Morphogenetic System Theory as a Holistic Meta-Methodology for Socio-Technical System Evolution: A Review

1. Introduction

The objective world is replete with forms—living forms such as fish, insects, birds, and animals, and non-living forms such as risks, laws, curves, and learning communities. Form, as the structured manifestation of objective things under certain conditions, is characterized by direct recognizability, non-conservation, non-measurability, repeatability, and persistence [1]. Form identification can be divided into qualitative and quantitative approaches. Although qualitative identification is simple and applicable, it cannot accurately record or describe morphological differences, struggles with large-scale data processing, and lacks mathematical tractability. Consequently, developing quantitative methods for morphological identification has become an inevitable trend in disciplinary advancement [2].

In the field of quantitative morphological identification, scholars have conducted extensive research along diverse pathways. Mathematical morphology [3], grounded in set theory and topology, detects and transforms images through structuring elements, enabling quantitative description and filtering of geometric forms. Catastrophe theory [4], centered on singularity theory, seeks to provide mathematical models for morphological evolution in nature, with particular emphasis on describing discontinuous changes at critical points of systems. Self-organizing mapping net [5], representative of unsupervised learning, map high-dimensional data to low-dimensional topological structures through competitive learning, demonstrating unique advantages in pattern discovery and feature extraction. Bionic pattern recognition [6], characterized by high-dimensional geometric analysis, explores computer cognition of complex forms from the perspective of image-based thinking. These methods have achieved significant advances in their respective domains, yet share a common methodological limitation: they treat forms as outcomes to be processed—to be classified, filtered, or recognized—rather than as traceable processes. In other words, while considerable attention has been devoted to what forms are, the question of how forms become—their formation and evolution—still lacks systematic mathematical tools for description.

To address this gap, Professor Germano Resconi of the Catholic University of the Sacred Heart in Milan proposed the Morphogenetic System Theory (MST) in 1998 [7]. This theory integrates multidisciplinary knowledge from artificial intelligence, physics, mathematics, biology, philosophy, and cognitive neuroscience, and is committed to interdisciplinary unification, with particular emphasis on the unity of algebra and geometry. Its core innovation lies in redefining form as a morphogenetic process from input field to output field, and in achieving the recording, preservation, and reconstruction of forms through the closed loop of "write" and "read" operations, thereby elevating form identification from static classification to dynamic process analysis. MST synthesizes multidisciplinary achievements—including Sheldrake's morphogenetic field theory [8], Gestalt psychology [9], Turing's Turing machine model [10] and morphogen concept [11], Jessel's secondary source theory [12], Gabor's holography [13], Fatmi and Resconi's new computing principle [14], Dubois and Resconi's hyperrecursion theory [15], Wille's formal concept analysis [16], Ricci-Curbastro's tensor calculus [17], as well as Resconi's own general system logical theory [18] and agent uncertainty theory [19]. It aims to provide a unified mathematical model for morphogenesis phenomena and has been applied to risk analysis, uncertainty analysis, quantum mechanics, holography, and other fields.

Although MST has achieved progress in risk analysis, system identification, and image recognition [1,20,21,22,23,24], its broader methodological value—particularly its potential as an analytical framework for socio-technical systems—has not yet been systematically revealed. Socio-technical systems—such as learning communities, legal institutions, and organizational forms—exhibit typical morphogenetic behaviors: they undergo five morphogenetic stages: germination, differentiation, coupling, explosion, and updating. These evolutionary trajectories align closely with the core logic of MST, yet a systematic review that positions MST as a holistic meta-methodology for socio-technical system evolution remains lacking.

To this end, this paper has three objectives: (1) to trace the academic development of MST and distill its core philosophical and methodological tenets; (2) to compare MST with seven related methods—information diffusion principle, attribute theory, grey system theory, catastrophe theory, mathematical morphology, self-organizing map net, and bionic pattern recognition—in order to clarify its distinctive methodological contributions; and (3) to examine two cases—learning community morphogenesis and legal morphogenesis—as empirical anchors for testing MST's applicability in socio-technical systems. Through this work, we argue that MST is not merely a mathematical formalism, but a holistic meta-methodology oriented toward socio-technical system evolution, providing a formal analytical tool for systems thinking and interdisciplinary research.

2. Morphogenetic System Theory: Foundations and Framework

2.1. Development Trajectory

The development of MST can be divided into three stages.

