Case Studies in Unearned Confidence, Part 2: The Disastrous Applications of Bandura’s “Self-efficacy” in Health Research

Abstract

Albert Bandura’s concept of self-efficacy has dominated health psychology for nearly five decades. This article argues that self-efficacy theory is not merely theoretically imprecise but ontologically misconceived, and that the two principal alternatives proposed to correct it , collective efficacy and social capital , repeat the original error at a different scale. All three constructs share a common architecture: they locate the determinant of health behaviour in a container, whether individual or collective, and treat that container as the proper unit of intervention. Living Value Theory dissolves this container logic entirely. What appears, through the individualist lens, as a cognitive property and, through the collectivist lens, as a community asset is better understood as a pattern of interrecursive commitment operating across the five mediations of coordinated social life: embodiment, being-with, dwelling, multimateriality, and multisymbolization. Interventions designed to restructure these mediational configurations will be more effective, more honest, and more equitable than any programme designed to boost confidence, whether in one person or in an entire community.

I. The Emperor Has a Questionnaire

Molly is in her late 30s. She lives in one of the UK’s “most deprived” areas. She has more than a dozen diagnoses and is taking a vast range of medications on a daily basis: antidepressants, beta blockers, weight loss pills, antidiabetics, medications for abdominal cramping, and on and on (Ecks 2021). In one of the consultations at the NHS Wellbeing Centre in her local area, Molly reflected on how she is handling her health on a daily basis: “My self-care is really bad, I know that it is really bad.” She professes to eating spicy Indian food even though she knows it will irritate her bowels. She takes Imodium daily even though she knows that this medication should only be taken in acute cases of diarrhoea. Molly admitted that she struggles to resist unhealthy snacks: ”Pringles are my downfall.” Molly drank three to four litres of Coca-Cola throughout the day: ”I am really bad with fizzy drinks.” She knew that surgary soft drinks could cause obesity. She did not know that these drinks could literally kill her because of her diabetes. If you asked an applied health researcher what is going on here, they would, in all likeliness, tell you that Molly has “low self-efficacy.” Her health would be vastly better if she had more “self-efficacy”: more confidence that she can take control of her diet and her life. They would recommend to the clinician to find ways of bolstering Molly’s self-efficacy. This article shows that the dominant paradigm of “self-efficacy” is catastrophically wrong when applied to health.

There is a particular kind of intellectual dominance that sustains itself not through superior explanation but through institutional inertia, measurement convenience, and the capacity to absorb criticism without changing. Self-efficacy theory in health has achieved exactly this kind of dominance. A search of PubMed returns more than 50,000 results. The construct appears in the foundational texts of health promotion, in nursing curricula, in the design logic of apps for smoking cessation and diabetes management, in the training manuals of community health workers from Lagos to Glasgow. It is, by any measure, the most pervasively applied psychological construct in the history of health research.

The source of this dominance is Albert Bandura’s 1977 paper proposing that an individual’s belief in their capacity to execute the behaviours required to produce a specific outcome is the primary cognitive mediator between knowledge, motivation, and action. The framework was immediately attractive to public health precisely because it was tractable: self-efficacy was measurable via Likert-scale questionnaire, modifiable through four specified mechanisms (mastery experience, vicarious learning, verbal persuasion, physiological feedback), and apparently applicable across any domain in which human beings did or failed to do things. It offered clinicians and researchers a handle. The fact that this handle was attached to the wrong door has gone largely unexamined.

This article is not a balanced review. The balanced review has already been written, more than once, and has had essentially no effect on the field’s practice. The present argument is a structural one: self-efficacy theory is wrong in a specific, diagnosable way, and the two bodies of work that have arisen to correct it , collective efficacy research and social capital theory , are wrong in the same way at a larger scale. The error runs through all three frameworks. It is not an error of emphasis or of insufficient nuance. It is an ontological error about where human capacity is located and what it is composed of. Until that error is addressed, adding variables to the model, subdividing the construct, or shifting the unit of analysis from the individual to the group will continue to produce frameworks that are administratively convenient and theoretically hollow.

II. What Bandura Found and What He Thought He Found

Intellectual honesty requires acknowledging that Bandura discovered something real: explicit declarations of future orientation, made in socially significant contexts, do indeed influence subsequent performance differences. Versions of this have been “replicated” in thousands of studies. A 2016 meta-analysis by Sheeran and colleagues, examining 204 experimental studies, argued that self-efficacy had a medium-size effect on health-related behaviours. Specific interventions, such as the Chronic Disease Self-Management Program, built explicitly on Bandura’s framework, have probably not made people’s health worse. The critique that follows here is not saying that all these researchers have not been studying something. The critique is that all of them have completely misrecognized what exactly they were studying.

