The IPS-based theory of sociology

Sociology studies how households, workplaces, movements, and communities cohere and collapse. This essay shows that every major sociological finding — from Durkheim’s anomie and Putnam’s social capital through Karpman’s drama triangle at group scale and Weber’s bureaucracy — is a reading on the Fractal Persistence Equation (FPE) at level \(L+1\) and \(L+2\): the social layer between individual psychology and the nation-state.

Formal derivation: information_persisting_systems.md. Vocabulary: glossary.md. Individual scale: ips_psychology.md. State scale: ips_geopolitics.md. Full fractal stack: fractal_layers.md.

1 The equation at the social layer

At the social layer, members are \(\Phi\) (people, skills, trust stock, tacit know-how), the collective is the node, and markets, states, and larger institutions are \(\Psi\).

Why a household, firm, or movement is an IPS

The equation applies at the social layer for the same reason it applies at the agent layer: a household, workplace, or movement satisfies the four clauses of Definition 2.1 in the formal paper, only now the Markov blanket is social rather than physical.

1. Boundedness. The collective has a Markov blanket \(\partial\Sigma\) defined by membership rules, roles, and the boundary of what the group will and will not share with outsiders. Conditional independence (2.1) holds: given the group’s shared observations and collective actions, its internal state (who is in good standing, what the group believes, what has been settled) is statistically separable from the rest of the world.
2. Drive. The collective is a non-equilibrium steady state: it imports free energy (wages, volunteer hours, member attention, food, energy) and exports entropy as decisions made, work done, waste, and churn. Stop the fluxes and the pattern dissolves — a household with no income, a firm with no payroll, a movement with no meetings. The \(\tau_d\) bound of Theorem 5.1 sets how long the group survives a flux interruption.
3. Internal model. The collective carries a shared generative model \(q_\mu\) — the official story, the charter, the role system, “how we do things here.” Coordination is well-described as the group acting on this model; when the model mismatches reality, the group generates collective surprise. This is the formal content of “culture.”
4. Identity preservation. The collective persists as the same coarse-grained pattern (a recognisable family, firm, or movement with a continuous history) over \(\tau \gg \tau_{\mathrm{relax}}\), where \(\tau_{\mathrm{relax}}\) is the group’s coordination-update cycle (a meeting, a quarter, an election). Members come and go; the pattern persists. This is what makes the group, not its current roster, the node.

Because the four clauses hold, the collective is governed by (4.4). Sociology is not a separate law beneath the thermodynamics — it is the set of substitutions that translate each FPE term into the vocabulary of collective life.

A group persists not because everyone is happy but because \(\mathcal{R}^{(\text{collective})} \ge 1\) — the pattern of roles, stories, and couplings continues as a recognisable node. The group fails when the denominator grows faster than the numerator, regardless of how compelling the official story sounds.

The mechanism is Theorem 5.1 read at this scale: when usable income \(P_{in}\eta\) falls short of required dissipation \(\omega\mathcal{E}_\Sigma(1+\mathcal{D}_{KL}+\Gamma)\), the collective’s internal free-energy reservoir (its trust stock, cash reserves, member goodwill) depletes at rate \(\dot{\mathcal B} \le -\delta P_{out}^{req}\). The “borrowed time” is the bounded quantity \(\mathcal{B}/(\delta P_{out}^{req})\) in (5.1): a firm burning cash, a family burning goodwill, a movement burning activist energy each have a finite runway before the macrostate leaves its equivalence class — bankruptcy, rupture, or dissolution.

The two operations

Almost every effective social repair — family meeting, post-mortem, union negotiation, truth commission — reduces to:

1. Lower \(\mathcal{D}_{KL}\) — align the collective model with observables. Say what is actually happening.
2. Lower \(\Gamma\) — resolve, renegotiate, or exit unresolved friction. Settle what is owed, or cleanly decouple.

The thermodynamic reason these two — and only these two — recur is that they target the only two denominator terms the collective can actively reduce. \(\omega\mathcal{E}_\Sigma\) is largely fixed by the group’s structure and environment (you cannot will away regulatory complexity or market volatility); \(P_{in}\) is bounded by budget and circumstance. But \(\mathcal{D}_{KL}\) and \(\Gamma\) are informational debts that the collective services either by paying the Landauer cost of an honest update once or by paying the continuous suppression cost forever.

