What is the anthropic loop?
The anthropic loop follows, through five conditional states, what becomes of a displaced cost: displacement, accumulation, saturation, return, bifurcation. It extends K. William Kapp’s cost-shifting and is distinct from Bernard Stiegler’s neganthropy — and from the AI company Anthropic.
Page updated
The author’s analytical framework
On this page · 8 sections
The anthropic loop is formalised by economist Stéphane Lalut in the working paper AWP-07 (2026), in explicit dialogue with K. William Kapp’s cost-shifting, Nicholas Georgescu-Roegen’s bioeconomics and Bernard Stiegler’s neganthropy.
The five moments of the loop
A social system — an institution, an economy, a territory — produces its order by exporting disorder. The loop follows that disorder through five states. None is compulsory: the load may never accumulate, the receptacle may not saturate, and no return may occur within the period studied.
- E1 — Displacement: disorder is exported toward a receptacle — a spatial, temporal or social elsewhere; the three axes can combine.
- E2 — Accumulation: the load accumulates if inflows exceed what the receptacle absorbs or regenerates; otherwise it can stabilise or decline.
- E3 — Saturation: the receptacle reaches a capacity limit, or contestation imposes an effective constraint on the continuation of the transfer.
- E4 — Return: disorder may come back toward the emitting system — rarely in the form in which it was exported.
- E5 — Bifurcation: the system re-displaces toward a new receptacle, reconfigures itself, or partly collapses.
Four readings of the anthropic loop
Each reading asks the same four questions: what is observed? what proves a transfer? which causal links are established? what trajectory can be drawn? The answers differ, and so does the shape. Each reading shows what is observed and what can be explained; that these loops form an overall dynamic remains a research hypothesis. Dashes: conditional transitions, as in the diagram · filled disc, thick line: documented · “?”: hypothesis, not established · grey dotted: not assessed.
Rome — a reconstructed historical trajectory
An illustrative reading, after Anthropie (2025) and AWP-02: at the scale of a civilisation the loop cannot be tested, for want of a bounded receptacle and of a latency named in advance.
In the reading Anthropie offers, Rome buys order at its centre by having the price paid elsewhere: the Pax Romana is financed by the taxes and tribute of the provinces, African grain, the labour of slaves and twenty years of soldiers’ service. The book describes this burden as an accumulating debt, that of an empire ever harder to govern, until Rome exhausts the reserves that fed the transfer, “when there was no longer enough elsewhere to burn”. Disorder then returns to the centre in other forms: a debased coinage in the third century, then a war entrusted to federates, one of whom, Alaric, sacks Rome in 410. Neither this reliance on federates nor Constantine’s Christian turn prevents the fall of the West in 476; the East holds until 1453. Saturation and return stay marked with a “?”: this is the book’s thesis, and other historians give the leading role to the invasions. The fall of the West is a fact; that it was the outcome of this loop remains an interpretation.
Public debt — open accumulation
An open loop, after AWP-09: the public accounts document the first two states, in a balance-sheet sense; the final net burden remains unestablished.
From 1995 to 2025, French general government debt rose from 57.8% to 115.6% of GDP, mainly through deficits excluding interest, and less than one fifth of the deficits matched a net increase in public assets. Later generations therefore inherit debt without equivalent wealth on the other side: it is in this narrow, balance-sheet sense that the accounts establish a shift over time. The debt-to-GDP ratio does not by itself measure the burden passed on, and it does not rise in one direction only: since 2017 growth has reduced it every year except 2020, without reducing the debt in euros. The rise in interest since 2022 can be read as the start of a return; but refinancing the debt at higher rates is enough to explain it too, and nothing establishes that the state’s capacity to carry its debt is saturated. Who finally pays will depend on the adjustments chosen.
Over time, and up close. From 1995 to 2025, the ratio rose by 57.9 points of GDP, despite 9 years of decline out of 30. Up close, interest (+73.9) and growth (−72.6) almost cancel out; the rise comes mainly from deficits excluding interest (+50.5). The force that shapes the trajectory is not the one seen oscillating. This accounting decomposition describes without attributing a cause: the year-by-year detail.
Environment — coexisting regimes
After AWP-04, reread by AWP-07: the new displacement is documented, what follows is not.
The fossil regime is not a past turn: it carries on, and the transition coils around it, since each energy source has been added to the previous ones instead of replacing them, as Jean-Baptiste Fressoz shows and AWP-04 follows. The figure therefore draws two active strands, and the turn in full is the transition. Its displacement is documented, though not quantified in net terms: towards the mining territories of lithium and cobalt, and towards the future, through panels, batteries and nuclear waste that will still need handling decades from now. It shows in the accounts: France has cut its territorial emissions far more than its carbon footprint, which includes imported goods. The strain on the Atacama aquifers or waste with no mature recycling route are signals, not named thresholds: hypotheses, marked with a “?”, and no return is established. The rebound effect, finally, does not close the circle: the consumption that an efficiency gain induces is a new load, which feeds the turn under way.