What Bandura thought he found was a general internal cognitive mechanism: a belief, located inside the individual, that causes that individual to persist in the face of difficulty and thereby to achieve better outcomes. This story has four components, each of which is wrong. First, self-efficacy is not located inside the individual. Second, it is not a belief in any philosophically useful sense. Third, it does not cause performance through any specifiable internal mechanism. Fourth, it is not a general mechanism that operates uniformly across domains. Each of these errors has consequences. Together, they produce a framework that cannot explain its own most important findings and cannot identify the conditions under which its interventions will work.

Consider what it actually means to say that a patient with Type 2 diabetes has high self-efficacy for blood sugar management. Bandura’s framework says this means that the patient has an internal cognitive property, a confident belief, that will cause them to monitor their diet, take their medication, and exercise regularly. Living Value Theory says something different. What the patient has is a particular mediational configuration: a body with certain current metabolic parameters, a set of social relationships with family members, healthcare providers, and peers who share or do not share their condition, a living environment with or without access to affordable healthy food, a material infrastructure of lancets and testing strips and insulin that may or may not be reliably available, and a symbolic framework within which diabetes management is understood as worthwhile or shameful, as part of a coherent identity or as an unwanted intrusion on a normal life. The self-efficacy questionnaire captures some signal from this entire configuration and reduces it to a number between one and ten. Then the field treats the number as if it were the cause of the configuration it imperfectly reflects.

This is not a measurement problem to be fixed with better scales. It is a category error. The questionnaire is being asked to measure a cognitive property that does not exist as described. What it is actually doing is something more interesting: it is functioning as what the South Asian philosophical tradition calls a sankalpa, a formal declaration of future intention made in a socially significant context. The declaration does not express a pre-existing internal state. It constitutes a new social fact. The patient who rates their medication adherence self-efficacy at eight out of ten in the presence of a healthcare provider has made a public commitment that reorganises the recursive ecology within which their subsequent behaviour will occur. The provider now holds an expectation. The patient now inhabits a world in which they have said something that can be confirmed or disconfirmed. The declaration is not a report; it is a performance. Self-efficacy research has been studying ritual commitment while believing it was measuring cognitive states.

III. The Recursivity Error and Its Systematic Consequences

The most consequential error in Bandura’s framework is not the individualism, though that is serious enough. It is the failure to distinguish between different types of recursive processes and the systematic conflation of mechanisms that operate quite differently. Living Value Theory distinguishes three fundamental recursivity domains: non-recursive processes, in which the elements of the process do not respond to expectations about them; self-recursive processes, in which the organism’s relationship to its own states is part of what shapes those states; and interrecursive processes, in which multiple parties recursively affect each other’s expectations, orientations, and behaviours.

The distinction matters because self-efficacy’s predictive power is not uniform across these domains. In non-recursive domains , the purely physical aspects of clinical outcomes, the pharmacokinetics of medication, the cellular mechanisms of wound healing , the patient’s confidence has no direct causal effect. At most, it may influence whether they attempt the task and persist in attempting it. The underlying process is indifferent to their beliefs. In self-recursive domains , pain experience, anxiety, fatigue , the organism’s relationship to its own states does genuinely shape those states, and self-efficacy research captures something real here, though still not through the mechanism Bandura described. In interrecursive domains , which constitute the overwhelming majority of health behaviour contexts , what the patient believes and declares affects what others believe and do, which affects what the patient subsequently does, which affects what others subsequently believe, in a loop that can only be understood at the level of the coordinative ecology rather than the individual mind.

Because the field has not made this distinction, it has produced two systematic errors. The first is the overclaiming of self-efficacy’s causal power in non-recursive domains, where it functions at most as a proxy for readiness and not as a cause of outcome. The oncology patient’s confidence in their ability to tolerate chemotherapy does not alter the pharmacology of the drug. It may affect their behaviour around appointments, side-effect reporting, and help-seeking, all of which are interrecursive. But the conflation of these distinct mechanisms under a single framework has produced interventions targeting internal beliefs in domains where the actual determinants of outcome are structural and material.

The second systematic error is the misidentification of interrecursive effects as cognitive ones. When self-efficacy interventions work in social domains , chronic disease self-management groups, addiction recovery programmes, rehabilitation settings , they work because they restructure the interrecursive ecology: they create communities of shared commitment, establish social witnessing of declared intentions, build accountability structures that make giving up socially costly. The Chronic Disease Self-Management Program is effective not because it installs a belief but because it creates a group of people who have made commitments to one another in a context that makes those commitments real. The self-efficacy framework describes the shadow of this process and calls it a cognitive mechanism.

