This is an extended version of a keynote lecture delivered on 9 May 2026 at the conference "Rejecting the Future: Affect and Mental Health" at the Hanse-Wissenschaftskolleg. The lecture develops Living Value Theory (LVT) as an account of why modern predictive institutions systematically fail in certain domains while succeeding brilliantly in others. The key concept is mixed futurity: the ontological condition in which non-recursive, self-recursive, and inter-recursive processes unfold simultaneously within the same mesocosmic situation. Modern institutions, such as laboratories, clinical trials, bureaucracies, audit regimes, algorithmic platforms, are predictability machines that achieve projectable stability by suppressing recursive dynamics and amplifying non-recursive regularities. The lecture calls this ontological surgery. Such surgery succeeds where the domain is primarily non-recursive but produces systematic failure when applied to constitutively inter-recursive domains. Antidepressant clinical trials are the paradigmatic case: they are sophisticated recursive suppression apparatuses that fail not because they are poorly designed but because mental illness is an irreducibly recursive domain that resists non-recursive treatment. The lecture concludes by calling for ontological literacy: the capacity to identify the recursive structure of the domain one is working in and to calibrate cognitive and institutional strategies accordingly.
1. The Problem of Systematic Failure
In Ivan Drago's Soviet sports laboratory, everything is measured. Biometric sensors record every physiological variable. Scientists track oxygen uptake, punch force, and recovery time. Pharmacologists optimise the body's chemical performance. A prediction is produced: the fight can be calculated. Rocky Balboa trains in a Siberian barn. He chops wood, runs through snow, and climbs mountains in darkness. No sensors. No outputs. No projections. Rocky IV knows which preparation survives contact with reality. What the film cannot say is why.
This lecture provides an answer. It does so not as an exercise in reading a 1985 Cold War film but because the two training regimes exemplify a structural contrast that runs through modern knowledge systems. Modern institutions are extraordinarily good at prediction in some domains and extraordinarily bad at it in others, and the failures are not random. Financial crises, pandemic mismanagement, the persistent gap between antidepressant efficacy in trials and effectiveness in clinical practice, the chronic underperformance of social policy interventions, the radicalisation effects of algorithmic recommendation systems, these are not accidents. They follow a pattern, and the pattern demands a structural account.
Several existing accounts notice parts of the problem. James Scott (1998) argued that state legibility destroys the local practical knowledge, metis, that sustains actual coordination. Nassim Taleb (2007) identified the systematic underweighting of tail risks by Gaussian models. Philip Tetlock (2005) demonstrated that expert forecasters barely outperform chance, and that strong theoretical frameworks actively impede performance. Ian Hacking (1995) showed that human classifications loop back onto the people classified, making the objects of social science unstable in ways that the objects of physics are not. Each of these contributions is important. None supplies a structural account of why the failures cluster where they do, why prediction works in structural engineering and fails in psychiatry, why pharmacokinetics is tractable and social trust is not.
The argument of this lecture is that what is missing from these accounts is an adequate ontology, a theory of what kinds of future-structure actually exist in the world and how different domains are constituted by different mixes of these structures. Living Value Theory (LVT) provides this ontology through the concept of mixed futurity: the idea that the mesocosm, the lived field of coordination in which human life takes place, is characterised by the simultaneous presence of non-recursive, self-recursive, and inter-recursive processes, each with radically different predictability profiles. Modern predictive institutions succeed, when they do, by suppressing recursive dynamics and amplifying non-recursive regularities, a procedure this lecture calls ontological surgery. The surgery succeeds where the domain tolerates it; it produces systematic failure where the domain's most important features are precisely the recursive dynamics that have been suppressed.
Antidepressant clinical trials are the paradigmatic case developed here. They represent the most technically sophisticated attempt yet to extract predictive stability from a constitutively recursive domain. The trial design is not naive; it is an extraordinarily refined recursive suppression apparatus. But the domain it is trying to suppress is one in which recursive dynamics are not incidental features to be eliminated but constitutive properties of the phenomenon itself. The result is a structural paradox at the heart of psychiatric research: trials succeed scientifically to the extent that they reduce the very recursive complexity that makes psychiatric suffering what it is.
The lecture proceeds as follows. Sections 2 and 3 introduce LVT's foundational concepts: the mesocosm, the five mediations, the levels of recursivity, and the principle of recursive fluidity. Section 4 develops the account of recursion domains and the concept of mixed futurity. Section 5 analyses modern institutions as predictability machines performing ontological surgery. Sections 6 and 7 develop the antidepressant trial case. Section 8 proposes ontological literacy as the minimal response to the systematic failures the lecture has diagnosed.