Stage One: Theoretical Incubation (pre-1998). This stage primarily involves the emergence and development of concepts such as secondary sources, holography, morphogens, hyperrecursion, morphogenetic fields, tensor calculus, and formal concept analysis. The idea of secondary sources can be traced back to Huygens' principle [25], which provided the physical prototype for morphogenesis. Jessel [12] calculated secondary sources to infer wave propagation and generate desired fields, and co-published A General Systems Logical Theory with Resconi in 1986[18], laying the foundation for the later introduction of internal and external source concepts. Gabor's holography [13] revealed a computational paradigm in which the part implies the whole, and in 1954 he further elaborated the mathematical principles of intelligent machines [26]. Fatmi and Resconi proposed a new computing principle [14] based on this, establishing the basic framework of superposition computation. In 1992, Dubois and Resconi proposed hyperrecursion theory [15]. Turing proposed the Turing machine [10] and later the reaction-diffusion model and the concept of morphogens [11], using mathematical equations for the first time to describe the spontaneous emergence of form. Sheldrake proposed the hypothesis of formative causation [8], arguing that biological morphogenesis is guided by morphogenetic fields—an idea that influenced Resconi and prompted him to extend Sheldrake's hypothesis from biology to a computable and verifiable general system theory. Wille's formal concept analysis [16] provided the data structure prototype for MST's morphogenetic information tables. The concepts of covariant components, contravariant components, and invariants from Ricci-Curbastro's tensor calculus [17] were inherited and simplified in MST.

Stage Two: Proposal and Refinement (1998–2008). In 1998, Resconi published The Morphogenetic Neuron [7], marking the formal establishment of MST. Over the following decade, the theory was extended to quantum computing [27], semantic search [28], robotics kinematics [29], and morphogenetic neural networks [30]. In 2003, the projection operator was proposed as a closed-loop control mechanism for systems, and the core ideas of the morphogenetic system became essentially clear [31]. In 2007, Resconi et al. [20] published Morphic Computing: Concept and Foundation, introducing the concept of "morphic computing" for the first time, which marked the maturity of the theory. This stage was characterized by "new methods for old problems," consolidating the theoretical foundation.

Stage Three: Maturation and Application (2009–present). Research focus shifted to applications. Resconi provided a geometric interpretation of risk analysis [32] and explored the synergy effects of risk analysis in combination with information diffusion theory [33]. Jing Tian et al. [34] proposed morphogenetic updating algorithms and explored their application in probabilistic risk updating. Resconi et al. presented a logical method for computing information matrices [35] and published the monograph Introduction to Morphogenetic Computing [36], systematically summarizing the theoretical framework. In recent years, MST's applications have expanded from risk analysis to traffic safety [37], education and teaching [38], and construction project risk field construction [39], demonstrating cross-disciplinary applicability. In recent international developments, Dodig-Crnković proposed a theoretical framework of cognition as embodied morphological/morphogenetic computation [40], and later proposed a morphological info-computational framework [41], elevating morphic computing to the philosophical height of "cognitive underlying logic" and systematically arguing for morphic computing as a unified logical framework connecting physics, chemistry, biology, and cognition [42]. The closed-loop self-regulation between tissue shape and morphogen patterns revealed by Kaity and Lobo [43] resonates across disciplines with Resconi's morphic computing loop. MST's dialogue with adjacent fields such as morphogenetic engineering and computational systems biology is expanding its theoretical influence. This stage marks MST's transition from theoretical construction to an applied methodology oriented toward both natural and social systems.

2.2. Core Architecture

MST is built upon three mathematical spaces: the reference space (the space in which observed objects exist), the object space (an M-dimensional Euclidean space composed of M objects), and the attribute space (an N-dimensional non-Euclidean subspace composed of N basic attributes). Both the object space and the attribute space are subspaces of the reference space, and the attribute space is also a non-Euclidean subspace of the object space.

Knowledge in MST is represented by a morphogenetic information table (Table 1), which extends the cross-tabulation concept in formal concept analysis and records the degree of connection between objects and attributes.

The morphogenetic information table expresses the three spaces in tabular form. When entering subsequent morphic computing, the information is represented in matrix form:

$$H=\left[\begin{array}{cc}\begin{array}{c}{h}_{11}\\ \begin{array}{c}{h}_{21}\\ \cdots \\ h_{M1}\end{array}\end{array}& \begin{array}{c}{h}_{12}\\ \begin{array}{c}{h}_{22}\\ \cdots \\ h_{M2}\end{array}\end{array}\end{array}\begin{array}{cc}\begin{array}{c}\cdots \\ \begin{array}{c}\cdots \\ \cdots \\ \cdots \end{array}\end{array}& \begin{array}{c}{h}_{1N}\\ \begin{array}{c}{h}_{2N}\\ \cdots \\ h_{\mathit{MN}}\end{array}\end{array}\end{array}\right]，X=\left[\begin{array}{c}{x}_1\\ \begin{array}{c}{x}_2\\ \cdots \\ x_M\end{array}\end{array}\right]$$

