What Cannot Be Undone, What Cannot be Completed
On the assumptions of reversibility and totalizability in the modern world view, and their role in the metacrisis.
The metacrisis names the condition in which ecological disruption, democratic erosion, technological acceleration, geopolitical instability, and psychological exhaustion have ceased to appear as independent challenges, and instead form a single, interwoven field. What weighs on this field is certainly its density, but perhaps far more its persistence, the growing sense that each intervention, however well-intentioned, fails to bring closure, and often deepens the entanglement it seeks to resolve. As argued for in previous essays, this can be seen as a deeper mismatch between how reality unfolds and how we as moderns have learned to relate to it. Beneath the surface diversity of the metacrisis lies two shared assumptions, both defining of the modern world view, that seem to have received scant examination in this context: first, that the world is in principle reversible, that damage can be repaired without remainder, and second, that the world is in principle totalizable, that reality can be rendered sufficiently complete in representation to be governed and controlled. This essay argues that a big contribution to why the metacrisis persists as a predicament, and not a problem, is because modern institutions manifests a world view grounded in reversibility and totalizability.
The counterpoints to these two assumptions are irreversibility and non-totalizability. These are often treated as practical inconveniences, limits imposed by insufficient knowledge, immature institutions, or technical constraints that will be overcome. But a different reading is possible, one that treats them as features of reality, and not as failures of our understanding and frameworks. This does not require that irreversibility and non-totalizability be absolute in every conceivable ontology, nor that no local reversals or closures ever occur. Rather, no reality within which action unfolds over time can make reversibility or totality its governing condition without erasing the very features that make action meaningful.
In an effort to see why this is so, consider why reversibility cannot serve as a governing assumption for responsibility and meaning. Responsibility presupposes that actions bind futures in ways that cannot be annulled, and without this binding, agency flattens into action with no real stakes. When reversibility becomes structural, commitment loses its binding force and responsibility migrates away from the agent toward the faculties that promise correction. Meaning arises precisely where action proceeds without the guarantee of reversal, where consequences must be carried rather than reset, and where action closes off alternatives rather than preserving them indefinitely. Reversibility and meaning cannot function as the background condition upon which action relies in advance, thus irreversibility is a precondition for action to count as action at all. Responsibility and meaning also rely on non-totalizability, on the fact that action unfolds in a reality that cannot be exhaustively represented in advance. If all relevant consequences could be fully anticipated, calculated, and contained within a complete model, action would reduce to execution, and responsibility would shift from the agent to the model system. Meaning arises precisely because action must proceed without full knowledge of what it brings about, and responsibility persists because no representation can absorb or preempt the consequences of acting. Thus, irreversibility and non-totalizability are the conditions of possibility for action, meaning, and responsibility: if every act could be rolled back without cost, nothing would fully matter.
Irreversibility names the fact that time does not just pass, it commits. Events do not simply occur, they bind the future to a past that cannot be undone without becoming something else entirely. This intuition finds one of its earliest and most rigorous philosophical expressions in Bergson’s (1911) account of duration. Bergson critiqued the conception of time as a sequence of interchangeable instants that could in principle be replayed in reverse. To him, time as lived reality is a qualitative becoming in which the past is conserved and carried forward into the present. Duration is irreversible not because we lack the means to rewind it, but because it is creative, each moment adds something that did not exist before. Time is the internal tension by which reality differentiates itself, rather than a container within which change occurs. As such the present is never free of the past, nor the future fully separable from what has already taken place. Once time is understood in this way, irreversibility ceases to be an unfortunate limitation and becomes the very medium through which novelty, experience, and meaning arise.
The political and social implications of this become explicit in Arendt’s (1958) analysis of action. Arendt distinguishes between fabrication, where an agent remains in control from beginning to end, and action, which unfolds among others and produces consequences that escape the agent’s intention. Action is irreversible and unpredictable, and for that very reason it requires stabilizing responses, like trust, promises and agreements. These responses are practical recognitions of a world in which deeds cannot be undone. Modern society, however, increasingly organizes itself around activities that simulate reversibility: making, managing, optimizing, iterating, and redesigning, practices whose legitimacy depends on the assumption that errors can be corrected without residue and that consequences remain local and revisable. But society is no longer local, and the drive for reversible actions now devalue the forms of action that generate meaning through exposure to consequence.