2. The Mesocosm and the Five Mediations
Living Value Theory begins with the concept of the mesocosm: the lived field in which coordination actually takes place. The mesocosm is not a scale between micro and macro analysis, nor a level of society between individual and structure. It is the level at which reality becomes available to living beings through mediation, the field in which anything that enters into coordination with a living being becomes part of that being's world. Black holes and quantum fields are not part of most human mesocosms, but they can be, for astrophysicists who bring them into their practical and symbolic coordination. Conversely, things that exist without being mediated into coordination do not enter the mesocosm at all. Entry is through mediation, not through presence.
The mesocosm is not defined by identity, territory, ownership, or group membership. It is defined by ongoing processes of coordination. This makes it irreducible to the organism-environment binary that has structured so much of biological and social thought. LVT does not situate a self-contained organism against an external environment; it begins from the coordination processes through which living beings and their worlds are co-constituted. Jakob von Uexküll's (2010 [1934]) concept of the umwelt partially anticipates this move, but LVT departs from it precisely at the point where Uexküll retains the organism-environment distinction. LVT begins from coordination, not from the organism that coordinates.
Coordination in the mesocosm occurs through five irreducible mediations. These are not analytical categories imposed from outside but the actual modes through which coordination takes place; none can be derived from another, and they are typically co-present in any given situation.
Embodiment names the lived body's ongoing coordination: sensation, proprioception, interoception, metabolism, fatigue, pain, arousal, illness, balance, rhythm, and skill. Embodiment is not an object among others; it is the condition of access to all other mediations. Being-with names coordination with other living beings: co-presence, attunement, dependency, care, threat, imitation, rivalry, intimacy, authority, shame, hierarchy, and shared attention. Social life does not begin with explicit norms or symbolic agreements; it begins with coordinated presence. Dwelling names the spatial and temporal conditions of inhabitation: landscape, distance, altitude, climate, gravity, seasons, light cycles, ecological patterns, and planetary constraints. Dwelling refers to non-human environmental conditions, not to houses, roads, or built infrastructure, which belong to the fourth mediation. Multimateriality names coordination through materials and artefacts: tools, surfaces, substances, technologies, infrastructures, animals, plants, clothing, food, medicines, machines, documents, devices, and built environments. Multisymbolization names coordination through symbols: language, number, image, diagram, code, narrative, classification, law, theory, money, ritual formula, label, sign, and model.
Multisymbolization is the last of the mediations to evolve and the most distinctive feature of human mesocosmic life. Other living beings coordinate through embodiment, being-with, dwelling, and to a limited extent multimateriality. Humans have massively expanded multimateriality and, most consequentially, invented multisymbolization. Numbers, diagnostic categories, digital platforms, and financial models are all forms of multisymbolization. The capacity to symbolise, to maintain relations to what is absent, to re-present the world, to coordinate action through shared symbolic systems, is generative and extraordinary. It is also the source of one of the fundamental error patterns in modern knowledge: the mistake of treating the symbol for the coordination it represents.
LVT defines value as what sustains, repairs, intensifies, or improves coordination in the mesocosm. This has a counterintuitive consequence: the most valuable coordinations are often the least visible. When coordination succeeds fully, it withdraws into the background, it becomes the smooth and effortless taken-for-granted world within which everything else takes place. This principle of invisible value runs directly against the institutional logic of modern knowledge systems, which consistently overvalue what can be named, measured, displayed, and audited, and systematically undervalue what works too well to demand attention. The gap between what is symbolically visible and what is mesocosmic valuable is not incidental. It is a structural feature of what symbols are, and it produces distortions whose consequences range from the trivial to the catastrophic.
3. Levels of Recursivity and Recursive Fluidity
LVT distinguishes five levels of recursivity that describe how coordination becomes available to itself, how the mesocosmic field comes to be felt, named, stabilised, and reflected upon.
L1 is seamless coordination. At L1, life proceeds without explicit attention or deliberation. Actions are performed without awareness of performing them. Skills are enacted without thought. Social arrangements operate without needing to be named. This is the Heideggerian background of everyday life: the world in which equipment is ready-to-hand and transparent in use. L1 is characterised by invisible value. The most effectively functioning coordinations leave least trace because they have become conditions of ordinary life rather than achievements that must announce themselves.
