“Stress” generally refers to stimuli that evoke a defensive response—encapsulated famously for humans as “fight or flight.” While fight or flight most vividly pertains to the actions of an autonomous, individual animal, we suggest that this concept may be adapted to the defensive actions of cells, cell communities, organs, organ systems, individuals, and communities of individuals. As such, we hypothesize that when stress is sustained beyond certain individual resilience parameters, one enters a regime of learned helplessness (33), which has been connected with societally adverse behaviors such as the adherence to conspiracy theories (34).

Generally, an organism that interprets stress as an “action” occurring against it reacts in a manner analogous to fight, flight, or surrender. Given a suitable provocation (i.e., stressor), cells and unicellular organisms may express a range of defensive biochemical and biological responses, including the synthesis of various molecular compounds, structural alterations (such as encystment), and, at the extreme, apoptosis or other modes of programmed cell death (35). Below, we present a working definition of inflammation and detail the interrelationships between inflammation and stress.

In multicellular organisms, stress is transmitted and amplified by inflammatory cells and their products. Indeed, given an appropriate and sufficient stress signal, most cells of the body will mount a reaction that qualifies as inflammatory (36–38). White blood cells, for instance, are deployed en masse within the body to counter an immunogenic provocation. Whereas the chemicals generated by unicellular organisms are directed almost exclusively “outward” (compounds that alter gene expression, i.e., epigenetic actions, being an exception), those produced by multicellular organisms may exert effects directed both outward and inward. Examples of inward effects include those mediated by compounds that bind to receptors within the same body to amplify, reduce, or otherwise alter the choreography of cells or the cascade of other chemical mediators and any chemotactic effects that alter the targeted actions of dedicated immune cells (2). In higher organisms, cells initiate alarm/danger signals—also known as damage-associated molecular pattern (DAMP) molecules—that activate tissue-resident inflammatory cells to secrete inflammatory chemokines and cytokines (4, 39), which in turn stimulate the production of oxygen and nitrogen radicals and other cell toxins (35). This cascade leads to the creation of more DAMPs by the inflammatory cells as well as damage to additional parenchymal, i.e., organ-specific, cells, driving a feed-forward loop of damage → inflammation → damage at the cell/microbiota, tissue, organ, and, ultimately, whole-organism level (2, 40) ( Figure 1 ). External stressors can also trigger epigenetic effects, i.e., producing stable changes in the genome (41). For example, depression-related, stress-associated epigenetic changes have been reported in various genes, including those involved in glucocorticoid signaling, serotonergic signaling, and neurotrophins (30).

The embedded, multiscale, feed-forward character of inflammation is necessary to respond rapidly to threat or injury, possibly across generations, but is also a key driver of pathology across all scales of organization (42–44). This is coupled with pre-existing causes of inflammation in the form of acute and chronic diseases (7). The enhanced state of alarm in our society means that a large proportion of the human population is potentially predisposed to stress-induced inflammation at any given moment. Importantly, stress may be either constructive (i.e., eustress) or destructive (i.e., distress), and inflammation may be either well-regulated and enhance function or dysregulated and promote dysfunction. This realization, in turn, offers an initial basis to prioritize and manage stress responses and to differentiate between essential, constructive inflammatory responses versus dysfunctional inflammation. The aforementioned “inward versus outward” paradigm can be elaborated further inward, given that individuals can be thought of as complex “communities of the self” (45), especially when considering the microbiome (46) in addition to our diverse cells and tissues. Similarly, given the newly gained access to information broadcasting through social networks, humans have strongly amplified their ability for outward action.

We consider biological agents as comprising embedded networks, from genes and neurons to thoughts and behaviors (47). Our cells are subject to nearly constant stress, and the imperative to maintain homeostasis is thought to have led to the evolution of multiple mechanisms for sensing, responding, and adapting to both exogenous (e.g., due to microbial infection, injury, or noxious stimuli) and endogenous (metabolic) stressors throughout our lifetime; indeed, the induction of stress response mechanisms is an important feature of aging (43, 48–50). Moreover, when stressed, our commensal microbiota can become virulent and promote further pro-inflammatory stress (51). Homeostasis itself is regulated through allostatic processes (52) that adapt homeostatic control loops to internal needs and external opportunities, finding organismic stability through continuous homeostatic change. However, these regulatory loops can go awry when perturbed, such as in settings of sepsis (51, 53) or trauma/hemorrhage (54).

We propose that these same principles of stress and inflammation pertain to communities, or populations, of individuals acting in concert ( Figure 1 ). In support of this idea, the constellation of stressors, including climate change and environmental pollution, has been linked to a significant reduction in growth, survival, and microbiome diversity of plant communities (55). In the human context, a key example of a community-wide acute stress response is any version of civil strife or warfare, while a chronic one is a reduction of birth rates and life expectancy. As we extend this paradigm to the population level, we note that a stress or provocation (e.g., war or pandemic) incites a coordinated response among the component parts (e.g., civilians, politicians, individual soldiers, or healthcare workers) of a superorganism engaging with the same, small suite of response options (i.e., fight, flight, or surrender). At the biological level, individuals communicate stress to other individuals in a process called “stress contagion” (56, 57)—e.g., via pheromones (58)—and thereby likely exacerbate their inflammatory responses. We hypothesize that this is a further cause of community-wide, and possibly global, stress communication, in addition to shared exposure to infectious stressors and highly prevalent chronic inflammatory diseases ( Figure 1 ). Indeed, the hypothesis that psychopathology can be understood in terms of symptom networks provides a direct causal link between biological substrates and pathways of stress and psychological, social, and cultural pathways (47). We advance that these pathways have become perturbed pathologically not only through the aggregate effects of the modern exposome (12, 59) but also through the potential amplification of such effects via high-throughput information channels in the Internet age (60, 61).