IV. The Measurement Is the Intervention

One of the most carefully avoided implications of self-efficacy research is that the measurement instrument produces what it purports to measure. This is not a methodological quibble. It is a structural feature of the entire research programme that undermines the validity of thousands of studies.

If a self-efficacy questionnaire is not a neutral measurement of a pre-existing internal state but a formal declaration that constitutes a new social fact, then any study that measures self-efficacy at baseline and post-intervention and attributes behavioural change to a change in self-efficacy scores has confounded the effect of the intervention with the effect of the measurement. The patient who completes a post-intervention self-efficacy questionnaire rating their confidence in diabetes self-management at eight out of ten has made a public commitment, in the presence of a researcher, that will enter their recursive ecology and affect their subsequent behaviour. The behaviour change then attributed to the self-efficacy intervention may be partly or substantially caused by the measurement itself.

The evidence for this interpretation is embedded in findings that the field has consistently failed to theorise. Self-efficacy questionnaires predict performance better when completed immediately before a performance than when completed well in advance. They predict better in public contexts than in private ones. They predict better when administered by a credible authority than by an anonymous researcher. Each of these findings is inexplicable on the assumption that self-efficacy is an enduring internal cognitive property: if it were such a property, its predictive power should not vary with the timing or social context of its measurement. Each finding is entirely explicable on the sankalpa account: declarations made closer to the performance moment, in more socially significant contexts, in the presence of more authoritative witnesses, constitute stronger commitments and reorganise the recursive ecology more powerfully.

The practical implication is that the field has been running not experiments but rituals, without knowing it. The clinical trial in which patients complete a self-efficacy questionnaire before an intervention is not measuring a baseline cognitive variable and then assessing the intervention’s effect on it. It is administering two doses of commitment ritual: one at baseline and one at follow-up, with the intervention sandwiched between. The attribution of effects to the intervention rather than to the measurement protocol is, in many studies, unsupported. This does not mean the interventions are ineffective. It means the field does not know why they are effective, and this ignorance has consequences for how they are designed, scaled, and adapted.

There is a further dimension to this problem that has received almost no attention. The mediation analysis, the analytic workhorse of much self-efficacy intervention research, assumes that changes in self-efficacy scores transmit the effect of the intervention to the health outcome. If the intervention raises self-efficacy, and raised self-efficacy produces better behaviour, then self-efficacy is the mechanism. But if the self-efficacy questionnaire is itself doing coordinative work , if completing it at follow-up is itself a commitment act that influences subsequent behaviour , then the mediation model is circular. The measurement of the mediator is not independent of the mediator’s effect. The entire causal architecture collapses. Studies that have used this design, which is to say most of the most influential self-efficacy intervention studies, have produced results that cannot be interpreted in the way the authors claim. This is not a minor methodological limitation. It is a fundamental problem of experimental design that follows directly from the ontological error at the framework’s core.

V. Neoliberalism’s Favourite Construct

The political economy of self-efficacy theory is not incidental to its intellectual failures. The two are connected. Self-efficacy flourished in a specific ideological climate , the consolidation of neoliberal governance from the 1980s onward , and it flourished precisely because its ontology was politically convenient. A framework that locates the determinant of health behaviour in the individual’s internal cognitive state is a framework that systematically obscures the structural determinants of health. If people fail to manage their chronic disease, eat unhealthily, smoke, drink, or fail to exercise, the self-efficacy framework implies a deficit in their internal cognitive machinery. The solution is modification of that machinery through training, counselling, and skills development. Structural change is not indicated because structure is not the problem. The individual’s beliefs are the problem.

This framing has real consequences. As Card and colleagues have documented, neoliberal health policies that emphasise individual responsibility and behavioural change have been associated with widening health inequalities, systematically benefiting those with the social and material resources to act on their beliefs while leaving unchanged the conditions that make behaviour change impossible for others. The person without safe spaces to exercise, affordable healthy food, stable employment, reliable childcare, or healthcare access cannot simply adjust their self-efficacy upward and overcome these structural barriers. The self-efficacy framework, applied to this person, does not merely fail to help. It actively misidentifies their problem and thereby obstructs the identification of actual solutions.