L2 is felt misalignment. Coordination becomes unsettled. Something feels off before it can be named or described. The glass that the hand expects the table to support slides to the edge and falls. A bodily sensation signals that something has changed before any articulate description is available. A social interaction produces an unease that cannot yet be put into words. L2 is both diagnostic and generative: it registers that L1 coordination has been disturbed, and it opens the possibility of remediation, repair, or transformation. Crucially, L2 is ontologically distinct from what follows it. The felt misalignment is real and operative before any symbolic articulation of it exists, it is prior to language, not merely pre-linguistic in the developmental sense.
L3 is symbolic articulation. Something becomes named: this pain, this problem, this feeling, this word. Clinical notes describing a patient's sleep disruption, relational difficulties, and change in appetite are L3 articulations. L3 can name immediately available mesocosmic states or felt misalignments without requiring abstraction from many cases or generalisation across populations. It stays close to the particular.
L4 is abstraction and stabilisation. Symbolic articulation moves beyond the immediate into generalisation, classification, or rule-like organisation. General concepts (illness, disorder, depression), institutional categories (patient, diagnostic code, treatment protocol), and decision-guiding thresholds (normal/pathological, eligible/ineligible) all operate at L4. L4 is the level at which coordination is made actionable across populations and institutions. The most socially powerful L4s combine abstraction with decision-guiding force: 'major depressive disorder' is an L4 category that triggers not merely description but prescription.
L5 is meta-recursive reflection. At L5, the categories and frameworks of L4 themselves become objects of analysis. Critique, genealogy, comparison, and theoretical innovation operate at L5. LVT itself is an L5 enterprise: a symbolic reflection on the conditions, limits, and harms of symbolic reflection.
These levels are not stages that culminate at the highest. Human life moves continuously and fluidly across all five levels, and health, understood in LVT terms, is precisely this capacity for fluid movement. Recursive fluidity is the ability to articulate what has gone wrong, remediate it, and return to seamless coordination at L1. When embodied states require attention, hunger, fatigue, cold, pain, they arise at L2, are addressed at L3 or through practical action, and the successful remediation allows return to L1. Health is not the absence of disturbance but the capacity for appropriate movement across levels in response to disturbance.
Recursive fluidity clarifies what goes wrong in certain institutional practices. Consider a sleep-tracking application. When a person sleeps four and a half hours, this produces an L2 state, a felt sense of tiredness. The application, however, does not report the L2 state. It converts physiological data into an L4 classification: 'danger zone,' 'severe deficit,' 'long-term harm.' The felt misalignment is bypassed, and an institutional verdict is imposed directly. The person's self-recursivity, the ordinary capacity to assess one's own bodily states, is colonised by a symbolic system that claims more authority than the person's own felt sense of whether they have rested. The result is not merely inconvenient. It disrupts recursive fluidity by locking the person into an L4 assessment they cannot interrogate from below.
4. Three Domains of Recursion and Mixed Futurity
Domain recursivity describes the recursive structure of a particular coordination, not the essential property of a mediation. Any mediation can, in principle, be involved in any form of domain recursivity; what matters is the structure of the specific coordination in question.
Non-recursive entities do not respond to being coordinated with, described, or intervened upon. Gravity, temperature, material fatigue, pharmacokinetic regularities, and circadian rhythms are non-recursive in the relevant sense: they do not change in response to how they are understood, measured, or described. This does not mean that humans engage with them passively. Fear responds to gravity. Exhaustion responds to cold. A drug dose does not respond to the hope of the person who takes it, but that person's hope responds to the drug dose in ways that alter the pharmacological encounter profoundly. Non-recursive entities are engaged recursively even as they do not themselves respond to that engagement. They are part of the mesocosm through the recursive mediations through which living beings engage them, not as independent presences.
Self-recursive loops are processes in which a system responds to its own ongoing states. Embodiment is characteristically the site of self-recursive coordination: pain amplifies vigilance, which amplifies sensitivity to further pain; fatigue alters effort thresholds; sleep affects mood, which affects appetite, which affects energy, which affects sleep. The lived body is a massively self-recursive system operating across multiple sensory modalities, temporal scales, and physiological registers simultaneously. Self-recursive loops can be constrained and predictable, normal homeostatic regulation operates within relatively stable parameters, or destabilised and volatile, as in the self-amplifying dynamics characteristic of certain pathological states.
Inter-recursive coupling is the form of domain recursivity that most distinguishes living social life from everything else. In inter-recursive coupling, two or more systems respond to each other in ways that genuinely constitute the unfolding situation rather than merely influencing it. A conversation between two people exemplifies inter-recursivity: each utterance responds to and transforms the field within which the next utterance is produced. Neither participant knows what the other will say; what the other says depends on what is said to them; and what is said to them was generated by the same reciprocal process. The conversation cannot be predicted because it has not yet been constituted by the recursive exchange that will produce it.