This is the content of Theorem 5.2. Each nat of collective delusion divides the group’s lifetime budget by a factor of \(e\) (equation 5.2). The mechanism: a wrong shared \(q_\mu\) forces the group to act on predictions that mismatch reality, generating surprise on every coordination cycle; by the Crooks/Jarzynski identity (3.9) each nat of surprise costs at least one nat of excess dissipation per cycle. The group pays Landauer three times — once to act, once to be wrong, once to suppress the mismatch — instead of once to update. The honest update (a truthful post-mortem, a candid family meeting, a commission’s findings) is a one-time erasure cost (\(k_B T\ln 2\) per bit, 3.4); suppression is that same cost charged every cycle, compounded by the \(\Gamma\) it generates. Over any horizon longer than a single cycle, paying once is cheaper than paying forever.

\(\Gamma\) — structural fatigue — is the backlog of micro-conflicts the collective has absorbed but not repaired (grievances, silos, frozen roles, unpaid relational debts). It enters the denominator additively and, because each unresolved item keeps generating fresh surprise, the outstanding set \(C\) grows quadratically in cost: \(n\) outstanding conflicts do not cost \(n\) times one conflict, they cost roughly \(n^2\), since each pair can co-trigger (the same conversation re-opens both wounds; one grievance primes the next). Operation 2 (settle or exit) does not just remove one item; it removes the cross-terms it was generating. This is why early repair is wildly cheaper than late repair, and why groups often feel “suddenly” terminal — the cost curve is convex and the phase transition in \(\mathcal{R}\) through 1 is sharp.

Rituals, bylaws, and “culture” are implementations. The persistence ratio is indifferent to vocabulary.

2 Households and families

A household is a level-\(L+1\) node whose \(\Phi\) is the nervous systems of its members and whose persistence depends on shared models staying cheap enough to coordinate. Families persist not because everyone is happy but because the pattern of roles, stories, and couplings continues.

Classical concepts, FPE read

Emotional labour (Hochschild) is \(P_{in}\eta\) spent on others’ denominators — lowering \(\mathcal{D}_{KL}\) and \(\Gamma\) for the unit while the labourer pays the cost on their own substrate. The mechanism is the same conservation enforced by (3.3) that drives people-pleasing inward (see ips_psychology.md §4): the household is a NESS and must export its entropy. Honest coordination work — de-escalating a conflict, translating a partner’s complaint into something hearable, absorbing a child’s distress — is a legitimate entropy-export channel that lowers the collective’s \(\mathcal{D}_{KL}\) and \(\Gamma\). When that work is unequally distributed, the labourer’s own substrate pays the thermodynamic bill: their \(\mathcal{B}\) depletes, their \(\Phi\) erodes, and by Theorem 5.1 their personal \(\mathcal{R}\) drifts toward 1 while the household’s looks stable. This is substrate export upward from individual to collective appearance — the group’s \(\mathcal{R}\) clears by borrowing against a member’s \(\Phi\).

The identified patient holds unresolved friction for the subsystem so adults avoid the two operations with each other. The child’s \(\Phi\) pays — symptoms, school failure, acting out — while parental \(\Gamma\) stays artificially low. Same geometry as the drama triangle Victim at group scale.

Intergenerational transmission is \(\Phi\) routing through time. When parents run high \(\mathcal{D}_{KL}\) or export \(\Gamma\) without repair, the next generation inherits higher default \(\Gamma\) in close coupling, depleted \(\Phi\) (hypervigilance, chronic stress load), and learned priors that “truth is dangerous” or “repair never happens.” The mechanism is Theorem 5.3 (fractal composition): the child is a level-\(L-1\) sub-IPS of the parent’s level-\(L\) node during dependency, so the parent’s persistence condition supervenes on the child’s. But composition runs both ways — the child’s learned internal model \(q_\mu\) is set by the parent’s behaviour, and that prior becomes the child’s default policy for managing \(\mathcal{D}_{KL}\) and \(\Gamma\) in their own future dyads. A parent who suppresses rather than updates teaches, by example, that suppression is the affordable move; the child inherits not only depleted \(\Phi\) but a \(q_\mu\) that will keep generating \(\Gamma\) until it is itself revised. Parent persistence supervenes on child persistence, and child persistence is shaped by parent modelling (Theorem 5.3 in the formal paper).

Dysfunctional equilibria persist when short-term \(\mathcal{R}\) for role-holders beats the cost of the two operations: avoid truth (hide \(\mathcal{D}_{KL}\)), avoid repair (defer \(\Gamma\)), export to one member. Each arrow is entropy export — someone else pays later.

