English integrated edition · v003
Chapter 39: What This Theory Is Not: Society Is Not a Class Plane but a Differentiation Network
Society forms and exists as a differentiation network. The differentiation network has real standing in social ontology, replacing class entities and a pre-existing class plane. Here, a “differentiation network” means the real, dynamic, open, relational social structures and handling processes, situated in relational fields, that form through the joint operation of people, materials, records, nodes, and relational pathways. It is emphatically not an image temporarily adopted by a researcher, still less an analytical tool that can be replaced at will. At the same time, it is not a fixed, closed, pre-completed, personified ultimate entity that exhausts society; it is not a final explanation of the universe, nature, life, and everything that exists; nor is it a closed, unrevisable final naming of the essence of society.
This theory requires its own limits to be drawn clearly, rejects total images of all kinds, and calls for an account of its real content, the conditions that sustain it, and its theoretical boundaries. The first task is to prevent familiar grand terms from simply being brought back in: if class, identity, or the state is still treated as a pre-existing social skeleton into which differentiating units are then fitted, the questions that motivate the theory have been rewritten. Analysis should begin with the differences, those handling them, and the consequences that actually arise in an event, rather than deciding in advance which image society must conform to. Concepts such as relational field, weight, community, compression, and calibration jointly constrain inference. They require analysts to disclose their materials and scope, specify the actors and actions, and revise their judgments when counterexamples arise.
Social Differentiation Network Theory does not install an eternally unchanging entity beneath society, nor does it treat class, the state, the market, the nation, the family, the platform, identity, or a list of differentiating units as society’s ultimate constituents. Treating a network or differentiating units as yet another fixed ultimate entity to replace the old underlying entity would merely repeat the practice it opposes. Substantialist ontology often first assigns an event a center, then arranges subsequently observed materials around it as subordinate elements. Social Differentiation Network Theory proceeds in the opposite direction: first identify which people, organizations, materials, and procedures are involved in an event, then trace the connections they form.
The author and interpreters who use this theory must trace how the people concerned, organizational members, and concrete procedures handle differences, and determine which materials support a judgment, which nodes change consequences, and which counterexamples require correction. These constraints keep the theory grounded in reality and prevent interpreters from closing relational structures into a new ontological image. Even when analysis temporarily reaches a stable conclusion, it may not infer that an eternal entity lies behind all events.
The fact that relational structures have real efficacy does not mean that they are things that act. School personnel receive materials, family members bring forward memories of care, and institutional personnel make certifications; these concrete actions join into continuing processes that form and sustain social reality. The word “network” is shorthand for the relational form of these real connections. It does not turn the expression itself into an entity. Apart from its carriers, materials, and events, it has no independent will and cannot produce consequences on its own.
For precisely this reason, society is not a class plane, and the consequences of concrete events cannot be accounted for by a single class axis. The real differences involved in a child’s school admission, unpaid wages, and platform labor may include household registration, student registration, materials, accounts, income, family burdens, local recognition, state enforcement, and community boundaries. These differences do not disappear because the class entity is rejected, nor do they constitute the whole or a part of a class entity. In concrete relational fields, people, organizations, and procedures extract, assign weight to, connect, transmit, and write back differences, thereby forming a changing differentiation-network order. These processes account for how social reality forms and exists.
The class plane projects different events onto the same main axis. Qualifications tied to materials in school admission, proof and enforcement in an unpaid-wage case, and account ranking in platform labor are all finally rewritten as manifestations of class position. This compression strips concrete nodes of their independent explanatory standing and pushes responsibility toward an abstract entity. The differentiation-network order, by contrast, retains the direction and variation of relations. A difference may receive greater weight at one node and be downweighted at another; an existing record may be retrieved by later personnel, or it may be refuted by new materials. Because the order is formed through concrete handling, branching, interruption, reorganization, and write-back can occur without assuming a single center. Rejecting the class plane still permits recognition that some labor and resource differences are extremely important in particular events. Analysts must explain where that importance comes from, how it is transmitted into consequences, and under what conditions it changes. This preserves the concrete forms of inequality better than assigning every consequence to class in advance.