Digital — displacement and lock-in
After AWP-06: an open loop, established on its first two states.
Digital services look immaterial because their matter has been moved out of the user’s field of view: the service, AWP-06 writes, “has not lightened matter, it has learned to make it work elsewhere”. The displacement runs along all three axes at once: towards the territories that host data centres, towards the future through grid and land commitments that run for several decades, towards populations, from mines to end-of-life sites, who neither produce nor benefit from it. The load builds up through one-off decisions, each reasonable on the spot, whose sum narrows the territory’s room for manoeuvre; optimising one register, electricity for instance, can shift the constraint to water, land or matter. Beyond that, nothing is established: neither saturation nor a return to the operators, and whether to maintain, redirect, close or dismantle these infrastructures remains an open decision. The cognitive load shifted onto users is not coded here: AWP-06 treats it as a receptacle, not yet as a constituted debt.
The diagram also carries three paths outside the sequence. Transfer arrest leads straight from accumulation to bifurcation: the emitter stops displacing before any saturation, when the transfer costs it more than it brings in. The early return toward the emitter S occurs during displacement or accumulation — leaks, immediate backlash — without changing the state of the loop. Finally, at every passage, counter-tendencies — regeneration, repayment, technical substitution, a political reaction — can delay, deflect or prevent what follows; and the load can be internalised: the system deals with it at home, and the loop stops.
One full turn: asbestos
The asbestos industry produced its order by making others bear a risk (E1): the bodies of workers and local residents, on the social axis, and the building stock. The load accumulated out of sight (E2): the latency of mesothelioma is “often of the order of 30 to 40 years”, according to the French Senate’s 2005 fact-finding mission. Saturation did not come from a lack of room but from contestation (E3): lawsuits, the press and independent expertise wore down the doubt the industry had kept alive. The return took a different form from the one that left (E4): as spending on compensation and asbestos removal. It reached the emitting firms only in part: its burden falls mostly on all the employers of the general social-security scheme, including those that had never used asbestos. The outcome (E5) was forced internalisation, through the national ban of 1997: this turn opened no other.
When turns follow one another
For one turn to open another, the outcome must be a new displacement: the emitter has changed, and the cycle starts again from another configuration. That is when the loop becomes a spiral. AWP-04 reads the energy transition this way. The fossil regime shifted part of its costs into the atmosphere, and so into the future; the response to its saturation opens a new circuit, spatial and material — lithium from Chile, cobalt from the Democratic Republic of the Congo, far from the centres of consumption — to which the rebound effect adds a new load. Yet the “Environment” button on the figure shows that this new turn does not replace the old one: the fossil regime remains an active strand there, with the transition coiling around it, because energy sources add up more than they substitute for one another. Each re-displacement wears down the available elsewheres a little more, unless they replenish fast enough: a turn is therefore not necessarily worse than the one before. That these turns nest, the saturation of one level spreading to the level above, from territory to world-system, remains a research hypothesis: a drift observed over time, at a single level, does not demonstrate it.
Kapp: cost-shifting is systemic
The first lineage is economic. As early as 1950, in The Social Costs of Private Enterprise, K. William Kapp established that shifting costs onto third parties is not an accidental market failure but a systemic practice: private enterprise is structurally incentivised to move part of its costs onto workers, neighbours, the environment, the future. The anthropic loop extends Kapp on two points: it generalises the mechanism to any order-producing system — not only the firm — and it follows what becomes of it: the displaced cost does not vanish into a memoryless elsewhere, and the loop asks under what conditions it accumulates, saturates its receptacle and returns.
Georgescu-Roegen: there is no infinite elsewhere
The second lineage comes from bioeconomics. In The Entropy Law and the Economic Process (1971), Nicholas Georgescu-Roegen established that no economic process cancels disorder: every production of local order is paid for by degradation somewhere. The anthropic loop keeps this result as background — the guarantee that a “displacement without remainder” does not exist — without importing thermodynamics: its receptacles are not energy reservoirs but social, territorial and attentional spaces, and its saturation is read in institutions, not in joules. The lesson retained is that of the closed system: there is no infinite elsewhere.
Stiegler: an acknowledged homonymy
The third lineage is also a demarcation. Bernard Stiegler used the word “anthropy” for human-generated entropy — the share of disorder that human activities inject into living and technical systems — and opposed to it neganthropy: the forms of knowledge, care and bifurcation capable of producing local negentropy. That is a grammar of quantities: produce less anthropy, more neganthropy.
