Balleine and Dickinson, 1998).

This experimental refutation of the representationalist account of the cathexis of incentive value is potentially relevant to the refutation of the representationalist account of active inference, as noted in the FEP literature (Ramstead et al., 2020). The refutation of the representationalist accounts in the domains of incentive value and active inference speaks to the incompatibility between the relational worldview and realist worldview, to be further discussed below.

As we abstractly denoted the relational worldview as the notion of Dependent Origination—Effect is an interactive product of Cause and Condition, a realist worldview presents a stark contrast. Realism can be defined as follows:

In other words, the realism assumes a non-participatory universe wherein an object’s ultimate real ontology (a, b, c and so on) is identical to its properties (F-ness, G-ness, H-ness and so on) that can be observed and ascertained, independent of the subject/observer/apparatus involved in the observation. In an abstract expression, a realist worldview can be expressed as follows:

The incompatibility between the realist and relational worldviews is clearly evident in preceding discussions offered here. First, in contrast to the non-linearity in the abstract expression of relational worldview (Eqs 1–4), Eq. 5 is a linear function. Second, in contrast to Eqs 1–4, the subject or the apparatus used to make observation plays no roles at all in Eq. 5.

This incompatibility may point to the innate possibility of the cessation of suffering, i.e., healing. We postulated in Ho et al. (2021) that an invalid belief—the belief that self or other is fixed and unchanged after any person-to-person interactions—is the root cause of entrapping one’s awareness in an unhealthy state characterized by rigidity and inflexibility. The consequence of invalid beliefs in one’s awareness is that one will misattribute the cause of his or her subjective experiences (qualia) to the object itself, rather than the interactive product of his or her active inference engine by the object. In this misattribution, one judges the value of the object based on one’s own qualia (“this object makes me feel good/bad”), which one misbelieves to be identical to the object’s ultimate existence (“this object is good/bad in and of itself”), such that one would believe that he or she can only have or not have the object, rather than participate both in the creation of the qualia and in the unfolding of how the object appears to be in accordance with the ultimate nature of the reality, Dependent Origination.

In fact, invalid beliefs are deeply embedded in the realist worldview or beliefs. In Buddhism, such realist beliefs are called conceptual thoughts (Vikalpas), and the substantial bases underlying these realist beliefs are called mental fabrication or superimposition (Prapañca). In the root text of Madhyamaka Philosophy, Arya Nagarjuna provided the diagnosis and treatment for cyclic suffering caused by Vikalpa and Prapañca, stating:

We have discussed the types of conceptual thoughts and levels of mental fabrication, as well as how these processes adversely influence the active inference processes of the brain in thinking about compassion elsewhere (Ho et al., 2021).

According to Buddhism, suffering is rooted in invalid beliefs and the cessation of suffering is guaranteed because all views laden with invalid beliefs are incompatible with the ultimate nature of reality (Tenzin Gyatso, 1997, 2009). Arya Nagarjuna’s prescription of cessation of suffering, as quoted above, can be paraphrased as follows: since the ultimate nature of reality—whether it is called Emptiness, Dependent Origination, or Participatory Universe in the context of the relational worldview—is incompatible with the realist worldview that can result in suffering, the cessation of suffering is always possible through the realization of the ultimate nature of reality by living in this participatory universe in the absence of invalid beliefs in one’s awareness and active inference engines.

Among the initial steps toward this goal, what we hoped to clarify through the abstract expressions presented above is the critical importance of cultivation of an inward, reflective contemplation, including being aware of one’s own prior beliefs in his or her internal states, the difference between the relational and realist worldviews underlying his or her prior beliefs, and the incompatibility between the realist worldview and the relational worldview concerning the nature of reality. Promoting such awareness is primarily an educational, contemplative work on changing beliefs and behaviors that are thought to hinder healthy dyadic interactions. By applying the abstract expressions and formal dyadic model to the study of the parenting interventions as an example domain, we indeed found that the cultivation of inward contemplation (i.e., Component 3 in the coding system) played a critical role in the efficacy of parenting interventions for reducing stress.