A differentiating unit is not a new class, an identity label, a fixed factor, or a social atom. It is a unit of difference that can be distinguished, represented, and compared. Relevant people or procedures may extract it, assign it weight, connect and transmit it, and write it back within a concrete relational field, thereby changing a position, qualification, risk, evaluation, responsibility, action, or consequence. Differentiating units can be subdivided, combined, and represented anew as relations and events change; there is no final closed list. An interpreter cannot first list a set of isolated variables and then declare that society is made of them. The interpreter must state who extracts a difference from which materials, who records or adjudicates it, and which later people or procedures carry the existing result into a new event.
The same difference can be represented differently in different relational fields. Residential circumstances may be recorded as an address, a filing, an account of living together, or local testimony; the facts of labor performed may be carried by wage records, traces of work, and testimony from others. These representations cannot automatically substitute for one another. Analysts need to specify which kind of material the current procedure accepts and which facts of life have failed to enter the record.
Nor is a differentiating unit an indivisible smallest particle. Care can be differentiated further by time, responsibility, bodily exertion, and the ways memory is carried; it can also form new combinations with property, evaluations by relatives, or work arrangements. Whether further subdivision is needed depends on the question at hand. Stopping too early occludes consequences; subdividing without limit loses the main line of judgment. The value of differentiating units therefore lies in making differences traceable, not in supplying a permanently applicable table of social elements. Interpreters who fix commonly encountered differentiating units into a list still fall back into factor ontology. New events can lead to new representations of existing differences and may also bring out differences that previously went unnoticed.
Concrete people or procedures handle, transmit, and reorganize differences along relational chains at multiple nodes. The real relational structures and processes thereby formed are what this book calls a “network.” This requires separating the real network, the theoretical image, and the acting subject. The differentiation network is neither a decorative metaphor nor simply the network diagram drawn by a researcher: that diagram is a limited representation and may be wrong or revised. Family relations, community ties, institutional authority, materials, and records may exist before a particular event, and may operate before an analyst discovers, names, or draws them. Even a researcher’s erroneous image does not cancel these real connections. But the prior existence of a local relation does not imply a fixed, pre-completed total network covering every event.
A differentiating unit is a unit of difference that may enter into consequences. Materials and records carry existing determinations; concrete people or procedures at nodes preserve, identify, certify, or transfer differences according to rules; and extraction, weight assignment, transmission, and write-back are still carried out by people, organizational positions, technical systems, and procedures. The network itself has no will and is not an entity that acts on its own. Only by retaining these divisions of labor can the network avoid becoming the empty claim that “everything is connected.” The foregoing judgment in social ontology thus retains its grounding in reality without personifying the relational structure or treating it as closed. Concrete applications must still be scrutinized against materials and counterexamples, and remain subject to revisions of scope.
In a platform setting, enterprise personnel and procedures form, preserve, and retrieve records of accounts, ratings, recommendations, customer evaluations, and appeals, then carry the results of handling into income opportunities. Family members may preserve or recount care, property, silence, and evaluations of responsibility, raising them again in new disputes. Past records do not act automatically. Only when current personnel or procedures retrieve them anew and assign them weight can they enter present consequences. Preservation as differences that can be extracted again in the future constitutes write-back; actual subsequent extraction and renewed effects depend on further conditions. Immediate consequences do not presuppose write-back. A node must not be understood as a dot on a diagram. It is a position where concrete handling occurs and may be carried by people, an organizational position, or a procedure. Materials may be received, recast, certified, rejected, or preserved at a node. Different actions can send the same difference into different relational chains and thus produce different consequences.
A relational chain is not a simple sequence of before and after. An earlier act of handling may constrain a later procedure, or it may be refuted or recast by later materials. Acceptance of residence materials does not guarantee a place; retrieval of an account rating does not necessarily decide an appeal. Analysts must distinguish connection, transmission, and causal inference. They cannot claim that a consequence has been demonstrated merely because events occur next to one another. A network claim has content only when it can be re-expanded. If an interpreter cannot say who extracted the differentiating unit, where materials were transferred, and how a determination was written back, “network” is only rhetoric. Re-expansion remains limited: the analyst need not restore all background, but must unfold enough to verify the judgment at hand.