The field’s response to this critique has been to add structural variables to models as moderators and mediators of self-efficacy effects. This is worse than insufficient. It preserves self-efficacy as the primary explanatory variable while treating structural determinants as background conditions that affect how much self-efficacy matters. The poverty and food insecurity that prevent behaviour change are thus repositioned as factors that reduce self-efficacy, rather than as the actual causes of the behaviour pattern. The reframing is not merely inaccurate. It is ideologically functional: it continues to locate the problem in the individual’s internal states, even while ostensibly acknowledging structural context.

A particularly revealing symptom of this problem is the persistent conflation of capability and motivation in self-efficacy measurement. When a single parent working two jobs rates their exercise self-efficacy at three out of ten, they are almost certainly not reporting a belief that they lack the physical capacity to walk or ride a bicycle. They are reporting something about their motivational configuration: exhaustion, competing time demands, the absence of childcare, the unavailability of safe walking routes, the impossibility of prioritising exercise over paid work and child supervision. The self-efficacy questionnaire captures this motivational state accurately. The theoretical framework then misidentifies it as a capability deficit and proposes motivational counselling to address it. The result is an intervention that targets the wrong level of the problem, leaves its actual determinants unchanged, and then attributes its limited effectiveness to the patient’s insufficient engagement with the programme.

VI. Collective Efficacy: The Same Error, Scaled Up

The failures of self-efficacy theory have not gone unnoticed. A substantial body of research has proposed collective efficacy as a corrective: the shared belief of a group or community in its capacity to achieve desired goals. Albert Bandura himself introduced the concept, and it has been elaborated extensively in community health research. The argument is straightforward: if individual self-efficacy is insufficient because health behaviours are shaped by social context, then we need a construct that captures the group’s collective confidence in its capacity for coordinated action. Communities with higher collective efficacy, the research shows, have lower rates of obesity, depression, and risk-taking behaviours, independent of individual-level self-efficacy.

This move appears to be a genuine advance. It is not. Collective efficacy makes exactly the same ontological error as individual self-efficacy, applied to a different container. Instead of locating the determinant of health behaviour in an internal cognitive property of the individual, it locates it in a shared cognitive property of the group. The group now has a belief about its capacity. This belief is measurable, modifiable, and causally efficacious. The intervention logic follows exactly the same template: assess the collective efficacy deficit, design programmes to raise collective confidence, measure the improvement, attribute the behaviour change to the cognitive shift. The unit of analysis has changed. The framework has not.

The problems that afflict individual self-efficacy afflict collective efficacy in amplified form. If self-efficacy questionnaires do not measure internal cognitive properties but perform commitments, what do collective efficacy questionnaires do? They perform community commitments. Neighbourhoods surveyed about their collective capacity to maintain public order, resist drug use, or promote healthy eating are being asked to make, in aggregate, declarations of communal intent. These declarations can reorganise the recursive ecology of community life in meaningful ways. But the attribution of subsequent behaviour change to the measured cognitive property , rather than to the social act of collective declaration itself , is no more defensible at the group level than at the individual level. The questionnaire is again being confused with the phenomenon it purports to measure.

More fundamentally, collective efficacy research has no adequate account of what a community is or how its shared beliefs are constituted, transmitted, and contested. The concept treats the community as if it were a large individual: a bounded entity with unified internal states that can be measured by asking its members to rate their confidence. But communities are not individuals. They are configurations of interrecursive relations, material conditions, historical sedimentations, institutional structures, and symbolic systems that produce different outcomes in different configurations. Two communities with identical collective efficacy scores may be organised through entirely different recursive logics , one through a dense religious institutional infrastructure, another through neighbourhood-based mutual aid networks, a third through fear of shared stigma , and these differences will determine what their shared confidence can and cannot accomplish. The collective efficacy framework cannot see these differences because it has already collapsed them into a single number.

The deepest failure of collective efficacy is that it relocates the individual-society binary rather than dissolving it. Individual self-efficacy posits an individual, defined by internal cognitive properties, in a social environment that shapes those properties but is external to them. Collective efficacy posits a collectivity, defined by shared cognitive properties, in a social environment that shapes those properties but is external to them. The binary is preserved. The individual has simply been scaled up into the group, and the group is now treated as the unit with the relevant internal properties. This is not a theoretical advance. It is a structural repetition of the original error at a higher level of aggregation.

VII. Social Capital: The Sinkhole Beneath the Bunting

Social capital occupies a different institutional niche from collective efficacy, but it commits the same fundamental error with even greater theoretical ambition and even less theoretical precision. Robert Putnam’s account, which has been most influential in health research, proposes that the density of civic associations, the level of generalised trust, and the extent of voluntary participation in community life constitute a form of social capital that can be accumulated, maintained, and spent, and that communities with higher stocks of this capital enjoy better health outcomes.