Non-totalizability sharpens this picture by challenging a second, equally entrenched assumption: that reality can in principle be fully represented, mapped, or closed into a complete system. The modern imagination is deeply invested in the hope that with sufficient information, computation, or formal rigor, the world could be rendered fully legible: through perfect representations of reality and unifying theories of everything. This hope has been dashed at the level of formal reasoning: Gödel’s (1931) incompleteness theorems demonstrate that any sufficiently expressive formal system cannot be both complete and consistent, as there will always be truths that cannot be proven within the system itself. This might seem a temporary failure awaiting better axioms, but is a structural limit. Closure generates its own outside.
Gödel’s result sits in continuity with the earlier discovery of Cantor [1915] (1955) that infinity cannot be domesticated into a final totality. No matter how far enumeration proceeds, a larger infinity can always be constructed, and the continuum resists completion. Read philosophically, Cantor’s work suggests that non-totalizability is a trait of what is counted, rather than an accident of method, and that reality does not present itself as a finished whole awaiting capture, but as an inexhaustible field whose very being exceeds any act of summation. Non-totalizability is not a defect of our descriptions, it is a feature of what is described. Unsurprisingly, we find similar pointers to the limits of the epistemic from quantum theory and its uncertainty principle: not only are there physical limits to what can be known about a system, but to acquire knowledge we must interact and any interaction changes what we sought knowledge about. The relevance of these results is not that reality is a formal system, but that every attempt to render it complete must pass through form, the epistemic, and it is there that closure fails.
Once irreversibility and non-totalizability are treated as ontological rather than epistemic, the modern world view comes into focus as a systematic misalignment. The main problem is not that modernity lacks insight into these aspects, but that these assumptions suffuse the background of its workings, and that these insights rarely shape the design logic of its dominant institutions, infrastructures, and technologies. The assumptions of reversibility and totalizability operate as selection pressures, shaping which designs scale, which interventions are rewarded, and which failures are interpreted as reasons for revision rather than as signals of mismatch. This world view largely treats the world as something that ought to be predictable and engineerable. This diagnosis aligns with Rosa’s (2020) account of modernity as a project of making the world fully available and controllable, a project that systematically generates uncontrollability precisely because it refuses to accept limits as constitutive rather than accidental. Problems are approached as optimization challenges, uncertainty as a temporary lack of data, and unintended consequences as correctable errors. Furthermore, systems organized around reversibility and closure are materially advantaged, because they align with administrative control and decision-making. This orientation has produced extraordinary capacities, but it rests on a silent wager, that reality will ultimately submit to closure and control.
The Nobel Prize laureate Prigogine challenged this wager from within the sciences themselves. Against the classical image of nature as governed by reversible laws, Prigogine argued that irreversibility is less an anomaly to be explained away than a fundamental feature of physical reality as it is realized through time, especially in far-from-equilibrium systems where order emerges through dissipation and where history matters (Prigogine & Stengers 1984). In such systems, time both measures change and actively participates in it, in that the path by which a system arrives at a state becomes inseparable from what that state is. It could be objected that irreversibility cannot be fundamental, since the basic equations of physics are largely time-reversal invariant, and because entropy increase is statistical rather than absolute. From this, irreversibility appears as an artifact of description, coarse-graining, or incomplete knowledge, something that would dissolve under perfect control or infinite precision. But this objection mistakes formal reversibility for physical reversibility. That equations permit reversal does not mean that processes are reversible within the world the equations attempt to describe, except in extreme, localized scenarios. Reversibility is the exception, not the rule. The conditions required to reverse a real process, exact state reconstruction, perfect isolation, and complete coordination across interacting degrees of freedom, are themselves physically unrealizable and would constitute new, irreversible interventions.
What physics shows is that irreversibility does not arise from a single asymmetric law. It arises instead from interaction, instability, and the accumulation of history, from the fact that systems are embedded in a world that records what has happened and does not offer a standpoint outside itself from which time could be undone. It arises from the whole. Formal symmetry does not negate irreversibility, it presupposes it: symmetry is discovered from within a world where time has already done its work, rather than discovered from outside time. Nature is as such a process that produces novelty irreversibly, and explanation does not eliminate time, it must reckon with it.