Within these concrete processes, family members may put care, property, evaluations by relatives, silence, and responsibility forward again in a dispute. Local residents and community personnel may retrieve familiarity, reputation, testimony, and interaction records in a local event. Members of a national community may maintain boundaries through language, rites and customs, shared memory, shared harm, shared protection, and demands of responsibility, forming intergenerational write-back. These processes cannot all be replaced by relations construed as subordinate to class or by state registration. Personnel in state institutions may handle registration, qualifications, materials, and enforcement in concrete procedures. Personnel and procedures of platform enterprises may handle accounts, ratings, rankings, and visibility. Movement organizers and participants may connect dissatisfaction, risk, memories of action, and imaginings of the future. Social Differentiation Network Theory traces these real actions and their relational consequences; it does not treat the state, platform, movement, or network itself as an actor.
The processes operating within events also include interruptions. Materials may not be received, records may not be transferred, and members may refuse an existing responsibility. Interruption is not an exception outside the network; it changes whether a difference can be transmitted and who bears the consequence, so analysts must trace connection and rupture together. The differentiation-network order may also reform over time. A past determination retrieved by current personnel may be supported or limited by new materials; new handling leaves records that can be put forward again in the future. A historical index provides an entry for retrieval but does not automatically prescribe the conclusion of every event. Describing society through relational structures does not mean that every person is directly connected to everyone else. Many consequences travel through materials, organizational positions, and layers of relays, while other relations are completely broken. Analysts must identify the actual relays instead of replacing concrete explanation with “everything is connected.”
Identity names can serve as entry points for analysis, but cannot replace the tracing of mechanisms. Faced with an “outsider,” an interpreter must determine who retrieves materials concerning residence, filings, qualifications, or local recognition, and in what event; through which nodes these materials pass; and how they change opportunities. Whether landlords cooperate with filings, how school personnel examine proof, and which procedures community personnel follow in processing applications can all change the real weight of the same name. An identity does not itself extract materials or confer consequences. All handling is still carried out by concrete people and procedures.
The convenience of using a name to gather similar experiences must not be mistaken for the existence of a shared mechanism. Two people both labeled “outsiders” may face entirely different landlords, material requirements, school nodes, and local relations. A shared name only indicates where to look; it is not enough to predict the same weights or consequences. “Platform laborer” is also only an entry point. In order-taking, ranking, or appeals, enterprise personnel and the relevant procedures may retrieve account ratings, response times, completion rates, and customer evaluations, then carry the results into subsequent recommendations and income opportunities. Who forms the records, how they are retrieved again, and how they affect household cash flow must be traced. Without these connections, classification still does not amount to explanation.
Identity itself may also be remade in the course of handling. People submit materials, staff select fields, and institutional procedures form records; when later personnel retrieve them again, the visible content of the identity may change. This is not identity growing by itself. The force of classification comes from concrete handling and from the way its results enter subsequent relations. Administrative records, people’s own accounts, and other people’s evaluations are each formed in different ways. Simply mixing these names into an identity collage does not explain how qualifications, responsibility, or risk change. Nor does an identity submit itself to examination. Interpreters must still determine who extracts it on the basis of which materials, who or what procedure gives it consequences, along which relational chains it is transmitted, and which differences it occludes. Only when the relevant records are retrieved again in future events can they participate in new determinations and write-back.
Differences not covered by a name—family support as a fallback, bodily condition, local familiarity, language ability, and past records—may change the real position associated with the same identity. Adding more identity names without tracing how they enter consequences leaves classification where it was, however fine its distinctions become. Listing poverty, household registration, education, family, the state, the market, the platform, locality, and the nation, together with class, school, or movement, may look comprehensive while still explaining no consequence. A longer list is not necessarily closer to reality.
Factor lists usually place matters from different levels side by side. Poverty may be a condition of life; household registration and student registration may appear as records; families and states involve many kinds of members, institutions, and procedures. Placing them in the same column readily occludes how differences are represented and obscures who has the authority to receive, reject, or reinterpret materials. A list may be retained as a reminder, but analysis must enter the relational chains: how differences connect, at which node a connection is interrupted, which part of an inference new materials may change, which differentiating units gain weight, which are downweighted or occluded, and which counterexamples would overturn the current judgment. Names cannot answer who did what or what consequences followed.