The appeal of this framework to health researchers is intelligible. Epidemiological evidence has consistently shown that social connectedness predicts health outcomes across a wide range of conditions. Communities with stronger social ties tend to have lower mortality, lower rates of mental illness, better recovery from acute illness, and more effective management of chronic disease. The social capital framework offers a conceptual vocabulary for naming and measuring what appears to be at work in these associations. The problem is that the vocabulary it offers is wrong in a way that makes it impossible to specify the actual mechanisms at stake.

The word “capital” does enormous theoretical damage that the concept’s proponents have consistently underestimated. Capital is a non-recursive symbolic form: it accumulates, transfers, converts, and depreciates in ways that are indifferent to the social relationships through which it flows. A pound coin does not change its value based on whether it passes between friends or strangers, employers or debtors, communities with histories of mutual trust or communities with histories of exploitation. This indifference to relational context is not a deficiency of money. It is money’s defining achievement. By abstracting from social thickness, money becomes fungible, comparable, and calculable. These are powerful properties for financial analysis and catastrophic properties for the analysis of human relationships.

Being-with is the precise opposite of a capital form. A relationship between two people is never simply a quantity of trust or reciprocity that can be converted into other forms of value. It is a specific, historically embedded, morally charged, recursively responsive configuration that changes its meaning, its practical consequences, and its emotional texture depending on context, history, obligation, shame, anticipation, and the recursive responses of both parties. The kinship network that provides survival support in conditions of extreme poverty is not the same as the civic association that facilitates access to employment information, even if both produce the same social capital score. They operate through entirely different recursive logics, create entirely different obligations, and produce entirely different practical constraints.

The health literature on social capital has documented this problem empirically without recognising its theoretical implications. Studies have found that dense social networks in low-income communities can simultaneously reduce chronic disease risk, increase social support for illness management, restrict individual mobility, create unsustainable obligations, and produce the concealment of health information when disclosure would generate claims that network members cannot meet. The same network density that predicts better diabetes control in one context predicts worse mental health in another, and the social capital framework has no resources for explaining the difference because it can only see the density and not the recursive logic through which density produces outcomes.

Most damagingly for health equity, the social capital framework has served as a soft austerity concept that translates structural deprivation into a cultural deficit. When a low-income community has poor health outcomes, the social capital diagnosis identifies a deficit of civic engagement, generalised trust, or voluntary participation and prescribes more associational activity. This translates the material questions , why are people in this area unable to access healthy food, safe housing, stable employment, or adequate healthcare? , into a cultural question about whether residents attend enough committee meetings. It asks people to network their way out of institutional abandonment. In health research, this substitution has had real consequences: community health interventions built on social capital frameworks have directed resources toward trust-building workshops and civic engagement initiatives in contexts where the actual determinants of poor health outcomes were poverty, housing insecurity, and lack of access to healthcare infrastructure.

What the collective efficacy and social capital literatures have together demonstrated, despite themselves, is that the social conditions of health behaviour are not adequately captured by any single construct, at any level of aggregation. The consistent failure to specify the mechanism through which collective beliefs or relational assets produce health outcomes is not a gap to be filled by further research within the existing frameworks. It is a symptom of the frameworks’ constitutive inadequacy. No measure of community-level confidence can specify what forms of being-with are actually operating, what obligations they create, how they interact with the material and institutional conditions of the community, or whether the dense social ties in a given setting tend toward collective flourishing or toward the kind of compulsory redistribution that Carol Stack documented in her ethnography of urban kinship networks, where the same density that enabled survival also made individual accumulation nearly impossible. The concept of collective efficacy has no resources for distinguishing these cases. It can see the density of shared belief but not the recursive logic through which that belief produces, or fails to produce, health.

VIII. Three Constructs, One Binary

The common architecture of self-efficacy, collective efficacy, and social capital is now visible. All three locate the determinant of health behaviour in a container: the individual’s internal cognitive state, the group’s shared cognitive state, or the community’s accumulated relational assets. All three treat this container as the proper unit of measurement and the proper target of intervention. All three produce frameworks that are administratively convenient , because they can be measured with questionnaires, expressed as numbers, aggregated across populations, and targeted with discrete interventions , and ontologically inadequate, because the phenomena they purport to explain are not located in containers of any kind.