This is not complexity in the technical sense, and the distinction is important for social theory. Complexity theory describes systems with many interacting variables whose future states are difficult to calculate but are, in principle, deterministic and projectable. A weather system is complex: it does not respond to how it is described, and its dynamics are, at least in principle, projectible, even if in practice they become intractable beyond short horizons. A conversation is inter-recursive: its future states are not merely difficult to calculate but are genuinely constituted through the interaction itself. The weather is complex but non-recursive; a conversation is inter-recursive and genuinely open. Confusing the two has significant consequences for how institutions are designed and how failures are understood. A system that fails because it is too complex might be improved by better computing power or finer-grained data. A system that fails because it treats inter-recursive domains as if they were complex non-recursive ones will fail in the same way regardless of the data it accumulates.
Mixed futurity names the ontological condition of the mesocosm: the simultaneous presence of non-recursive constraints, self-recursive loops, and inter-recursive couplings in any given situation. The mesocosm is never purely non-recursive. Gravity acts on a body, but the body responds to gravity through fear, anticipation, skill, and injury, all of which involve self-recursive and inter-recursive processes. A drug has pharmacokinetic properties that are non-recursive, but it is taken by a person who interprets its effects, responds to the meanings of taking it, and acts within a relational and institutional field that shapes what the drug can do. A diagnostic category describes a pattern, but it also loops back onto the person diagnosed, reorganising their self-understanding, their anticipation of the future, and their relationship to their own suffering.
Mixed futurity implies that cognitive responses to the future must also be mixed. Prediction, the projection of future states from prior states using symbolic models, is the adequate cognitive response where the domain is predominantly non-recursive or tightly constrained in its self-recursivity. Attunement, reading a situation as it unfolds and responding from within it, without requiring prior symbolic stabilisation of what the situation is, is required where the domain is constitutively inter-recursive. These are not competing epistemologies or rival research programmes. They are responses to genuinely different ontological conditions. Preparation, the construction of prepared topologies of possible futures, the scripting of contingencies, the design of redundancies, is a practical strategy for managing mixed conditions, but it is not a third mode of futurity alongside prediction and attunement. Preparation remains symbolic: it maps what might happen using L3 and L4 categories. What distinguishes it from prediction proper is that it holds the map lightly, ready to abandon the prepared topology when the inter-recursive situation demands something outside it.
Rocky's barn training is not anti-scientific. It is calibration for an inter-recursive encounter that Drago's sensors cannot model, and it works not because it is more authentic or more embodied but because it has better ontological fit with the domain it is preparing for.
5. Modern Institutions as Predictability Machines
Modern institutions are predictability machines. This is not a sociological metaphor but an ontological description. Laboratories, clinical trials, bureaucracies, financial models, audit regimes, and algorithmic platforms are all systems for extracting projectable stability from a mesocosm characterised by mixed futurity. They achieve this through ontological surgery: the selective suppression of self-recursive and inter-recursive dynamics and the amplification of non-recursive regularities.
Ontological surgery is extraordinarily powerful where the domain tolerates it. Physics, structural engineering, pharmacokinetics, and short-range meteorology are domains in which non-recursive regularities are sufficiently dominant that suppressing recursive dynamics yields genuine predictive stability. The table will still be here in ten minutes. The drug will clear the bloodstream at a calculable rate. The bridge will bear its calculated load or it will not. These knowledge systems have achieved excellent ontological fit between their methods and the domains to which they are applied, and their success is real.
The problem emerges when ontological surgery is performed on domains constituted by the dynamics being suppressed. Hartmut Rosa (2024) touches on this in his account of the transition from situations to constellations in modern social life. Rosa argues that modern institutions have systematically replaced the open inter-recursive situation, in which what happens is genuinely co-constituted by participants, with institutionally scripted encounters that constrain deviation from expected scripts. The supermarket interaction, the administrative procedure, the bureaucratised clinical consultation, all are designed to function as if the inter-recursive human being on one side of the counter were a non-recursive entity generating predictable outputs. Rosa correctly identifies the suppression of situational inter-recursivity as a defining feature of modern institutional life. LVT provides the structural explanation for why this happens and why it systematically fails: institutions are designed for prediction, and prediction requires the amplification of non-recursive regularities, which in turn requires the suppression of the recursive dynamics that make the domain what it actually is.