In a child’s school-admission situation, parents provide proof of residence, landlords cooperate with or refuse a filing, and school personnel examine materials according to procedure and handle school places and application schedules. The results of application processing may be preserved by the relevant admissions personnel or procedures. After students register, student-registration personnel separately preserve student-registration information and records of the corresponding changes in accordance with the rules, for future retrieval. This relational chain changes qualifications, time pressure, and future opportunities, but it is not a fixed checklist. How materials operate must still be judged in light of the event, procedures, and counterexamples. Early in the application process, whether materials are complete may directly affect their acceptance; at later stages, school places, continuity of student registration, or a family’s relocation may become the principal consequences. Weight changes dynamically with events and handling. It is neither an inherent property or score of a difference nor something that can be read directly from a factor’s name.
Writing every factor into the text without stating which enters consequences at the current node still evades the distinction between primary and secondary. Weight is neither a permanent ranking nor a numerical allocation. It must be judged by how personnel or procedures handle materials and how the results are transmitted into subsequent relations. Whether an identity is retrieved, a piece of material is recognized, or a difference gains weight through a new record must each be traced step by step. This may reveal several relational chains, but explains reality better than piling up names. The same holds for taking orders, school admission, and unpaid wages.
The class entity does not exist. Real differentiating units—labor conditions, wages, income, employment, property, debt, housing, available exits, relations of domination, and risk—do exist and may produce grave consequences in concrete relational fields. But these real differences do not constitute the whole or a part of a class entity. Opposition to the class entity is in no way a denial of the objective reality of unpaid wages, exploitation, poverty, injustice, or the shifting of risks. It is a refusal to bind concrete conditions a priori to a name purporting to designate an entity.
An interpreter who adopts class theory extracts a set of high-weight differentiating units, assigns them weight, connects them, subjects them to low-dimensional compression and projection, and names them. The local explanatory capacity comes entirely from these real differentiating units and the processes of handling them; the class image itself is only a low-dimensional output. An interpreter may indeed identify certain labor and resource conditions through what the image marks, but this local explanatory efficacy can never prove the existence of a class entity, still less serve as an excuse to ignore household registration, student registration, accounts, local recognition, family burdens, state enforcement, community boundaries, and historical write-back.
The local explanatory capacity formed by real differentiating units and the processes of their extraction, weight assignment, connection, and compression must be strictly distinguished from class theory; the class image is only a low-dimensional output. Class theory is not Model A or Model B. From the outset, its basic structure belongs to the Model C direction of total-image subsumption. Interpreters who adopt class theory elevate a limited image into society itself and thereby demote, dissolve, or cancel other valid explanations. Class theory does not become Model C only after overuse, expansion, or abuse. This distinction must therefore never be softened into the compromise statement that “class exists, but is not complete enough.”
Real differences that can be traced separately—labor, income, debt, property, housing, domination, and risk—may connect with one another in a concrete event and may simultaneously carry greater weight. Yet these connections still generate no class entity. Local distinction extraction cannot become a refuge for class theory. Faced with a class image, analysts may bring some of the labor and resource differences it marks back into a concrete relational field and examine, one by one, mechanisms involving household registration, student registration, family, national community, state enforcement, local recognition, and platform procedures. If all the materials are ultimately taken back into the total class image, local analysis has become only an entrance to subsumption.
Opposition to class theory in no way removes responsibility for inequality. Unpaid wages, deprivation of resources, shifting of risks, and relations of domination can all produce real and grave consequences. The point is that interpreters should directly trace how concrete differences are extracted and assigned weight. They must not replace concrete responsibility with a non-existent class entity, nor allow the class name to occlude other mechanisms of oppression that are present.
Social Differentiation Network Theory does not bring other explanations to an end. Explanations concerning the state, market, family, nation, school, platform, and movement may point to real differences within limited scopes. This does not entitle them to be judged equally effective in advance or to receive priority solely on the strength of their traditional standing. Model A and Model B designate only two limited explanations on the same level; neither may be presumed superior. Real societies, institutions, and outcomes are the objects they explain and compare. The labels A and B must not be treated as two separate real societies or institutions. Interpreters may compare limited images, but whenever they use an image, they must disclose what they have extracted, downweighted, or occluded, which events their conclusion applies to, and which counterexamples would force them to revise it. They must still compare the images on the basis of materials, scope, consequences, counterexamples, and capacity for revision.