The metacrisis can be understood as the point where this ontological reality collides with social and technological systems built on the opposite assumption. Climate systems do not respond smoothly to intervention, they cross thresholds. Social trust does not behave like capital, it cannot be stored and reinvested without decay. Information abundance does not automatically produce shared understanding, it often fragments it further. Institutional complexity does not yield proportional control, it breeds brittleness and diminishing returns. Each attempt to impose reversible solutions on irreversible processes and totalizing frameworks on non-totalizable realities intensifies the very dynamics it seeks to manage. As such, much of what we experience as crisis arises from the persistent refusal to embrace the reality we inhabit, and accepting what we can’t or shouldn’t control.
Technology, and especially artificial intelligence, amplifies this tension by collapsing representation, decision, and execution into a single accelerating loop. This collapse alters institutional incentives, favoring systems that act fastest on their own representations, even when those representations are necessarily partial and lagging. AI systems excel at producing legibility, translating complex, qualitative aspects of the world into structured data and actionable outputs. They reinforce the belief that with enough computation, uncertainty can be dissolved into optimization. Yet their deployment is itself irreversible, as infrastructure is built, skills atrophy, incentives harden, and power concentrates in ways that cannot simply be rolled back if the experiment disappoints. The digital, despite its appearance, rearranges material resources, energy flows, incentive and cognitive landscapes, and institutional dependencies in lasting ways.
AI also intensifies the drive toward totalization. By presenting the world as something that can be exhaustively represented and anticipated, it invites institutions to treat models as substitutes for reality rather than as partial instruments within it. Decisions begin to occur in the space of representations, while the world they represent becomes a delayed or secondary signal, often registering only when thresholds are crossed or failures cascade. This dynamic did not begin with AI, but AI accelerates it by operationalizing representation at a speed and scale that leaves little room for corrective friction. AI subtly reshapes what counts as relevant, actionable and real, narrowing attention to what fits the model while expanding confidence in the adequacy of the picture produced.
As AI systems increasingly train on data generated by other AI systems, researchers have observed forms of degeneration, loss of variation, and compounding distortion, a phenomenon often described as model collapse (Shumailov et.al. 2024). What appears as a technical problem reflects a deeper dynamic: when representation feeds on itself, the attempt to totalize reality into a model begins to erode the very ground it depends on. What Gödel and Cantor showed formally, the phenomenon of model collapse enacts operationally, that any system which treats its own outputs as exhaustive of reality will, over time, forfeit contact with what exceeds it. Furthermore, recent work on epistemological fault lines between human and artificial intelligence shows that apparent alignment at the level of outputs conceals a deeper structural divergence in how judgments are produced: LLMs generate fluent conclusions without grounding, causal modeling, metacognition, or value-sensitive commitment, creating what the authors call Epistemia, “the experience of knowing without the labor of judgment” (Quattrociocchi et.al. 2025). Closure undermines contact with the novelty of reality, and the appearance of understanding can coexist without that which makes understanding accountable to the world. There is little reason to think that this structure will change within the current paradigm of AI development, because the paradigm itself is defined by the pursuit of closure and control.
The result is a peculiar asymmetry. AI promises reversibility at the level of decision and control, while entrenching irreversibility at the level of structure and consequence. It promises total vision, while deepening dependence on representations that can never be complete. This asymmetry follows directly from treating non-totalizable reality as if it were amenable to closure, and irreversible processes as if they could be managed through iterative correction. AI becomes less a neutral tool than an amplifier, bringing into sharp relief the ontological assumptions that already govern our responses to the metacrisis.
This is where Jonas’s (1984) warning becomes especially relevant. Jonas argued that modern technology introduces a new ethical condition, as action now operates on timescales and magnitudes where consequences may be irreversible and extend far beyond the horizon of prediction. What is decisive in his argument is that technological power alters the ontological situation of action itself, by binding present decisions to futures that cannot answer back, and by making responsibility asymmetric in time. Traditional moral frameworks, developed in contexts where harm was local and recoverable, become inadequate when technological power threatens the conditions of life itself. With asymmetry, Jonas’s point comes not to be only about ethics, but it is also metaphysical. Irreversibility changes what responsibility means. One implication is that ethical adequacy can no longer be measured primarily by efficiency, optimization, or expected benefit, but by exposure to consequence. Where actions bind futures that cannot respond or repair themselves, responsibility shifts from achieving outcomes to preserving conditions of possibility. This privileges caution, restraint, asymmetry of burden toward the powerful, and the selective refusal of interventions whose primary justification rests on their purported reversibility. As such, the claim made earlier that responsibility presupposes irreversibility acquires its full force under technological conditions.