The individual-society binary that all three frameworks presuppose is a particular feature of modern Western social thought, not a feature of the phenomena they study. Individual self-efficacy assumes an individual whose internal properties are shaped by social experience but who remains the locus of the relevant causal variable. Collective efficacy assumes a community whose shared properties are shaped by structural conditions but which remains the locus of the relevant causal variable. Social capital assumes a stock of relational assets that is generated by social life but which can be abstracted from that life and treated as an independent explanatory variable. In each case, a boundary is drawn between an inside and an outside, the relevant causal variable is placed inside, and the outside is treated as context that shapes the variable but is not constitutive of it.

It is worth noting that more sophisticated versions of all three frameworks have been developed in response to exactly this criticism. Socio-ecological models of health behaviour, multilevel analyses of self-efficacy within structural contexts, mixed-method studies that supplement survey data with ethnographic attention to lived conditions: the field’s methodological toolkit has become more elaborate. But methodological sophistication does not dissolve a conceptual error. A multilevel model that treats self-efficacy as the individual-level variable and neighbourhood deprivation as the community-level variable has not dissolved the individual-society binary. It has formalised it as a statistical architecture. The binary is now operationalised as levels in a hierarchical model, with the individual’s cognitive property as the level-one unit and the structural condition as the level-two modifier. This is a more sophisticated form of the original error, not a transcendence of it.

This binary has been extraordinarily productive for a particular style of health research: the large-scale survey study that measures individual or community attributes, correlates them with health outcomes, and produces effect size estimates that can be compared across populations and contexts. It has been almost entirely unproductive for understanding what is actually happening in the processes that determine health behaviour, because what is happening is not located inside individuals or communities but in the recursive configurations of embodied, social, material, and symbolic life through which action is coordinated.

The capacity to manage a chronic illness is not inside the patient. It is distributed across the patient’s body and its current physiological state, their relationships with healthcare providers and family members and peers with similar conditions, the material environment of their home and neighbourhood, the symbolic frameworks through which illness management is understood and valued, and the institutional configurations that make relevant knowledge, medication, and support available or unavailable. None of these elements is separable from the others without losing what matters about each. The patient’s “self-efficacy” is a name for one snapshot of this entire configuration, taken at a particular moment, in a particular social context, and treated as if it were a property of the patient rather than a feature of the configuration.

IX. What an Adequate Framework Requires

An adequate framework for understanding health behaviour must begin not from the individual or the community as the unit of analysis but from the mesocosm: the specific, situated configuration of interrecursive processes through which action is coordinated in a particular context. This is not a demand for descriptive complexity for its own sake. It is a theoretical requirement that follows from what health behaviour actually is.

Living Value Theory proposes that human coordination is constituted through five irreducible mediations: embodiment, being-with, dwelling, multimateriality, and multisymbolization. These five mediations are not additive components that can be measured separately and summed. They are dimensions of a single coordinative reality that are distinguishable analytically but inseparable practically. Any adequate account of health behaviour must attend to all five simultaneously.

Embodiment is not merely the physical body that carries the disease. It is the entire history of physical practice and attunement through which a person’s body has become capable of or resistant to particular forms of action. The insulin-dependent diabetic’s capacity to manage their condition is partly embodied in the habitual patterns of eating, monitoring, and injecting that have or have not been integrated into their physical routines over years of practice. No questionnaire about confidence captures this. No collective efficacy score reflects it. It requires attention to the specific trajectory of embodied practice that has produced the current configuration.

Being-with is not social support as an additive variable. It is the specific recursive logic of the patient’s actual relationships: with healthcare providers who may be attentive or dismissive, with family members who may facilitate or undermine illness management, with peers who may model effective practice or normalise its absence, with communities of shared condition that may generate solidarity or stigma. The interactive quality of these relationships , how each party’s behaviour shapes and is shaped by the others’ , determines what is possible within them. Self-efficacy questionnaires and social capital surveys both flatten this recursive texture into aggregate scores. The flattening is not a pragmatic simplification. It is a loss of the information that would allow the framework to explain its own findings.

Dwelling is the geographic, architectural, and seasonal environment within which health behaviour occurs. The availability of safe walking routes, green space, fresh food, and healthcare facilities are not background variables that moderate self-efficacy. They are constitutive of the capacity for health behaviour. The patient who lives in a food desert does not merely face reduced self-efficacy for healthy eating. They face a materially different configuration of possibility from the patient who lives within walking distance of a farmers’ market. These are not the same capacity constrained by different contextual factors. They are different capacities, constituted by different mediational configurations.

Multimateriality is the entire world of objects, infrastructures, technologies, and physical resources through which health behaviour is enacted. Medication that is available, affordable, and easy to administer is not merely a contextual facilitator of self-efficacy. It is part of the capacity itself. The patient’s capacity to adhere to their antiretroviral therapy is partly constituted by the reliability and affordability of the medication supply, the presence of a refrigerator, the availability of clean water, and the physical design of the pill itself. These material facts do not moderate an underlying cognitive capacity; they are among its constitutive elements.