Limited use requires distinguishing explanatory images from real carriers. An abstract name can help with generalization. In causal judgment, however, actions must still be attributed to people, organizations, and procedures; they cannot be left for abstract nouns to perform on their own. Opposing the subsumption of reality by an image does not mean opposing every relation that image concerns. Analysts must retain mechanisms and consequences. What they reject is elevating a local image into the only answer. Retaining other explanations also does not mean that the author abandons the theory’s own judgment. Social Differentiation Network Theory still insists on returning to the real processes through which differences are extracted, assigned weight, connected, transmitted, and written back.
Explanations supported by different images need not have the same scope. Whether they can be used depends on the current event and materials; no permanent hierarchy can be assigned in advance. An explanation using a state image may have greater explanatory power in matters of material certification and enforcement. Personnel in state institutions can handle registration, certification, and enforcement; the state does not perform registration on its own. Criticism of procedures cannot deny their real efficacy. Nor may the use of this explanation fuse different institutions and procedures into a single will or ignore families, markets, localities, and platforms. School personnel and procedures can handle student registration, evaluations, and qualifications; enterprise personnel and procedures can handle accounts, ranking, and visibility. Each form of handling has its place. When using a platform image, bodily conditions, care, state real-name identification, and market pressure must still be retained.
Other differences emerge in care and responsibility. Family members can preserve care, property, responsibility, and long-term evaluations; an explanation using a family image may better account for the care and responsibility involved. Criticism of family oppression cannot cancel care, and acknowledgment of care cannot make responsibilities among relatives inherently legitimate or imply that all members bear them equally. Market participants can shift risk through transactions and prices; the market does not transact by itself. The use of a market image must not ignore care, responsibility, memory, or dignity. Movement organizers and participants can connect dissatisfaction, risk, and memories of action, but those using a movement image must not flatten differences in positions of participation or costs of participation.
Names of nations may be compressed or put to use, but criticism of national essentialization cannot cancel real communities. Their formation may involve long-term interaction and multiple levels of relays, and may proceed through language, rites and customs, shared memory, shared harm, shared protection, demands of responsibility, experiences of boundaries, and intergenerational write-back. State registration is not the only source of a community. Interpreters using a national image may extract long-term memory, shared harm, and shared protection. Such an explanation must preserve real national communities while also confronting internal differences and oppression within communities. It must not assume that shared memory is entirely uniform or allow a shared name to swallow up different positions and harms.
A limited image that flattens its own internal structure also requires calibration. Different explanations can constrain one another: materials extracted through one image may become counterexamples to an explanation supported by another. A consequence within a family may also force an explanation in terms of state procedures to be unfolded again. Mutual constraint does not require combining several images into a larger total image; it makes their respective scopes of application clearer.
Real differentiating units—labor, wages, property, debt, housing, risk, and domination—and the processes of their extraction, weight assignment, connection, and compression can form local explanatory capacity; the class image is only a low-dimensional output. This local distinction extraction does not prove that a class entity exists and does not constitute a repair of class theory. As a total image, the basic structure of class theory belongs from the outset to Model C. It cannot be reduced to Model A or Model B in the name of “limited use,” nor thereby returned to the ranks of ordinary limited models.
If a person, family, group, state institution, or platform enterprise is first judged good or evil and the evaluation is then used in place of mechanism analysis, the explanation stops at moral judgment. The interpreter needs to trace who extracted a difference, formed a record, and assigned a consequence; who received recognition or suffered exclusion; and how the current determination was preserved as material that could be retrieved again in the future. Moral and responsibility judgments may arise, and real harm cannot be canceled. Family members may oppress others by using interpretive authority, memories of care, or arrangements of property. Personnel and procedures of platform enterprises may also produce injustice through accounts, ratings, rankings, appeal deadlines, and income opportunities. After identifying a consequence, one must still explain its formation mechanism rather than stopping at “this family is bad” or “the platform is bad.”
Mechanism analysis does not cancel moral responsibility. Who formed a record, who rejected materials, who set a procedure, and who continued the handling after knowing its consequences may all affect the judgment of responsibility. Only when analysts make actions and consequences clear can responsibility be located in real carriers, rather than being displaced by abstractions such as “that is how the institution works” or “that is how the market works.” In the same event, harm and responsibility may also be distributed across different nodes: one person makes the direct decision, another maintains the record, and a later procedure enlarges the result. Analysts should not place all responsibility on a single person, nor say that no one is responsible merely because the relational chain is long. The scope of responsibility must be judged by action, knowledge, position, and actual consequence.