The institutional suppression of recursion is never complete, and it fails in a characteristic way. Two forms of failure are most important. The first is porosity: recursive dynamics persist inside the institutions that claim to have eliminated them. The clinical trial participant responds to the meaning of participation, to the clinician's attention, to the social experience of being cared for; these recursive couplings produce the placebo response that the trial is designed to suppress. The bureaucratic client finds ways to game the categories that are supposed to process them neutrally; the audit metric becomes an incentive to optimise for visibility rather than for the underlying coordination the metric was designed to measure. The second failure is collapse: when the suppression conditions end, the full mesocosmic complexity reasserts itself. The drug that worked in the trial operates differently in daily life. The financial model that worked in the stable period fails catastrophically when market dynamics shift. The policy that was effective under controlled conditions produces unintended consequences when implemented at scale.
The most consequential failure is neither porosity nor collapse but misrecognition. Modern institutions mistake recursion-managed environments for reality itself. When the suppression conditions cease to hold, when people game the metric, when the market moves, when the pandemic spreads through channels the model excluded, the institution does not recognise this as the predictable failure of its suppression conditions. It experiences it as irrationality, non-compliance, volatility, or crisis. The vocabulary of failure (noise, confound, bias, irrational behaviour, treatment resistance) reveals what might be called ontological illiteracy: an incapacity to recognise that what has been managed, not eliminated, has simply reasserted its presence.
This analysis is compatible with Latour's (1993) critique of purification, but more specific. Latour argued that modern thought achieves its power through the separation of what it also mixes, nature and culture, human and nonhuman, fact and value. LVT agrees that purification is modern knowledge's characteristic operation, but the level of analysis differs. Latour's purification works at the level of categories. LVT's ontological surgery works at the level of recursion domains: it suppresses certain kinds of dynamic process in order to make others measurable. This explains not just that purification happens but what precisely is being purified and why the purification systematically fails in certain domains while succeeding in others, a question that Latour's framework cannot fully answer.
6. Antidepressant Trials: The Paradigmatic Case
Mental illness is the most recursively saturated domain in the biomedical sciences. This is not a claim about the severity or the irreducibility of subjective experience; it is an ontological claim about the structure of the coordination that has been disrupted. Depression, anxiety, psychosis, and related conditions are not things located in bodies awaiting detection and treatment. They are configurations of disrupted coordination operating across all five mediations, with dense inter-recursive coupling between them.
In depression: sleep disrupts mood, disrupted mood disrupts appetite, disrupted appetite disrupts energy, disrupted energy disrupts sleep, a self-recursive loop in embodiment, stuck at L2, that cannot be resolved by the ordinary mechanism of remediation and return to L1. The quality of social recognition and care shapes the trajectory of suffering in ways that cannot be separated from the biology; isolation intensifies embodied distress in a further inter-recursive loop. The person's relationship to light, seasonal change, and spatial accessibility shifts in ways that sustain or intensify depressive states. Food, medicines, and the material arrangements of daily life all become differently engaged. And the diagnostic label imposed at L4 reorganises the person's self-understanding at every recursivity level: the prognosis shapes what futures are imaginable; the public narrative about mental illness shapes how the person inhabits their own suffering; the expectation of pharmaceutical treatment shapes how somatic changes are interpreted.
Psychiatric symptoms are, in this account, often embodied self-recursive manifestations of fundamentally inter-recursive disturbances. The fatigue-withdrawal cycle, the sleep-mood amplification loop, and the pain-vigilance dynamic that dominate depressive presentations are not primary biological events underlying which the social context is incidental noise. They are the embodied resonance of disrupted coordination in being-with, dwelling, multimateriality, and multisymbolization. The person who has lost their job, whose relationship has broken down, who lives in inadequate housing, and who is subject to diagnostic narratives that compound stigma presents in the clinic with disrupted sleep, energy, appetite, and mood, not because the social conditions are unimportant but because they are expressed through embodied self-recursive loops that are the body's mode of registering inter-recursive disturbance.
The history of modern psychiatry can be read as a sustained attempt to reverse-engineer this situation: to identify the primary, non-recursive biological event underlying the recursive ecology. Biomarkers are the ultimate expression of this ambition. A biomarker, a measurable biological variable with a deterministic relationship to a clinical outcome, would be a non-recursive element in an otherwise recursive domain, allowing mental illness to be treated like a kidney stone: an object that is either there or not, with a mechanistic relationship to the treatment that targets it. This ambition is not without scientific integrity, but it reflects what LVT identifies as ontological illiteracy, the failure to recognise what kind of domain one is actually working in. Psychiatric research has been failing to find stable biomarkers for decades not because the technology is insufficiently refined but because the domain's most important dynamics are constitutively recursive and therefore resistant to non-recursive treatment. The domain does not lack adequate methods; it lacks adequate ontology.