Multisymbolization is the dimension of cultural meaning, institutional recognition, and symbolic framing within which health behaviour is understood and valued. Whether diabetes management is experienced as an expression of self-care or as a mark of personal failure; whether help-seeking for mental health problems is understood as strength or as weakness; whether HIV treatment adherence is supported or undermined by the stigma attached to the diagnosis: these symbolic conditions are not merely attitudinal variables that affect self-efficacy. They are part of the mediational configuration within which the capacity for health behaviour is or is not constituted.

X. Interrecursive Commitment as the Operative Mechanism

What self-efficacy theory has actually been tracking, without the conceptual resources to name it, is the effect of interrecursive commitment structures on subsequent action. When a patient in a chronic disease self-management programme rates their confidence in their capacity to manage their condition, when a group of community residents collectively affirms their neighbourhood’s capacity for coordinated action, when a person answers a social trust survey question, they are not reporting internal cognitive states. They are performing commitments that enter the social world and reorganise the recursive ecologies within which subsequent action will occur.

This is the logic of the sankalpa, the formal declaration of future intention that is found in some form across all known human cultures. The declaration does not work through magic and it does not work through internal cognitive modification. It works because it constitutes a new social fact: the future self is committed by the present self’s declaration, others who know of the commitment hold expectations based on it, and the social ecology within which subsequent action occurs is reorganised accordingly. Actions that honour the commitment are facilitated; actions that violate it generate friction, social cost, and identity consequence. The declaration is not a representation of a pre-existing state; it is a performance that creates a new state.

This account preserves everything that is empirically robust in the self-efficacy tradition while dissolving its characteristic theoretical confusions. It explains why self-efficacy questionnaires predict better when completed immediately before performance: the declaration is a more powerful commitment when closer to the moment of action. It explains why they predict better in public contexts: public declarations constitute stronger social facts with greater recursive consequences. It explains why the effects are stronger in interrecursive domains: it is precisely in these domains that the declaration can restructure the social ecology through which performance occurs. It explains why collective efficacy effects exist but are not simply self-efficacy effects aggregated upward: collective declarations constitute a different kind of social fact from individual ones, with different recursive properties.

Crucially, this account changes the design logic for health interventions. If the mechanism is interrecursive commitment rather than internal cognitive modification, the question is not how to raise a patient’s confidence score but how to create conditions under which meaningful commitments can be made and honoured. This requires attention to the social context in which commitments are made (who witnesses them, what authority they carry, what community of mutual accountability they enter), the temporal structure of commitment (when it is made relative to the performance moment, how it is renewed, how progress is recognised), the material conditions that make commitment-honouring possible (is the medication available, is the food affordable, is the space safe), and the symbolic frameworks that give commitment its weight (is illness management understood as a valued activity within the relevant cultural context).

XI. What Adequate Health Interventions Actually Do

The most effective health interventions in the literature are, on careful examination, not self-efficacy interventions. They are interrecursive commitment interventions that have been misunderstood as self-efficacy interventions because the field lacked the conceptual vocabulary to describe what they were actually doing.

Alcoholics Anonymous, consistently among the most effective long-term recovery supports for alcohol use disorder, works through the formal public declaration of sobriety before a witnessing community, the renewal of that declaration at each meeting, the recursive accountability created by ongoing group membership, the sponsorship relation that embeds each individual’s commitment in an interrecursive structure with another person, and the symbolic framework that gives sobriety its cultural significance within the community. None of this is explained by changes in individual self-efficacy scores. All of it is explicable as interrecursive commitment operating across the five mediations. The meeting is an embodied event in a specific dwelling, constituted through specific material arrangements, organised by a specific symbolic tradition, and effective through specific relations of mutual accountability and recognition.

The Chronic Disease Self-Management Programme works not because it installs confident beliefs but because it creates a group of people who have declared intentions to one another, who have made those declarations in the presence of peers with similar conditions, who have established recursive accountability with those peers over multiple sessions, and who have been given embodied practice at the specific skills the commitments require. The self-efficacy scores rise as a by-product of this process, not as its cause. The interventions that would follow from a genuine understanding of the mechanism would look quite different from those that follow from the self-efficacy account: they would prioritise the quality of the social witnessing, the durability of the commitment ecology, the material conditions that make commitment-honouring possible, and the symbolic frameworks that give the commitments their weight. They would not prioritise Likert-scale scores.