When moral evaluation precedes the materials, counterexamples are easily treated as excuses. Mechanism analysis requires first retaining the response of the person concerned, records of conflict, and procedural materials, and then determining how harm formed. This does not reduce the force of criticism; it prevents names of good and evil from replacing proof. At the same time, Social Differentiation Network Theory contains the author’s definite judgment about social reality and must not be weakened into an arbitrary analytical tool or a list of methods. Relational fields, materials, nodes, and relational chains can be entry points for research, but they are not the whole theory. Neither the author nor its users may replace an affirmative account of how society forms with operational convenience. What is drawn here is the boundary between method and theory.
In concrete research, this theory cannot replace historical research, institutional analysis, field investigation, economic analysis, legal and political judgment, or literature and lived experience. The author thereby states a bounded theoretical position and a requirement of calibration, enabling interpreters to make limited, traceable, comparable, and revisable judgments. It is not the ultimate answer that explains everything. This theory cannot obtain materials for a researcher. The experience of the person concerned, the background in which a document was formed, the actual operation of a procedure, and historical records each require their own research methods. Social Differentiation Network Theory supplies a direction for questions and judgments: it tells researchers which differences, carriers, and relational chains to trace, but it cannot replace actual verification.
Nor does it guarantee that every event will yield a complete conclusion. Some materials cannot be obtained, some relational chains have been interrupted, and other consequences are still forming. Analysts may offer a restricted judgment and state what remains unknown. If they add connections unsupported by materials merely to appear complete, the theory slides from limited explanation into self-closure. There is a clear boundary between “usable” and “explaining everything.” This theory can assist analysis of difference-handling in families, localities, nations, states, markets, schools, platforms, and movements, but it cannot guarantee in advance that the same set of concepts occupies the center of every question. The relational field changes, and the focus of analysis must change with the materials.
Limited judgment is not arbitrary choice, and still less does it mean that all explanations are the same. The interpreter must disclose the materials used, the relational-field window, the differentiating units extracted, the basis of weight, the nodes, relational chains, scope of application, and conditions for correction, so that different limited explanations can be compared by explanatory power, comparative power, revisability, and differentiation bias. Limitation requires the interpreter to distinguish the strength of judgments: one relational chain may already have several mutually supporting pieces of material, another inference may rest only on a statement by the person concerned, and some nodes may not yet have supplied a response. Keeping these states separate allows a judgment to stand provisionally while preserving room for later correction. Different explanations may answer different questions. An explanation using a family image may better explain care responsibility, while one using a state image may better explain a qualification procedure. Analysts must still check whether each extracts the key differences, whether it can be re-expanded to the materials, and whether it has rejected counterexamples. Comparison does not require compressing them first onto one scale.
Acknowledging that society is complex, that materials may be omitted, and that judgments need revision is a normal awareness of limitation. It does not belong to Model C. Complexity does not mean that all differences have the same weight at the same time: family care, qualifications tied to materials, income pressure, and local recognition can coexist, but the times and paths by which they enter consequences may differ. Analysts must identify the relational chains currently operating and mark where materials have not yet been obtained. A healthy awareness of complexity makes judgments more precise. It requires interpreters to acknowledge omissions, rather than lay out every difference in equal detail. Certain differences may temporarily remain outside the relational-field window, provided the reasons for omitting them are disclosed and counterexamples are allowed to bring them back into analysis. People, institutions, and concrete procedures can preserve records, retrieve materials, make certifications, transmit determinations, and write the results of handling back as differences that can be extracted again in the future. Differences, history, and records do not perform these actions themselves. Preservation as differences that can be extracted again in the future constitutes write-back; actual subsequent extraction and renewed effects depend on further conditions. Immediate consequences do not presuppose write-back. New materials or later events may still lead the people concerned and interpreters to question an existing write-back. Write-back is therefore not a permanent settlement.
Neither a multiplicity of differences and nodes nor the complexity of relational chains can justify abandoning judgment. The problem with relativism is not acknowledging multiple viewpoints but canceling the difference among viewpoints. If sufficient materials and missing materials, revisability and refusal of revision, limited validity and arbitrary piecing together are all treated as the same, an explanation cannot bear responsibility. Social Differentiation Network Theory retains multiple limited explanations while requiring them to disclose what can be examined.