Ostracism in families is imposed decoupling: cutting a member’s shelter \(\Psi\) from the collective while often avoiding the real conflict. The silent treatment, shunning, and cut-off lower collective \(\Gamma\) briefly while the target’s \(\mathcal{E}_\Sigma\) rises and \(\Phi\) erodes. Unlike voluntary exit (Hirschman), ostracism inverts the sequence — the group decouples first, while often demanding loyalty from the target without restoring shelter. That is among the most expensive coupling geometries: high \(\Gamma\), high \(\mathcal{D}_{KL}\), no voice, no legitimate exit.

3 Workplaces and organisations

Organisation theory asks how firms coordinate thousands of strangers. Weber described bureaucracy; March and Simon modelled bounded rationality; contemporary management tracks psychological safety and moral injury. FPE compresses the question: does this level-\(L+2\) node still satisfy \(\mathcal{R} \ge 1\)?

Firm mortality is exponential

Firm mortality approximates exponential decay (Daepp et al., 2015) — consistent with a persistence ratio that drifts stochastically until it crosses one. The link to Theorem 5.1 is direct: a firm is a level-\(L+2\) IPS whose internal reservoir \(\mathcal{B}\) (cash, trust stock, tacit know-how) depletes whenever \(P_{in}\eta < \omega\mathcal{E}_\Sigma(1+\mathcal{D}_{KL}+\Gamma)\). Under stochastic drift of the ratio, the first-passage time to \(\mathcal{R}=1\) is exponentially distributed when the drift is small and roughly constant — exactly the regime empirical firm-mortality data exhibit. This is not random; it is the FPE: when \(\mathcal{D}_{KL}\) and \(\Gamma\) compound unchecked and \(\Phi\) (employee trust, tacit know-how) drains faster than it can be replaced, collapse is the predicted outcome, and the shape of the hazard rate (exponential, not Gaussian) is the signature of a depleting reservoir crossing a threshold rather than a smooth decline.

KPI theatre as institutional delusion

When metrics diverge from witnessed \(P\) on the floor, workers hold high \(\mathcal{D}_{KL}\) between institutional \(Q\) and daily reality. Energy that could lower collective surprise instead goes to maintaining the fiction — paid as burnout by the employees. The mechanism is Theorem 5.2: each nat of maintained gap between the official \(Q\) and floor \(P\) divides the firm’s effective lifetime budget by \(e\). The cost is not abstract — it is paid as excess dissipation on every coordination cycle (3.9), routed onto employees’ \(\mathcal{B}\) and \(\Phi\) as chronic stress and burnout, because the firm as a NESS must still export the entropy its false model generates. KPI theatre is not “lying” alone; it is multiplicative denominator inflation: every decision made on false \(Q\) raises future \(\Gamma\) when reality arrives, and because \(\Gamma\) grows quadratically in the unresolved set (§1), the reckoning is sharp rather than gradual.

Moral injury as forced \(\mathcal{D}_{KL}\)

Moral injury occurs when an agent must act on institutional \(Q\) while their own model reflects witnessed \(P\) — selling a product they know fails, enforcing a policy that visibly harms members. The agent pays glucose and sleep to suppress surprise instead of updating the organisation’s model. Individual denominator stress from organisational-layer delusion divergence.

The \(\omega\) vs \(\eta\) trade-off

Specialisation raises \(\eta\); coordination raises \(\omega\). Late-stage organisations often add process without \(\Gamma\) reduction — meetings, dashboards, approvals — which is \(\omega\) without return. The classic mistake: cut \(\Phi\) (layoffs) to fund \(\omega\) (consultants) while leaving \(\mathcal{D}_{KL}\) untouched. The result is a denominator increase from two directions simultaneously.

4 Social capital and trustworthiness

Robert Putnam distinguished bonding social capital (dense ties within groups) from bridging capital (links across groups). In FPE terms, social capital is stored coordination capacity — a component of collective \(\Phi\) and a precondition for cheap coupling between nodes.

High social capital does not mean “everyone likes each other.” It means lower cost to run the two operations — honesty and repair are cheap because defection is visible and punished by the graph.

Trust stock depletes when: institutions preach fairness while members witness extraction (\(\mathcal{D}_{KL}\) exposed); grievances between subgroups harden into feud lines (\(\Gamma\) unresolved); precarious labour and commuting drain time for associations (\(\Phi\) exhausted); platforms capture bonding ties for profit without reciprocity (\(\Psi\) extractive).