Antidepressant clinical trials are the institutional form through which this ontological illiteracy is most elaborately and expensively expressed. The Cipriani et al. (2018) network meta-analysis, 522 trials, over 116,000 participants, represents the most sophisticated attempt yet to extract projectable stability from a constitutively recursive domain. The trial achieves this by performing ontological surgery on the recursive ecology of depression.
What the trial preserves is the embodied self-recursive loop of pharmacological action. The drug acts on the body, the body changes, the changed body is assessed through standardised rating scales. What the trial suppresses is everything else: the inter-recursive therapeutic relationship, the meaning of receiving treatment, the social experience of being cared for, the patient's active interpretation of their own changing states, the relational and institutional context within which the drug is taken. These suppressions are not incidental to the trial's design; they are its design. Randomisation suppresses selection effects. Blinding suppresses expectancy effects. Standardised scripts suppress therapeutic relationship effects. Placebo controls suppress meaning effects. The trial is a machine for isolating the pharmacological signal by eliminating the recursive noise, and it is extraordinarily well-designed for this purpose.
The trial's design implicitly acknowledges the power of recursive dynamics, because otherwise blinding and placebo controls would be unnecessary. If expectancy did not matter, there would be no need to suppress it. If the therapeutic relationship did not matter, there would be no need to standardise scripts. The entire apparatus of the randomised controlled trial is a recursive suppression machine whose existence testifies to the power of the dynamics it is trying to eliminate. And yet the trial continues to describe these dynamics as contamination, bias, confound, and noise to be removed, rather than as constitutive features of antidepressant response in its natural environment.
A hypothesis that follows from this analysis concerns the differential performance of antidepressants in comparative effectiveness data. The drugs with the highest efficacy rankings in the Cipriani et al. (2018) meta-analysis, amitriptyline, mirtazapine, agomelatine, are those whose primary therapeutic mechanisms operate through constrained embodied self-recursive loops: sedation and sleep architecture restoration in the case of amitriptyline and mirtazapine, circadian oscillator regulation in the case of agomelatine. These mechanisms run through the channel the trial is designed to preserve and measure. If SSRI effects travel partly through the reduction of rumination, modification of anticipatory threat-appraisal, or alteration of self-monitoring practices, processes that are primarily symbolic-interrecursive, then those effects share operating channels with what the placebo arm also activates. Drug-minus-placebo in such cases removes part of the drug's own effect alongside the placebo response. The result is not a smaller drug effect but a smaller measurable difference. This is a hypothesis about measurability, not about pharmacological effectiveness. It does not imply that SSRIs are ineffective; it implies that the trial is systematically underestimating effects that operate through the recursive channels it is designed to suppress.
7. The Limits of Suppression
The antidepressant trial never fully achieves the suppression it attempts. It is mesocosmically porous from the start: not a purified space that later re-enters the world, but a partially managed space in which recursive dynamics persist throughout.
Side effects undermine blinding. An active antidepressant typically produces somatic signals, dry mouth, altered sleep, sexual dysfunction, gastrointestinal effects, that allow participants and clinicians to infer treatment allocation. Once allocation is inferred, the expectancy loops that blinding was designed to suppress are reintroduced. The blinded trial that the protocol describes and the partially unblinded trial that is actually conducted are different experiments, and their outputs cannot be treated as equivalent.
The clinical encounter generates recursive dynamics that standardised scripts cannot eliminate. A clinician who conducts monthly structured interviews, who asks systematically about mood fluctuations, who is interested in the participant's trajectory, creates a relational field that is itself therapeutically active. The participant who is being regularly attended to, seen, asked about, and cared for within the institutional structure of the trial, is already in a different relationship to their suffering than they were before enrolment. This is the recursive coupling that produces placebo response, and it is generated by the trial's own structure, not imported from outside it.
Hope and disappointment circulate regardless of treatment allocation. The act of enrolment, being selected, randomised, and assigned to an arm, is emotionally significant in ways the trial protocol cannot neutralise. The possibility of receiving active treatment generates anticipatory structures that shape how subsequent bodily changes are interpreted. The possibility of being in the placebo arm generates its own recursive dynamics. These are not noises around a signal; they are part of what constitutes antidepressant response under real mesocosmic conditions.