Continuity of care , the sustained relationship between a patient and a specific healthcare provider over time , is among the most powerful predictors of health outcomes across virtually all chronic conditions. The self-efficacy framework has no good account of why this is so. The LVT account is immediate: continuity of care creates an interrecursive structure in which the patient’s commitments are held by a specific other who has the authority, the knowledge, and the ongoing relationship to recognise, affirm, and hold the patient accountable. The provider who has known the patient for five years inhabits a different recursive ecology with that patient than the provider who is meeting them for the first time. The capacity for effective illness management is partly constituted by this recursive history. It cannot be captured by measuring either party’s confidence.

The therapeutic relationship in chronic disease care is, on the LVT account, a sustained interrecursive commitment structure within which specific health behaviours are held, witnessed, and renewed over time. The general practitioner who asks at each visit not only about medication adherence but about the circumstances of daily life , the quality of sleep, the stability of housing, the pressures of work, the state of close relationships , is attending, however informally, to the mediational configuration within which illness management occurs. This attention is not merely supportive. It is diagnostically essential, because it is the configuration and not the confidence score that determines what is possible. The policy environments that have fragmented primary care, shortened consultations, and disrupted continuity of care have not merely affected patient satisfaction. They have degraded the interrecursive commitment structures through which effective chronic disease management is produced. No amount of self-efficacy intervention can compensate for the destruction of the recursive ecology within which those interventions could have their effect.

XII. Conclusion: Beyond the Container

The argument of this article can be stated simply: self-efficacy, collective efficacy, and social capital are three versions of the same mistake. They assume that the determinant of health behaviour is a property of a container , the individual’s cognitive interior, the group’s shared beliefs, the community’s relational stock , and they build measurement and intervention frameworks around this assumption. The assumption is false. The determinant of health behaviour is not a property of any container. It is a feature of the interrecursive configuration of embodied, social, material, and symbolic life through which action is coordinated. This configuration cannot be accessed through questionnaires about confidence, measured through survey items about community trust, or modified through programmes designed to raise scores.

None of this means that what these frameworks have discovered is worthless. The regularity that Bandura documented , that explicit forward-oriented declarations predict subsequent performance , is real and important. The epidemiological evidence that social connectedness predicts health outcomes is real and important. The observation that community-level conditions shape individual health behaviour is real and important. What is not real, and what has had damaging consequences, is the theoretical infrastructure built around these regularities: the cognitive mechanism, the capital metaphor, the container ontology, the individual-society binary.

Health research needs a framework adequate to what health behaviour actually is: a distributed, interrecursive, materially and symbolically constituted achievement that cannot be located inside any individual or any group but that emerges from the specific configurations of coordinated life through which people manage illness, maintain health, and recover from disease. Living Value Theory offers the beginnings of such a framework. The five mediations provide a vocabulary for describing what matters without collapsing it into containers. The three recursivity domains provide a basis for distinguishing mechanisms that operate quite differently and that have been systematically confused under unified theoretical frameworks. Interrecursive commitment provides an account of what self-efficacy research has actually been tracking: the coordinative effects of formal declarations of future orientation made in socially significant contexts.

The practical research agenda that follows from this reorientation is substantially different from what the field has pursued. It requires methods adequate to the phenomenon: sustained qualitative attention to the recursive configurations of actual health worlds, rather than survey instruments that reduce those configurations to scores. It requires intervention designs that attend to the mediational field rather than to the cognitive interior: programmes that restructure the material conditions, social relationships, symbolic frameworks, and commitment ecologies within which health behaviour occurs, rather than programmes designed to raise confidence scores. It requires evaluation frameworks that can assess whether a mediational configuration has been improved, not merely whether a questionnaire score has moved. And it requires an honest reckoning with the political economy of the frameworks it would replace: self-efficacy, collective efficacy, and social capital have all been institutionally sustained not merely by their intellectual merits but by their compatibility with governance frameworks that locate the source of poor health in individuals and communities rather than in the structural and material conditions of their lives. An adequate framework for health behaviour will be less convenient for funders who prefer not to engage with structural determinants. This is a feature, not a defect.

The field has spent half a century measuring the shadow of a ritual performance and calling it a cognitive mechanism. The health of the people it studies requires something more accurate. The last thing that Molly needs is to be schooled in “better self-efficacy.”

For a longer discussion of Molly’s case, see Ecks, S. (2021), Depression, Deprivation, and Dysbiosis: Polyiatrogenesis in Multiple Chronic Illnesses. Cult Med Psychiatry 45, 507-524. https://doi.org/10.1007/s11013-020-09699-x