Putnam’s civic decline is bridging \(\Phi\) depleted: households and firms face higher \(\mathcal{E}_\Sigma\) from isolation; politics becomes bonding-only (tribal) because cross-cutting ties atrophied.

Bourdieu’s capital forms (economic, cultural, social) are three distinct contributions to collective \(\Phi\) and individual \(\Psi\):

Bourdieu’s key insight — capital is convertible and field-specific — is that \(\Psi\) channels differ by field (academic, economic, social), and an agent must accumulate field-specific \(\Phi\) to receive shelter from that \(\Psi\). Misrecognition (what Bourdieu calls doxa) is a collective \(\mathcal{D}_{KL}\): the field presents its arbitrary rules as natural, lowering apparent \(\omega\) for those who internalised them and raising it for those who did not.

Goffman’s dramaturgical model (impression management, front/back regions, face-work) is the sociology of \(\mathcal{D}_{KL}\) management in real-time interaction. “Front stage” is the broadcast model \(Q\); “back stage” is the working model that may diverge. Face-work is the negotiation of whose \(\mathcal{D}_{KL}\) gets exposed and at whose \(\Gamma\) cost. Status rituals and interaction order reduce \(\mathcal{E}_\Sigma\) by compressing which \(Q\)-mismatches must be repaired live. When interaction order breaks down, \(\Gamma\) spikes and \(P_{in}\) is spent on social repair rather than productive work.

5 Norms, rituals, and solidarity

Norms are compressed models that keep pairwise \(\mathcal{D}_{KL}\) low enough for large-scale coordination — not morality bolted onto physics.

Rituals (weddings, funerals, stand-ups, elections, retros) are implementations of the two operations: 1. Synchronise models — “we agree what happened and what it means.” 2. Close or bound friction — “this conflict is settled / scheduled / contained.”

Failed rituals raise \(\omega\) without lowering \(\Gamma\): meetings that do not decide, holidays where fighting is forbidden but unresolved, elections that change faces without updating \(Q\). Empty ritual is denominator theatre — calm surface, rising hidden bill.

Mechanical solidarity (Durkheim — sameness as glue) lowers \(\mathcal{D}_{KL}\) cheaply inside homogeneous groups but can raise \(\Gamma\) at boundaries with outsiders. Organic solidarity (interdependence as glue) requires bridging — higher setup cost, higher \(\eta\) at scale. Both must be maintained through genuine repair, not performance.

Religion as IPS technology. Churches, mosques, parishes, and secular mutual-aid societies are historically \(\Psi\)-factories at \(L+1\)–\(L+2\): they produce recurring ritual, mutual aid, and shared \(Q\). Religion prevents anomie when it performs repair (confession with restitution, sabbath that bounds conflict, funeral rites that redistribute roles). It produces anomie when it becomes denominator theatre: more ritual and moral language, less actual repair, \(\Gamma\) exported to women, minorities, and youth via scapegoating or ostracism. Secularisation does not cause anomie; losing coordination capacity — the joules nobody pays to synchronise models and close friction — causes anomie. A secular society with strong bridging institutions can run lower \(\mathcal{D}_{KL}\) than a pious one with hollow ritual.

6 Anomie: when collective prediction fails

Durkheim used anomie for moments when society no longer supplies stable limits, legitimate means, or intelligible reward schedules. Merton sharpened the point: high cultural aspirations paired with blocked legitimate paths produce improvisation — fraud, status racing, withdrawal — that looks like moral failure but is predictive collapse.

Core mechanism: a shock (economic, technological, cultural) causes old norms to stop predicting outcomes. The official success story remains in place while \(\mathcal{D}_{KL}\) rises between promise and lived reality. Members must improvise — hustle, fraud, status racing, withdrawal — burning \(P_{in}\) on low-\(\eta\) competition. \(\Gamma\) rises; trustworthiness and solidarity fall; \(\Psi\) weakens; individual \(\mathcal{R}\) failures spread.

Why anomie can feel like freedom first. The opening phase of norm breakdown often feels like liberation from stale constraint — desire is less bounded, old authorities weaken. But if new coordination paths do not replace the old ones, this is exposure to raw \(\mathcal{E}_\Sigma\) with less shelter. Freedom without credible pathways becomes aspiration inflation, comparison without belonging, choice without guidance.