Here, the matrix H represents the context of morphogenesis (the column space is the subspace spanned by the basic attributes). The mixed attribute Xcorresponds to the input field. The process by which basic attributes interact to generate the mixed attribute X is called attribute morphogenesis. The orthogonal projection operator of the context is $Q_H=H{\left(H^{T}H\right)}^{-1}{H}^T$, and the projection of X onto the context H is Y. It satisfies the equation Y = HS_I = Q_HX, and its calculation formula is:

$$\begin{array}{ll}Y& =\left[\begin{array}{c}{y}_1\\ \begin{array}{c}{y}_2\\ \cdots \\ y_M\end{array}\end{array}\right]=\left[\begin{array}{cc}\begin{array}{c}{h}_{11}\\ \begin{array}{c}{h}_{21}\\ \cdots \\ h_{M1}\end{array}\end{array}& \begin{array}{c}{h}_{12}\\ \begin{array}{c}{h}_{22}\\ \cdots \\ h_{M2}\end{array}\end{array}\end{array}\begin{array}{cc}\begin{array}{c}\cdots \\ \begin{array}{c}\cdots \\ \cdots \\ \cdots \end{array}\end{array}& \begin{array}{c}{h}_{1N}\\ \begin{array}{c}{h}_{2N}\\ \cdots \\ h_{\mathit{MN}}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}{S}^1\\ \begin{array}{c}{S}^2\\ \cdots \\ S^N\end{array}\end{array}\right]=S^1\left[\begin{array}{c}{h}_{11}\\ \begin{array}{c}{h}_{21}\\ \cdots \\ h_{M1}\end{array}\end{array}\right]+S^2\left[\begin{array}{c}{h}_{12}\\ \begin{array}{c}{h}_{22}\\ \cdots \\ h_{M2}\end{array}\end{array}\right]+\cdots +S^N\left[\begin{array}{c}{h}_{1N}\\ \begin{array}{c}{h}_{2N}\\ \cdots \\ h_{\mathit{MN}}\end{array}\end{array}\right]\\ & =S^1{H}_1+S^2{H}_2+\cdots +S^N{H}_N=H{S}_I\end{array}$$

(1)

Morphic computing is the core of MST, consisting of write and read operations, analogous to the principle of holography: the write operation "writes" the implicit information in the input field into the internal source S_I(the contravariant component of X ), and the read operation "reproduces" the input field through the internal source and the context.

The write operation proceeds as follows:

$$\left\{\begin{array}{c}{S}_E=H^{T}X\\ S_I={\left(H^{T}H\right)}^{-1}{S}_E={\left(H^{T}H\right)}^{-1}{H}^{T}X\end{array}\right.$$

(2)

Here, S_E is the external source, the covariant component of X; S_I is the internal source, the contravariant component of X; X is the input field; H is the context; and H^T is the transpose of H.

The read operation uses the internal source to calculate the projection Y of X onto H:

$$Y=H{S}_I=H{\left(H^{T}H\right)}^{-1}{H}^{T}X=Q_{H}X$$

(3)

Here, X is the input field, H is the context, and Y is the projection of X onto H. S_I is the internal source, and Q_H is the orthogonal projection operator in H, expressed as Q_H = H(H^TH)⁻¹H^T. If M = N, then Y can accurately reproduce X, meaning that the unknown form can be fully recognized. IfM > N, then Y is the best approximation of X in the least squares sense, but not exactly equal to X. In general, there will be a deviation between Y and X.

Let D denote the length of Y. The scalar $D^2={\left(Q_{H}X\right)}^T\left(Q_{H}X\right)={\left(H{S}_I\right)}^T\left(H{S}_I\right)$ is an invariant, remaining unchanged under any unitary transformation. The error vector e = X − Y = X − Q_HXrepresents the difference between the input field X and its projection Y. The norm ‖e‖ of the error vector e measures the distance from the input field X to the attribute space, serving as a direct quantitative indicator of morphological alignment.

2.3. Philosophical Propositions

The computational logic of MST embodies three philosophical propositions:

• The unity of subjective and objective understanding.