Thus, the metacrisis can be seen as a failure of orientation through assumptions deep in the world view of modernity. This diagnosis does not exhaust the metacrisis, nor does it claim explanatory priority over economic, political, or historical accounts. It is one lens among others, albeit one that cuts across domains rather than residing within any single one. The expectation that this diagnosis should yield a general solution or program is itself part of the misalignment being described, importing a demand for closure into a reality that resists it. The persistent desire for total explanations and reversible interventions is understandable, especially under pressure, but it also functions as a metaphysical anesthetic, dulling our sensitivity to consequence and commitment. When a world view built on reversibility and totalizability confronts an irreversible and non-totalizable reality, the resulting failures are rarely taken as signals to revise the world view itself. Instead, they are read as evidence that we need more control, better models, tighter integration, and faster intervention, which in turn hardens the very designs and systems that produced the failure in the first place. Each iteration strengthens the underlying assumption, narrows the range of acceptable responses, and deepens dependence on the very infrastructures that render reversal and revision increasingly difficult. This mismatch generates a constant sense of urgency combined with a chronic inability to resolve what we confront. Each intervention produces further consequences, each model both hides and reveals new blind spots, and the horizon of control recedes rather than approaches. This dissonance generates feedback loops in which the world view, the designs and incentive landscapes springing from it, and the traps they create mutually reinforce one another, contributing to and reinforcing wicked loops. Left unexamined, these loops deepen, as each failure justifies further entrenchment of the assumptions that made failure inevitable.
This implies the need for a different understanding of action itself. Optimization assumes reversibility, rule systems assume closure, and prediction assumes that what matters can be fully represented in advance. In a world structured by irreversibility and non-totalizability, judgment and wisdom are required because every alternative mode of action presupposes what is not available. Judgment denotes a situated capacity exercised under conditions of irreversibility, where criteria cannot be fully formalized in advance and consequences cannot be cleanly separated from their contexts. Wisdom is not a trait that scales through abstraction, but one that scales through cultivation and restraint: by limiting domains of intervention, slowing feedback cycles, and maintaining spaces where action remains answerable to lived consequence rather than to representation alone. Such capacities resist institutionalization precisely because they cannot be reduced to rules without distorting what they are.
The metacrisis is not only out there in systems and institutions. It lives in a shared difficulty of accepting that we act within an unfinished world larger than us and our control, one that cannot be mastered without remainder, and that whatever truly matters will matter precisely because it cannot be undone.
The universe is no narrow thing and the order within it is not constrained by any latitude in its conception to repeat what exists in one part in any other part. Even in this world more things exist without our knowledge than with it and the order in creation which you see is that which you have put there, like a string in a maze, so that you shall not lose your way. For existence has its own order and that no man's mind can compass, that mind itself being but a fact among others.
Cormac McCarthy - Blood Meridian
References
Arendt, H. (1958). The Human Condition. University of Chicago Press.
Bergson, H. (1911). Creative Evolution. Henry Holt and Company.
Cantor, G. [1915] (1955). Contributions to the Founding of the Theory of Transfinite Numbers. Dover Publications.
Gödel, K. (1931). On Formally Undecidable Propositions of Principia Mathematica and Related Systems I. Monatshefte für Mathematik und Physik.
Jonas, H. (1984). The Imperative of Responsibility. University of Chicago Press.
McCarthy, C. (1985). Blood Meridian or the Evening Redness in the West. Random House.
Prigogine, I. & Stengers, I. (1984). Order Out of Chaos. Bantam Books.
Quattrociocchi, W., Capraro, V. & Perc, M. (2025). Epistemological Fault Lines Between Human and Artificial Intelligence. https://arxiv.org/abs/2512.19466
Rosa, H. (2020). The Uncontrollability of the World. Polity Press.
Shumailov, I. et al. (2024). AI models collapse when trained on recursively generated data. Nature 631, 755–759 (2024). https://doi.org/10.1038/s41586-024-07566-y