Model C comprises mechanisms operating in two directions: total-image subsumption and the erasure of distinctions between explanations. When interpreters elevate a limited image into society itself, the mechanism of total-image subsumption comes into operation; when they use complexity to cancel comparison, counterexamples, and revision, the mechanism for erasing distinctions between explanations comes into operation. These mechanisms move in opposite directions but have a common function; they are not necessary stages through which a process must pass in sequence. Ordinary uncertainty, suspended judgment, insufficient materials, and valid objections are emphatically not equivalent to Model C. Complexity-based conflation illicitly substitutes “judgment is impossible” for “not yet complete.” Once all explanations are described as limited and equally unsuccessful, the strength of materials, errors of inference, and capacity for revision can no longer be distinguished. Valid models, erroneous models, and garbage models are thereby placed on the same level. The erasure of distinctions between explanations appears to respect complexity, but in fact cancels criticism of erroneous explanations. Subsumption and the erasure of distinctions move in opposite directions, yet both destroy the conditions for comparison, revision, and calibration among limited explanations.
The originator and users of Social Differentiation Network Theory face these two risks as well. If they treat differentiating units as final entities, elevate the network claim into an all-purpose image, or demote all other theories into subsidiary materials of their own, the theory will slide toward Model C’s total-image subsumption. Preventing total-image subsumption requires interpreters to state which questions are not to be answered by this theory alone. Historical context, legal responsibility, institutional detail, and the experience of the people concerned still require research through corresponding materials. The theory may organize questions, but it cannot reduce other research to a storehouse of examples for itself.
The other risk is to use social complexity and the limits of explanation to cancel comparison, counterexamples, and revision among models, presenting valid models, erroneous models, and garbage models as equally unsuccessful. Such a move will lead toward Model C’s erasure of distinctions between explanations. Healthy humility does not fall into this category. Interpreters must still disclose boundaries, offer provisional judgments, and accept calibration; they cannot cancel valid comparison in the name of humility. Preventing this erasure of distinctions requires interpreters to make judgments open to criticism. After acknowledging what is unknown, the interpreter must still say which material currently has greater supporting force, which relational chain is more likely to hold, and what counterexample would cause the conclusion to be withdrawn. If only “everything is complex” remains, the theory too loses its comparative capacity.
The originator and users of the theory must allow other limited explanations to be compared with it within concrete scopes. They must ensure that their judgments can be re-expanded, can accommodate counterexamples, and can be revised in the light of new materials. Self-calibration does not mean repeatedly rewriting the core position. Class entities do not exist; class theory belongs from the outset to Model C’s direction of total-image subsumption; a healthy awareness of complexity does not belong to Model C. These are established theoretical boundaries. Revisions at the level of individual cases address the corresponding concrete judgments and scopes on the basis of materials. They must never arbitrarily turn an explicit theoretical position into a vague compromise.
At the same time, when the foundational proposition itself faces a substantive challenge, the theory cannot evade examination by preemptively demoting the problem to an error in an individual case. It must undergo rigorous criticism and delimitation at the level to which the problem belongs. This self-limitation does not soften the judgment that class theory belongs from the outset to Model C in its total-image-subsumption form. On the contrary, it can effectively prevent Social Differentiation Network Theory from insulating itself against criticism.
This theory distinguishes two different levels: an affirmative judgment in social ontology and a way of judging constrained by reality. The former affirms that society forms and exists as a differentiation network and that the differentiation network has real standing in social ontology. The network is not the ultimate reality of the universe, a fixed and closed total network, or a super-entity that acts on its own. The latter requires interpreters to calibrate their judgments on the basis of materials and consequences when they confront concrete events.
The book’s judgments do not all stand on the same grounds, so their examination has layered requirements. Concepts must first remain consistent in their use across the text and be able to face boundary cases. An ontological judgment about how society exists must state which carriers, continuing relays, memory, responsibility, and consequences support real relations.
Once one enters a concrete event, materials, comparison, and contrary cases determine whether an empirical conclusion should be maintained, reduced in strength, narrowed, or withdrawn. A research approach must also allow readers to trace the basis on which a concept was formed, return to materials, nodes, and relational chains for re-examination, and retain the original judgment together with the reasons for its revision. Normative claims about what ought to be, priority, rights, or duties require their own value grounds, rules for conflict, and exceptions; they cannot be inferred directly from relational facts.