The context encodes prior knowledge (subjective understanding), while the input field captures empirical data (objective facts). The write operation uses the context to extract implicit patterns from the data (internal sources), while the read operation revises the context through projection. This iterative cycle reflects the essence of the scientific method: observe → describe with existing understanding → revise understanding if insufficient → understand. It also provides a quantitative measure of expert reliability—the context provided by experts must reasonably represent the morphogenetic process of the input field; only then is the context valid and the morphic computing complete. Morphic computing also demonstrates a path of "deriving the unknown from the known." An objective object can only be recognized when it resonates harmoniously with existing "similarity blocks" in the mind. Morphic computing mathematically replicates this psychological mechanism of image recognition, fully following the cognitive principle of "proceeding from existing practical experience to understanding new things through logical deduction."

2.

The unity of analysis and synthesis.

The fundamental philosophy of MST is holism. The context is a whole composed of basic attributes. The write operation is analysis: the input field is projected onto the context, where basic attributes interact and their connections with the input field are separated and expressed as internal sources. The read operation is synthesis: the basic elements combine with their internal sources to generate the output field—the projection of the input field. In simpler terms, the write operation decomposes the input field into basic attributes and their connection patterns, while the read operation recomposes them into a coherent whole. This dialectical relationship echoes the holistic philosophy of systems thinking: the whole precedes the parts and imposes constraints on them. It also resonates across time and space with the idea in the Tao Te Ching: "The Tao gives birth to one, one gives birth to two, two gives birth to three, and three gives birth to all things."

3.

Projection as measurement; structure over individual values.

Morphic computing measures the similarity between the input field and the selected context through projection. When the projection has high fidelity to the input field, it indicates that the morphological features of the input field can be adequately expressed by the context—a "good measurement." High-fidelity projection implies that the structural constraints of the context sufficiently capture the morphological features of the system. Consistent with field theory, understanding the structure of the field is more informative than observing the field value at any single point, because the field connects all elements into a global whole. The goal is not to obtain individual observations but to extract structural information from multiple observations. Based on field theory, understanding the structure of the field is more important than understanding the field value at any arbitrary point; particles at any position are affected by the field; the field connects all particles in the universe into a global entity, raising the issue of local invariance—every system is an open system, and local invariance will disappear. The solution lies in preserving invariance through deformation in non-Euclidean geometry: the behavior of the field can be replaced by spatial deformation, the reference space changes with time and space, and the behavior of the field is merely a virtual phenomenon.

3. Methodological Positioning of MST: A Comparative Analysis with Related Approaches

3.1. Information Diffusion Principle

The information diffusion principle was proposed by Huang Chongfu of Beijing Normal University in 1992 in his doctoral dissertation "Information Diffusion Principle and Computational Thinking with Applications in Earthquake Engineering" [44]. The information diffusion principle asserts that, given incomplete samples, there exists a reasonable diffusion function that can extract fuzzy information to more accurately estimate a functional approximation relationship. Although progress has been made in its application, its inherent regularity has lacked theoretical support. The information diffusion principle shares MST's "holistic" thinking—it uses an information matrix to express the overall structural information of samples, and the morphogenetic field can serve as a universal database encompassing it. MST provides a clear mathematical theoretical foundation for the information diffusion principle, and the information matrix can be seen as a prototype of the context. The information matrix is merely a way of presenting the information structure, but through the morphogenetic process in the context, the structural characteristics of the data can be further analyzed. The process of several basic fields superimposing to form an input field characterizes the morphogenetic process of the input field. MST's clear mathematical framework provides a theoretical foundation for the experience-based information diffusion principle, and the fields where information diffusion models are applied can be explored using MST.

3.2. Attribute Theory

Attribute theory was proposed by Professor Feng Jiali, and is a mathematical theory and method of thinking construction and intelligent simulation based on attribute computing [45]. In 2008, he explicitly stated in communication with the author that attribute theory and MST are dual theories. Both pursue the integration of qualitative and quantitative reasoning and the unification of algebra and geometry. Attribute theory achieves the transformation from quantitative to qualitative through attribute coordinate systems, complementing MST's quantitative morphological identification through "context + projection." Together, they can provide mathematical tools for Qian Xuesen's framework of "quantitative intelligence + imagery intelligence → wisdom (great wisdom)" [46].