The placebo effect is the most theoretically important finding in psychiatric pharmacology precisely because it reveals what the suppression apparatus cannot suppress. Placebo is not what remains when pharmacology is absent. It is the visible trace of recursive coupling between meaning, care, anticipation, embodiment, and self-monitoring that persists inside the apparatus. The scale of placebo response in antidepressant trials, consistently in the range of thirty-five to fifty per cent of drug response, and rising over several decades as trial infrastructure has become more elaborate, is not a methodological problem to be corrected. It is an ontological diagnosis of the domain: evidence that the recursive dynamics the trial is designed to eliminate are robust enough to produce clinically significant effects despite increasingly sophisticated suppression. The more intensive the trial's therapeutic context, the more recursive coupling is activated in both arms. Rising placebo response rates are the reciprocal of rising methodological sophistication.
When the trial ends, the world returns. The participant leaves the institution with its scripted interactions, its regular monitoring, its categorical certainty about diagnosis, and its carefully managed recursive environment. They re-enter a field in which being-with is fully inter-recursive, dwelling imposes its seasonal and ecological rhythms, multimateriality operates without the standardisation of dosing and administration, and multisymbolization is neither controlled by the trial's protocol nor anchored by the trial's narrative of improvement. The drug that produced an effect under suppression conditions now operates inside the full mesocosmic complexity that the trial excluded. This is the structural basis of the efficacy-effectiveness gap: the persistent finding that drug effects observed in trials are larger and more reliable than those observed in clinical practice. The gap reflects not poor compliance or inadequate prescribing but the mismatch between the recursion-managed environment in which efficacy was established and the mesocosmic environment in which effectiveness must be achieved.
The parallel with other predictability machines is instructive. Bureaucratic categories that are designed to process human beings neutrally are gamed, subverted, and adapted in ways that reorganise the institution itself; the metric becomes an incentive structure. Algorithmic recommendation systems model past behaviour and then reorganise the environment that generates future behaviour; the model produces the data it subsequently models. Audit regimes measure performance and then discover that institutions have optimised for legibility rather than for the underlying coordination the audit was designed to assess. In each case, the predictability machine achieves temporary local stability at the cost of systematic misrecognition of the domain. And in each case, when the suppression conditions cease to hold, the failure is experienced not as the predictable consequence of ontological surgery on a recursive domain but as irrationality, gaming, non-compliance, or crisis.
8. Conclusion: Towards Ontological Literacy
The argument of this lecture has three connected moments. First, the mesocosm is characterised by mixed futurity: non-recursive, self-recursive, and inter-recursive dynamics are always simultaneously present in lived situations, and these dynamics have radically different predictability profiles. Second, modern institutions are predictability machines that achieve projectable stability through ontological surgery, the selective suppression of recursive dynamics and the amplification of non-recursive regularities. This succeeds where the domain tolerates it and fails systematically where the domain's most important features are the dynamics being suppressed. Third, antidepressant trials are the paradigmatic case: they are sophisticated recursive suppression apparatuses operating on a constitutively recursive domain, and the systematic failure this produces, escalating placebo response, persistent efficacy-effectiveness gaps, the chronic inability to identify stable biomarkers, is ontologically predictable from the mismatch.
The conclusion is not that prediction is wrong, that science is ideologically motivated, or that clinical research should be replaced by qualitative description. LVT provincialises prediction rather than repudiating it. Structural engineering, pharmacokinetics under controlled conditions, and circadian biology are domains in which non-recursive regularities can be extracted and projected with excellent ontological fit. Prediction in these domains is not ontological illiteracy; it is appropriate cognitive response to the structure of the domain. The question is always: what kind of recursive domain is this, and what cognitive and institutional strategies have genuine mesocosmic fit with it?
Ontological literacy, the capacity to answer this question, is what psychiatry most urgently lacks. It does not lack funding, technical sophistication, or the genuine commitment of many of its practitioners. It lacks a framework for recognising what kind of ontological terrain it is actually working in. This matters because the appropriate interventions, and the appropriate forms of evidence for those interventions, are different in constitutively recursive domains from what they are in projectible ones. The pressure on the Yellow Tulip Project, an international initiative that brings people with mental health difficulties together for gardening and collaborative art, to produce randomised controlled evidence of its effectiveness is not a neutral methodological demand. It is a demand that the project redesign itself to produce measurable effects through the non-recursive channel that the trial can isolate, while ignoring the recursive dynamics, the being-with of collective engagement, the dwelling in outdoor space across seasonal change, the material resistance of soil and plant, that constitute most of what the project actually does. Supplying this kind of evidence would not demonstrate that the project works; it would demonstrate that the project can be made to look like something the trial can measure. Ontological literacy would make it possible to say this precisely rather than merely feeling it.