Anthropy in Lalut’s sense belongs to another grammar: it does not measure a production, it traces a route. The question is not “how much disorder do we produce?” but “where does disorder go, who absorbs it, and through which mediations does its route become invisible?”. AWP-07 fixes this acknowledged homonymy — same word, two concepts — and, in passing, returns the credit of the “boomerang effect” to Ulrich Beck, while demarcating the loop from its conceptual neighbours: David Harvey’s spatio-temporal fix, Stephan Lessenich’s externalisation society, system dynamics.
The apparatus’s own contribution: invisibility as a cost, return as transmutation
Two pieces of the apparatus carry its added value. The first is condition C3: the invisibility of a transfer is not a state of affairs but a maintenance cost — keeping a displacement out of sight requires spending, in devices and narratives — and this cost is observable before the return. The second is the transmutation of the return: disorder rarely comes back in the form in which it was exported — an environmental cost displaced returns as a health crisis, a social cost as political instability. This is precisely why returns are so rarely attributed to their original displacement: between export and return, the form has changed.
A formalisation built to be refuted
The formalisation — nine definitions, five states, three conditions, four forms of return, six propositions with differentiated status — has value only if it increases refutability. AWP-07 sets an ex ante protocol: name the receptacle and the order of magnitude of the latency before the analysis, so that one can neither assert an untraceable displacement nor certify a complacent internalisation. One case external to the corpus — asbestos — is coded in full to show how the grid is used, not to validate it.
Frequently asked questions
What is the anthropic loop?
The anthropic loop follows, through five conditional states, what becomes of the disorder a social system makes others bear: displacement toward a receptacle (E1), accumulation (E2), saturation (E3), a possible return toward the emitting system (E4), bifurcation (E5) — new displacement, reconfiguration or partial collapse. No passage is compulsory. It is formalised by economist Stéphane Lalut in AWP-07 (2026, DOI 10.5281/zenodo.21200286). When an iteration ends in a new displacement, the path does not return to its starting point: it is a spiral.
Is the anthropic loop related to Anthropic, the AI company?
No. Anthropic is an American artificial-intelligence company, and its engineering vocabulary (agent loops, 'loop engineering’) is unrelated. The anthropic loop is a social-science concept: the cycle through which social systems displace disorder, saturate their receptacles and face its return. Only the Greek root ánthrōpos is shared.
What is the link between the anthropic loop and K. William Kapp’s cost-shifting?
An acknowledged lineage. Kapp showed as early as 1950 (The Social Costs of Private Enterprise) that shifting costs onto third parties is systemic, not accidental. The anthropic loop extends cost-shifting by generalising it to any order-producing system and by asking under what conditions the displaced cost accumulates, saturates its receptacle and comes back — rarely in the form in which it was exported.
Is Stéphane Lalut’s anthropy the same as Bernard Stiegler’s anthropy?
No — the word is shared, the grammar differs, and AWP-07 fixes this acknowledged homonymy. In Stiegler, anthropy names a quantity (human-generated entropy, to be fought by neganthropy). In Lalut, it names a mechanism of circulation: the displacement of disorder by social systems — where it goes, who absorbs it, through which mediations its route becomes invisible. The first usage measures a production; the second traces a route.
Why engage with Georgescu-Roegen if anthropy is not thermodynamic?
Nicholas Georgescu-Roegen (The Entropy Law and the Economic Process, 1971), founder of bioeconomics, guarantees that no economic process cancels disorder. The anthropic loop takes this result as background, not foundation: it imports no thermodynamics — its receptacles are social, territorial and attentional. The lesson retained is that of the closed system: there is no infinite elsewhere.
Is the anthropic loop falsifiable?
That is the requirement AWP-07 sets for itself: an ex ante protocol requires naming the receptacle and the order of magnitude of the latency before the analysis, which disciplines both the claim that a displacement exists and the certification that a cost has been internalised. One case external to the corpus — asbestos — is coded in full to show how the grid is used, not to validate it.
Is the anthropic loop a metaphor?
No: it is a conceptual state machine, with definitions, transition conditions and testable propositions. The short definition is in the glossary.
Where can the loop be observed?
Wherever a local order rests on a saturating receptacle: public debt (a temporal transfer that can return as interest burden and budget constraints — AWP-03), the energy transition (externalised extraction, deferred waste, the rebound effect — AWP-04), digital infrastructures and attention (AWP-06).
How does it relate to social reversibility?
AWP-08 provides the loop’s micro-institutional counterpart: the stratified price of failure distributes unequally the capacity to defer, mutualise or absorb the cost of a failure — displacement of disorder seen from the people who absorb it. See What is social reversibility?
Further reading
- AWP-07 — The Anthropic Loop — the working paper that formalises the apparatus (DOI: 10.5281/zenodo.21200288, open-access PDF).
- What is anthropy? — the general concept and its distinctions (entropy, Anthropocene, anthropic principle, anthropology, Anthropic).
- The book ANTHROPY – A Big History of Civilization’s Hidden Costs (2026, 644 p.) — the complete framework, of which the loop is the theoretical keystone.