3.3. Grey System Theory

Grey system theory was founded by Professor Deng Julong in 1982 [47], specifically addressing uncertainty systems where "part of the information is known and part is unknown." Its core includes grey generation, grey relational analysis, GM models, grey decision-making, and grey control. Although grey system theory has wide applications in control, prediction, and decision-making, its essence lies in the GM(1,1) model. In the basic form of the grey differential equation of GM(1,1) (first-order, one-variable differential equation model), the expression y_i = ax_i + b is used, and its parameters $\left(\begin{array}{c}a\\ b\end{array}\right)={\left(B^{T}B\right)}^{-1}·B^T·Y_N$ are obtained using the least squares solution—a form structurally identical to MST's internal source computation$S_I={\left(H^{T}H\right)}^{-1}{S}_E={\left(H^{T}H\right)}^{-1}{H}^{T}X$. Here, B is the linear context and Y_N is the input field. However, GM(1,1) is limited to linear context; MST generalizes this to arbitrary context—polynomial, power, logarithmic, periodic functions, or any combination of basic fields—making it a superset of grey system modeling for quantitative identification of various forms of data.

3.4. Catastrophe Theory

Catastrophe theory was founded by French mathematician René Thom, emerging from the convergence of structural stability research and morphogenetic inquiry [4]. Catastrophe theory is based on mathematical theories such as singularity theory and stability theory, and is used to study phenomena of discontinuous change. It is also a theory of general morphology, attempting to provide mathematical models for the occurrence and evolution of forms in nature. Its mathematical proofs are quite deep, but its application models are relatively simple. Its applications are divided into "hard" applications (in natural sciences such as mathematics, mechanics, and physics) and "soft" applications (in fields such as biology and social sciences). Both MST and catastrophe theory address morphogenesis—the former is quantitative, focusing on the entire process from germination to renewal, emphasizing holistic computation; the latter is qualitative, focusing on threshold conditions and catastrophic phenomena, describing "critical conditions" of discontinuous change. The seven elementary catastrophe types can all be expressed using MST's context. Taking the cusp catastrophe as an example, its standard potential function isx⁴ + ax² + bx，and the field functions of the basic fields in the context are y = x⁴，y = x²，y = x, respectively, suggesting a potential formal unification between the two theories.

3.5. Mathematical Morphology

Mathematical morphology was jointly proposed in 1964 by French mathematician and geologist Georges Matheron and his doctoral student Jean Serra from the Paris School of Mines [3]. Mathematical morphology is a theory and technique based on set theory, lattice theory, topology, and random functions, used to analyze and process geometric structures. It is essentially a nonlinear filter, with the basic idea of using structuring elements to "probe" signals, preserving main shapes and removing irrelevant features (such as noise and burrs). Both mathematical morphology and MST address problems based on morphological thinking, but differ fundamentally: mathematical morphology is a signal-level method (set-theoretic operations on images), while MST is a system-level meta-methodology (projection-based identification in arbitrary measurable spaces). The former uses structuring elements for erosion and dilation operations to achieve morphological filtering, segmentation, and recognition of images; the latter uses tensor calculus to achieve quantitative morphological identification through projection. Mathematical morphology originated from engineering needs for analyzing ore properties in iron deposits—a practice-driven theory development; MST originated from modeling morphogenetic processes in neurobiology—a theory-driven application expansion. The mathematical foundation and application scenarios of mathematical morphology can provide a reference for MST, and MST's mathematical framework can encompass its application domains, extending morphological identification from 2D images to high-dimensional spaces, and elevating the idea of "shape-preserving denoising" from image processing to a general system methodology.

3.6. Self-Organizing Mapping Net

Self-organizing mapping net (SOM) are an important type of unsupervised learning neural network, proposed by Teuvo Kohonen of the Helsinki University of Technology in Finland in 1981 [5]. The early research on SOM has theoretical origins with Gabor's associative holographic memories (1969) [48], both involving the use of projection to detect invariant features and other issues. SOM and MST both extract structural information from data, emphasize self-organization, and involve projection. SOM maps high-dimensional input spaces to low-dimensional topological grids through competitive learning; MST projects the input field onto the context through the write operation. SOM preserves the topological structure of the input space through neighborhood functions; MST defines the "morphological constraint environment" through the column space of the context. SOM projects input vectors onto the best matching neuron (best matching unit); MST projects the input field onto the column space of the context. The core difference lies in the dimensionality of projection. In 2010, Resconi pointed out that SOM essentially involves only one-dimensional projection—each input vector is mapped to a single neuron in the competitive layer, compressing high-dimensional information into a one-dimensional discrete grid. This methodology is relatively simple, and its years of development have focused primarily on applications with limited theoretical advancement. In contrast, MST's projection involves multi-dimensional continuous projection—the input field is projected onto a multi-dimensional subspace spanned by multiple basic fields, preserving richer structural information and thus offering greater application potential. MST merits further in-depth research and development on the basis of SOM, extending the idea of "projection as measurement" from neural computing to broader fields of morphological identification.