The case of Stephen Laurie, the Edinburgh psychiatrist whose seven years of work with a patient was reconstructed in a collaborative autoethnography, illustrates what ontological literacy looks like in practice. What distinguishes Laurie's work is not a departure from psychiatric methods but a consistently careful refusal to escalate from L3 to L4 before the L3 articulation has been adequately developed. He stays close to the particular, to this person's sleep, this person's relationships, this person's changing situation, and moves to diagnostic categories and pharmacological decisions as lightly and as temporarily as the situation requires. The contrast, in the same account, is with the clinician who administers a standardised scale, produces an aggregate score, and issues a diagnosis and a prescription: the DSM in action as a machine for jumping from felt misalignment to institutional category without passing through the lived situation at all. The difference is not one of resources or goodwill. It is the difference between a practice that maintains ontological fit with the domain and one that does not.
LVT does not, finally, propose attunement as a wholesale replacement for prediction, or Rocky's mode as a wholesale replacement for Drago's. The best practices in high-recursivity domains hold prepared symbolic models, anticipatory topologies, clinical frameworks, institutional protocols, lightly enough to depart from them when the inter-recursive situation demands something outside the model. Duke's corner strategy is not worthless because the fight enters territory outside the prepared scenarios; it is simply that Rocky, not Duke, must read what is actually happening when it does. The ontological literacy that LVT proposes is not a romantic rejection of systematic knowledge but a structural account of which cognitive modes have genuine fit with which domains, and what it costs when the fit is wrong.
The ontological diagnosis is ultimately this. Modernity succeeded at prediction so brilliantly, in so many domains, that it acquired an institutional commitment to the methods that produced those successes, and then applied those methods to domains they were not designed for. The result is not failure in general but systematic failure in specific places: precisely the places where the recursive dynamics of the mesocosm most resist the surgery that modernity's predictability machines are designed to perform. Recognising this is not a counsel of despair. It is the beginning of a different kind of institutional design.
References
Bateson, G. (1972) Steps to an Ecology of Mind. New York: Ballantine Books.
Cipriani, A., Furukawa, T.A., Salanti, G., Chaimani, A., Atkinson, L.Z., Ogawa, Y., Leucht, S., Ruhe, H.G., Turner, E.H., Higgins, J.P. and Egger, M. (2018) 'Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis', The Lancet, 391(10128): 1357–66.
Ecks, S. (2013) Eating Drugs: Psychopharmaceutical Pluralism in India. New York: New York University Press.
Ecks, S. (2022) Living Worth: Value and Values in Global Pharmaceutical Markets. Durham, NC: Duke University Press.
Hacking, I. (1995) 'The looping effects of human kinds', in Sperber, D., Premack, D. and Premack, A.J. (eds) Causal Cognition: A Multidisciplinary Debate. Oxford: Clarendon Press, pp. 351–83.
Heidegger, M. (1962 [1927]) Being and Time, translated by Macquarrie, J. and Robinson, E. Oxford: Blackwell.
Kirsch, I., Deacon, B.J., Huedo-Medina, T.B., Scoboria, A., Moore, T.J. and Johnson, B.T. (2008) 'Initial severity and antidepressant benefits: a meta-analysis of data submitted to the Food and Drug Administration', PLoS Medicine, 5(2): e45.
Latour, B. (1993 [1991]) We Have Never Been Modern, translated by Porter, C. Cambridge, MA: Harvard University Press.
Rosa, H. (2024) Situation und Konstellation: Wider die Unverfügbarkeit des Lebens. Frankfurt: Suhrkamp.
Scott, J.C. (1998) Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. New Haven, CT: Yale University Press.
Taleb, N.N. (2007) The Black Swan: The Impact of the Highly Improbable. New York: Random House.
Tetlock, P.E. (2005) Expert Political Judgment: How Good Is It? How Can We Know? Princeton, NJ: Princeton University Press.
Turner, E.H., Matthews, A.M., Linardatos, E., Tell, R.A. and Rosenthal, R. (2008) 'Selective publication of antidepressant trials and its influence on apparent efficacy', New England Journal of Medicine, 358(3): 252–60.
Uexküll, J. von (2010 [1934]) A Foray into the Worlds of Animals and Humans, with A Theory of Meaning, translated by O'Neil, J.D. Minneapolis: University of Minnesota Press.
Walsh, B.T., Seidman, S.N., Sysko, R. and Gould, M. (2002) 'Placebo response in studies of major depression: variable, substantial, and growing', JAMA, 287(14): 1840–47.