Durkheim’s suicide types are individual-level \(\mathcal{R}\) failures indexed by which fractal failure predominates:

A secondary episode (bankruptcy, divorce, humiliation) rarely explains a suicide rate — Durkheim’s point. Rates track structure. To change rates, one must change \(\Psi\), collective \(Q\), \(\eta\), and \(\omega\) at the enclosing nodes, not only stabilise mood at \(L\).

7 Scapegoating and entropy export

Groups often lower collective \(\Gamma\) by exporting entropy to a sub-node’s \(\Phi\) or \(\mathcal{D}_{KL}\): scapegoating, the identified patient, the fall guy, the moral panic. The scapegoat function is social and often unconscious — the group gets short-term cohesion; the target pays in health, reputation, or life chances.

The mechanism is the same NESS conservation law (3.3) that drives emotional labour inward and people-pleasing onto the self, now operating downward across levels. The group is a far-from-equilibrium system and must export the entropy its unresolved \(\Gamma\) and suppressed \(\mathcal{D}_{KL}\) generate. The two honest operations — updating the shared model, settling the friction — would implicate the role structure itself, so the group routes the excess dissipation onto a level-\(L-1\) sub-node instead: one member’s \(\Phi\) absorbs what the collective refuses to pay. The scapegoat’s \(\mathcal{B}\) depletes, their \(\mathcal{R}\) drifts below 1, and by Theorem 5.3 the collective’s \(\Phi\) is now partly composed of a failing sub-node — a hidden liability the index headline (“the family is fine”) does not show.

Short-term collective \(\mathcal{R}\) can rise by exporting, because the local \(\Gamma\) is discharged without the cost of operation 1 or 2. Long-term \(\mathcal{R}\) falls when: skilled \(\Phi\) exits (brain drain, estrangement); hidden \(\mathcal{D}_{KL}\) leaks (everyone knows the story is false, so the suppression tax compounds per Theorem 5.2); the target’s collapse percolates upward through the composition graph (Theorem 5.3 in reverse — a critical sub-node failing drags the whole product down).

Entropy export is stable when leaders lack \(P_{in}\eta\) for honest assertion; when operation 1 would implicate the whole role structure; when \(\Psi\) rewards performative accountability (firing one person instead of changing the process); when members prefer bonding against an outsider to bridging repair. It is an attractor in the same sense as the drama triangle (ips_psychology.md §5): each move that avoids the two real operations offers a local thermodynamic bribe, and the configuration is locally stable until \(\Gamma\) grows enough to force a phase event.

Fractal ethics: if your \(\Psi\) requires someone else’s \(\mathcal{R} < 1\), you are borrowing persistence. Defection, revolution, burnout, and lawsuits are the collection mechanism — the bill arrives when the sub-node exits, collapses, or litigates, and the collective’s denominator suddenly inherits everything it had exported.

8 Diagnostic questions (any group)

For any group that feels stuck or is failing, ask:

1. \(\Phi\): Are members intact — sleep, skills, trustworthiness, turnover rate?
2. \(P_{in}\eta\): Is energy available and spent on coordination, not theatre?
3. \(\mathcal{D}_{KL}\): Where does the official story diverge from what members actually witness?
4. \(\Gamma\): What is unresolved and how large is set \(C\)? Who is holding friction for the whole group?
5. \(\omega\): Which process complexity can die without killing the node?
6. \(\Psi\): Is external shelter real or performative? What does it actually attenuate?
7. Export: Is the group lowering local \(\Gamma\) by draining a sub-node’s \(\Phi\)?

9 What sociology is not in IPS terms

• Not determinism. \(\mathcal{R}\) is a clearing condition, not fate. Groups and institutions can improve terms.
• Not reductionism. Culture, meaning, and solidarity are not “just” \(\mathcal{R}\) — but they run on \(\mathcal{R} \ge 1\) the same way thought runs on glucose.
• Not a management textbook. These are thermodynamic identities, not templates for manipulation.
• Not ahistorical. Which paths count as “legitimate \(\eta\)” is decided by the people in a given time and place. IPS names the cost of blocking them; it does not dictate the values.

See also

• information_persisting_systems.md — formal derivation, Theorem 5.3 (fractal composition)
• fractal_layers.md — full stack from body through self to polity
• ips_psychology.md — agent-layer accounting
• ips_geopolitics.md — state-scale accounting
• information_containers.md — family and company containers in the full atlas
• glossary.md — canonical FPE vocabulary