3.7. Bionic Pattern Recognition

Bionic pattern recognition is a new model of pattern recognition theory proposed in 2002 by Academician Wang Shoujue of the Institute of Semiconductors, Chinese Academy of Sciences, and was further developed into high-dimensional image geometry bionic informatics [6]. This theory aims to solve the problem of how computers can compute problems of image thinking with many independent variables, representing a new approach to developing algorithms for information science. Its core method is to compute starting from geometric figures on many planes in multi-dimensional space, replacing the solving of systems of equations with many independent variables. The theoretical analysis tool for bionic pattern recognition is the study of high-dimensional manifolds in point-set topology. The commonality between MST and bionic pattern recognition is their focus on "geometry and shape," emphasizing the role of image thinking in recognition. MST focuses on the morphogenetic process of forms, and the entire model involves both logical operations (write and read) and the image-thinking operation of reproducing the input field. The reference space, object space, and attribute space it introduces are all multi-dimensional spaces, representing a generalization of tensor calculus. Bionic pattern recognition is based on "recognizing" rather than "distinguishing" things, focusing on the topological properties of sample sets in feature space. MST can draw on the application-driven model of bionic pattern recognition to conduct cross-disciplinary research in high-dimensional image geometry computing and image thinking simulation, advancing morphogenetic theory from "quantitative morphological identification" to "morphological intelligent cognition."

Through comparison, MST is distinctive in three aspects: it provides a unified formal mathematical framework (projection + invariant) for morphogenetic processes; it captures both the "formation" and "evolution" of forms through write and read operations; and it operates at the system level, treating the context as a constraint environment rather than a mere data filter. This makes MST not a domain-specific algorithm, but a meta-methodology for systems whose forms and structures evolve over time.

4. MST as an Analytical Framework for Socio-Technical System Evolution

Although MST has been relatively well established in risk analysis and system identification [1,22,34,37], its recent extension to socio-technical systems reveals broader relevance. This section examines two cases—learning community morphogenesis and legal morphogenesis—to demonstrate MST's capacity for modeling institutional and social system evolution.

4.1. Case 1: Learning Community Morphogenesis

Educational institutions are complex socio-technical systems in which learning outcomes, peer interactions, and institutional norms co-evolve. MST provides a modeling framework for this co-evolution: the learning community constitutes the context (constraint environment); individual learner behaviors constitute the input field; and the projection fidelity (D²) measures the alignment between individual performance and the community's "ideal" context.

Tian et al. [38] operationalized this framework through the Morphogenetic Pyramid Model of Learning Community, integrating Vygotsky's zone of proximal development, Edgar Dale's learning pyramid, and Sheldrake's morphic resonance. The model divides the learning community into three levels (teacher learning community, teacher-student learning community, and student learning community) and uses GPA as a morphological indicator to track group structural convergence. In a six-month intervention involving 72 students in the course "Economic Law and Construction Engineering Regulations," each learning group's GPA increased from 2.47–3.37 to 3.37–3.57, and the overall class average rose from 2.93 to 3.52 (on a 5.0 scale). Within the MST framework, morphological transition is theoretically accompanied by structural convergence of the group toward the context. The observed simultaneous occurrence of "mean increase" and "dispersion narrowing" in this study resonates with the "structural convergence toward context" described by MST theory, indicating that the learning community achieved an observable morphological transition from "medium" to "good" at the group level.

Table 2. Comparison of learning community morphological indicators.

Value	T1	T2	T3	T4	T5	T6	C1
Pre-intervention	2.47	2.68	2.88	2.93	3.23	3.37	2.93
Post-intervention	3.57	3.55	3.51	3.37	3.56	3.56	3.52
Change	+1.10	+0.87	+0.63	+0.44	+0.33	+0.19	+0.59

Note: The form index value represents by students’ GPA (Grade Point Average). Pre-intervention means Initial form index value, Post-intervention means Final form index value, Change means deviation which equals Post-intervention minus Pre-intervention. T1 means Team 1, Team 2 means T2, and so on. C1 means Class 1. Data from reference [38].

Table 2. Comparison of learning community morphological indicators.

Value	T1	T2	T3	T4	T5	T6	C1
Pre-intervention	2.47	2.68	2.88	2.93	3.23	3.37	2.93
Post-intervention	3.57	3.55	3.51	3.37	3.56	3.56	3.52
Change	+1.10	+0.87	+0.63	+0.44	+0.33	+0.19	+0.59

Note: The form index value represents by students’ GPA (Grade Point Average). Pre-intervention means Initial form index value, Post-intervention means Final form index value, Change means deviation which equals Post-intervention minus Pre-intervention. T1 means Team 1, Team 2 means T2, and so on. C1 means Class 1. Data from reference [38].