These checks constrain one another but do not replace one another. If a relational-chain, weight, or write-back judgment in one event fails, the first requirement is to reduce the strength of, narrow, or withdraw that application-level conclusion. If a meso-level mechanism cannot account for the materials, its scope must also be adjusted. If a core term expands and contracts arbitrarily across settings so that different social structures can no longer be distinguished, the problem rises to a loss of conceptual discriminating power. The challenge faced by the overall proposition in social ontology is broader still: it cannot establish itself merely by naming every relation after the fact as a differentiation network, nor is it immediately refuted by a mistaken judgment in an ordinary case.
This foundational proposition must still accept real criticism. Concepts must remain consistent across contexts and distinguish themselves from the empty claim that “everything has relations.” They must make substantive distinctions among different social structures and provide explanatory content beyond “society is made of relations.” If social continuities or consequences remain insufficiently explained by extraction, weight assignment, connection, transmission, preservation, and write-back, the theory should acknowledge what it has not explained. If a competing social ontology can explain the same materials with fewer assumptions and clearer boundaries, that constitutes a genuine challenge. Faced with continuing counterexamples, the theory must be able to narrow its scope of application, adjust its concepts, or withdraw an invalid meso-level explanation.
If continuing counterexamples or a competing social ontology reveal problems that bear on the foundational proposition itself, the theory cannot preemptively reduce all these problems to misjudgments in individual cases or failures of meso-level mechanisms. It must openly explain why the relevant materials can or cannot still be explained by this theory, and on that basis clarify the foundational proposition’s boundaries of application, what remains unexplained, and which formulations may require revision. Accepting such a challenge does not mean that an ordinary case alone can overturn the foundational proposition, nor does it mean that the book thereby withdraws its established judgment in social ontology. Such examination both prevents the foundational proposition from insulating itself against criticism and preserves its theoretical level, which differs from that of a one-off empirical hypothesis.
When facing concrete events, this way of judging does not require a fixed order and does not constitute a table of steps, decision tree, score sheet, or algorithm. The concepts provide dimensions of calibration; they must not be arranged as an operating procedure that automatically generates answers. Analysis of some events starts with conflicting materials; for other events, an objection from the person concerned may provide the first opening into the relational field. Community mechanisms, weight, and historical indices may also appear in different sequences depending on the question.
In concrete relational fields, the responsibility of analysis remains grounded in real constraints. Interpreters must explain which people, materials, and consequences an event actually involves. In undertaking limited distinction extraction, they must also state which differentiating units they have extracted and what they have downweighted, and explain which real relations provide the basis for weight: weight is not a numerical value innately carried by a differentiating unit. Limited does not mean arbitrary; it makes responsibility for judgment more concrete. Analysts cannot select only differences that suit their position or arbitrarily change the scope after counterexamples appear. They must give reasons for their selections, preserve downweighted materials, and explain when their conclusions cease to hold.
This requires tracing nodes and relational chains through concrete carriers: who preserves records, who retrieves materials anew in the current event, who certifies or passes them on, and who preserves the results of handling as differences that can be extracted again in the future, thereby forming write-back. The past does not automatically govern the present, and write-back does not permanently determine the future. New materials and subsequent events can still open possibilities for objection, response, and revision. Likewise, identifying a community requires examining whether long-term interaction, memory, recognition, boundaries, shared responsibility, and shared burdens are actually sustained by members and their relay relations. Explanatory compression must also specify its scope and limits of validity. It must neither elevate a total class image or any single image into society itself nor use complexity to cancel valid judgment.
Social Differentiation Network Theory does not claim to have explained the whole of society. It retains a theoretical position that allows limited distinction extraction, disclosure of grounds, comparison of explanations, acceptance of counterexamples, and revision of judgments. Concrete people, organizations, and procedures can change the connection, transmission, and write-back of differences, bringing about change, branching, rupture, or reconnection in the differentiation-network order. Any theoretical judgment can hold only within the conditions set by its materials, scope, and counterexamples; the analytical window must also be adjusted to the question and materials. Others can still examine these judgments by following concrete questions: where do the materials come from, who handled the differences, which node changed the consequences, which records were retrieved anew in the current event, and which objections call for revision? As long as these questions remain open, the theory will not pass off the results of its own compression as society itself.