The mechanism can be explained through MST: the context (learning community norms) is updated through the write operation—teachers and peer leaders encode new study habits, collaborative behaviors, and assessment criteria into the shared context. The read operation manifests as improved learner performance, as individual trajectories more closely project onto the updated context. The observed ‘mean increase’ and ‘dispersion narrowing’ are consistent with the theoretical expectation of structural convergence, providing empirical support for the morphological transition at the group level.

4.2. Case 2: Legal Morphogenesis

Legal systems are institutional socio-technical systems in which norms, enforcement mechanisms, and societal expectations co-evolve over decades. Through years of teaching law-related courses to engineering management students, how to explain "how law is formed" to students lacking a legal foundation has always been a challenge. To address this, the author proposed the Five-Stage Morphogenetic Model [24] of legal morphogenesis (morphogenetic germination—differentiation—coupling—explosion—updating), with MST as its core theoretical tool—a theoretical innovation distilled from interdisciplinary integration and teaching practice. It redefines legal evolution as a "morphogenetic process," providing a "computable and verifiable" analytical framework for understanding the evolution of law from potential order to mature institutions.

Applied to the evolutionary analysis of China's Work Safety Law of the People's Republic of China, a clear morphogenetic trajectory can be identified: the germination stage (1949–1977, scattered administrative regulations), the differentiation stage (1978–2001, emergence of specialized legislation), the coupling stage (2002–2013, formal enactment and formation of the enforcement system), the explosion stage (2014–2020, the first major revision following catastrophic events), and the updating stage (2021–present, continuous iteration and improvement).

This case demonstrates MST's capacity for modeling long-term institutional evolution—understood as a sequence of context updates triggered by critical events such as the 2015 Tianjin Port explosion. When the existing legal context could no longer accommodate the input field (public demand for stricter safety enforcement), the error vector e exceeded the threshold, forcing a legal context update—a typical example of MST's updating algorithm at the institutional scale.

4.3. Synthesis of the Two Cases

Despite their vastly different temporal and spatial scales, both the learning community and the legal system exhibit the same five-stage morphogenetic pattern—germination, differentiation, coupling, explosion, and updating. This suggests that MST captures a general logic of evolution applicable to complex socio-technical systems. The context can be instantiated as classroom norms, legal codes, or organizational policies; the input field as learner behaviors, safety incidents, or stakeholder demands; and the error vector e as the misalignment between existing constraints and emerging requirements.

5. Discussion: Implications, Limitations, and Future Directions

5.1. Methodological Implications for Systems Science

MST provides a formal, traceable, and scalable framework for studying system evolution. Its unique contributions are:

• Traceability.

The write and read operations provide a transparent tracing path—every change in system state can be traced back to an update of the context or a change in the input field. In fields such as economic management, educational intervention, and disaster response, this feature has direct methodological value: when a system exhibits unexpected changes, decision-makers can locate intervention points based on context update records, rather than blindly searching through numerous variables. This combination of "process traceability" and "change attribution" distinguishes MST from black-box system analysis methods.

2.

Quantifiability.

The invariant D² and the error vector e provide real-time metrics of system health. Within the MST framework, system deviation from its expected trajectory is no longer a subjective judgment but a computable quantity. This enables system managers to establish early warning mechanisms at the group level—when the average group error continues to rise, the system signals deviation even before individual indicators show obvious deterioration. This feature has direct application value in risk monitoring and organizational health assessment.

3.

Intervention design.

MST provides normative guidance for intervention: to change a system, one must update the context, not merely modify individual components. In education, this means that "changing the examination system" or "reshaping class culture" is more likely to produce system-level effects than "talking to students one by one." In disaster risk management, it means that "restructuring the risk cognition framework" can more fundamentally reduce system vulnerability than "patching individual hazard points." This principle constitutes a strong departure from reductionist approaches that focus on parts rather than the whole.

5.2. Limitations

Four limitations require acknowledgment:

• Formal complexity.

The mathematical foundation of tensor calculus imposes a high theoretical barrier on practitioners, limiting the dissemination of the method. Some potential users may abandon it due to unfamiliarity with the mathematical language. Subsequent research could address this by developing graphical interfaces or simplified workflows to reduce the pressure of theoretical understanding on practical application, allowing algorithm implementation to be moderately separated from the theoretical framework.

2.

Empirical gap.

Although the learning community and legal cases are illustrative, systematic, large-scale empirical validation is still lacking. Most current cases are concentrated in educational settings, and its applicability in organizational management, public policy, and disaster response requires more empirical support.

3.

Lack of standardization.

There is no standardized algorithm or open-source toolkit available for MST applications. Different researchers may obtain divergent results due to differences in context construction methods, limiting replicability.

4.

Data dependence.

Defining the context and input field requires domain expertise. In contexts with ambiguous or contested boundaries (such as multi-stakeholder policy analysis), how to define system boundaries and select appropriate context attributes remains a practical challenge.

5.3. Future Research Directions

• Tool development.

Design user-friendly software or libraries to implement MST's write and read operations for common applications (such as educational analysis, policy modeling, and disaster assessment). Tools should support importing structured data (e.g., learning behavior records, risk event sequences) and automatically generate morphological evolution maps.

2.

Multi-system validation.

Apply MST to diverse socio-technical systems—organizational change, public health policy, disaster response—to test the generalizability of its methodology. Of particular interest is whether MST's risk evolution analysis can provide earlier warning signals than traditional probability models in disaster emergency management.

3.

Integration with computational methods.

Explore synergies with deep learning and graph neural networks, where MST could serve as an interpretable projection layer for morphological feature extraction. The predictive power of deep learning combined with MST's traceable projection structure may improve model interpretability while maintaining prediction accuracy.

4.

Longitudinal studies.

Track morphogenetic indicators across multiple cycles to validate the predictive capacity of the Five-Stage Morphogenetic Model and identify early warning signals of systemic crises. In educational settings, the trajectories of the invariant D² and the error vector e for the same learning groups across multiple semesters could be tracked to test the predictive validity of the Five-Stage Morphogenetic Model. In disaster management, the evolutionary path of risk forms from "latency" to "explosion" could be simulated to identify critical intervention windows.

6. Conclusions

This paper has employed a combination of literature review and theoretical analysis to systematically trace the four-decade trajectory of MST, to position it in relation to cognate approaches, and to articulate its value as an analytical framework for socio-technical system evolution. The following conclusions can be drawn from this review:

As a generalized system model, MST has matured through three consecutive phases—Stage One: Theoretical Incubation, Stage Two: Proposal and Refinement, and Stage Three: Maturation and Application. The morphogenetic system constitutes a form-generating architecture built upon three mathematical spaces, comprising input fields, output fields, contexts, internal and external sources, projection operators, and invariants. Its core architecture can be summarized as “three spaces—morphogenetic information table—morphic computing.” Morphic computing, in turn, embodies three philosophical propositions: the unity of subjective and objective understanding, the unity of analysis and synthesis, and the primacy of structure over individual values. The theory's most fundamental contribution lies in its formal unification of the contravariant components of tensors with projection operators in transformations—an achievement that positions MST as a holistic meta-methodology for analyzing the evolution of socio-technical systems.

Methodologically, MST is characterized by a holistic orientation that sets it apart from reductionist approaches: it prioritizes the comprehension of overall system structure over the precise specification of individual components. Through systematic comparisons with the information diffusion principle, attribute theory, grey system theory, catastrophe theory, mathematical morphology, self-organizing mapping net and bionic pattern recognition, MST proves distinctive in three key respects: its unified formal language, its capacity to capture both the “formation” and “evolution” of forms simultaneously, and its explicit system-level perspective.

At the level of application, the two case studies examined in this review—learning community morphogenesis and legal morphogenesis—demonstrate that MST is applicable across multiple scales, from micro-level educational interventions to macro-level institutional transformations. This suggests that a general logic of morphogenesis may underlie the dynamics of socio-technical system change.

Despite these contributions, several limitations warrant acknowledgment. Theoretically, the tensor calculus foundation imposes a significant cognitive barrier that constrains the accessibility of the approach to a broader community of researchers and practitioners. Methodologically, the existing literature remains dominated by conceptual exposition and framework construction, with a notable absence of standardized algorithmic procedures and replicable empirical protocols. Empirically, the majority of validation studies have concentrated on natural disaster risk, leaving the cross-domain generalizability of MST largely untested.

Looking ahead, as socio-technical systems in education, governance, and public policy increasingly resist conventional analytical tools, MST offers a promising—and quantifiable—way forward. Future research should prioritize the following directions: extending the application boundaries of MST in disaster emergency management, integrating the framework with deep learning and other computational methods to improve model interpretability, and developing standardized analytical toolkits that can lower the threshold for practical adoption.