We hypothesize that dysregulated inflammation ensues when multiscale stress response mechanisms are no longer sufficient to mitigate stress, and when the combination of internal and external stress exceeds our individual or collective stress resilience ( Figure 1 ). At baseline, the inflammatory state is constrained via multiple points of control (36) because the inflammatory response has evolved to ramp up vigorously through positive feedback to deal rapidly with relatively non-specific threats (2). In contrast, the resolution of inflammation occurs more slowly (62), suggesting that inflammatory responses are regulated by a race between the feed-forward processes that propagate self-sustaining inflammation versus inflammation resolution processes with fewer positive-feedback loops (40).

These multiscale induction and resolution feedback loops manifest within individuals both spatially and temporally (63). While inflammation is at some level inherently local (i.e., occurring in a tissue or organ), the inflammatory response can propagate across multiple tissues and manifest systemically (2, 42, 53, 54, 63), for example, in the pathology of critical illness (63–65) and other diseases. Notably, negative feedback on inflammation can manifest locally, systemically (if the inflammatory response spills over into the systemic circulation), and centrally (via neural regulatory mechanisms; see below) (6, 66, 67) ( Figure 1 ). We propose that this type of propagation—with the aforementioned causes and consequences—also applies to populations of humans responding to stress signals from other individuals and their natural and human-made environment.

We suggest that the rapid ramp-up of inflammation is the Achilles’ heel of this generally beneficial process, i.e., that the flow of inflammatory stress from the individual to the population and back is at the root of global-scale stress and its associated pathologies, including cognitive and thus behavioral dysfunction ( Figure 1 ). In complex biological systems, robustness and functional flexibility are realized through a paradoxical fragility in the so-called “constraints that deconstrain” (68, 69). In our paradigm, the tissue-embedded and inherent feed-forward nature of the inflammatory response—coupled with close interactions with the control systems of the CNS—creates both robustness and fragility. Robustness occurs via the multilayered, multiscale, dynamic, and reciprocal control mechanisms that link stress, inflammation, and neural control and behavioral function. Simultaneously, fragility occurs via dysregulated inflammation rippling rapidly throughout multiple tissues in a multiscale fashion, leading to dysfunction in multiple organs as well as dysregulated emotion, cognition, experience, and behavior—which can be transmitted between individuals.

The brain is a key substrate in this stress, inflammation, and dysfunction cascade. On one hand, the hypothalamic–pituitary–adrenal (HPA) and sympathetic–adrenal–medullary (SAM) axes link stress to brain function, including executive control (70) and the immune system (71). Indeed, there are indications of a direct link between mood disorders and elevated levels of inflammatory markers in the blood (72). In addition, the brain will drive the actions that act as stressors on the outside world, potentially setting in motion collective stress cascades in the population.

Our paradigm extends beyond a purely bottom-up perspective on stress and inflammation and includes a top-down component. For instance, alarm in the form of interpersonal threats—whether direct (73) or over long distances via digital media (22, 61, 74)—along with anxiety over the state of society, climate change, infectious disease, socioeconomic status, etc. (8, 13–15, 75), can be considered just as pro-inflammatory as explicitly biological stimuli such as severe infections or chronic illness. These interpersonal stresses propagate among individuals within and between populations, rippling inward and outward to impact global processes in multiple ways ( Figure 1 ). As in biological settings of inflammation, the repeated re-initiation of relatively low-level inflammatory responses can lower the activation threshold for positive feedback in multiple individuals within a community or society. This phenomenon, called “priming” in the cellular/molecular context of inflammation, is part of a broader preconditioning phenomenon that also includes the negative feedback process of tolerance/desensitization. At the population level, we suggest that priming leads to a state in which stress can explode across a society, even in the (apparent) absence of a single, defined pro-inflammatory stimulus, due to a diminished resolution response. The parallel process of tolerance/desensitization to ever-increasing personal and societal harm is driven by excess negative feedback (76, 77). Both processes comprise the phenomenon of preconditioning, can be modeled mathematically (78, 79), and are a key aspect of the global inflammatory paradigm advanced here.

According to our paradigm, the current intensity of stress and its multiscale transmission might be expected to translate into a comparable concentration of societal disorder and violence amplifying stress and inflammation. That this apparently does not occur (80) suggests a commensurate amplification of “controller” functions in modern societies, from the authority of science and medicine and trust in our institutions, to multinational organizations such as the United Nations and the North Atlantic Treaty Organization (NATO), to the interdependence fostered by global commerce (81). Yet all these control functions are currently being questioned and have arguably degraded. The challenge now is how humans can best prepare themselves for the emerging negative scenario: a new adversarial era of instability due to a combination of global stress and ecological collapse.

We propose that the nervous system and cognition in individuals and the forms of symbolic exchange it affords—and the conceptually parallel role that societal rules play in communities of individuals—link stress, inflammation, and individual and societal dysfunction. As noted above, higher organisms have evolved a multiscale negative feedback architecture to control inflammation, in part involving neural mechanisms. A key neural mechanism that regulates inflammation both locally and systemically involves the vagus nerve (26), with neural circuits acting to limit the degree to which inflammatory mediators are expressed following a pro-inflammatory stimulus (27, 28). Recent evidence suggests that the vagus nerve not only regulates the degree to which inflammation is induced, but also limits the spread of inflammation across tissues and organs (82). We extend this paradigm by hypothesizing that, under physiologic conditions, neural control of inflammation occurs in part via a type of “hardware abstraction layer,” a term we borrow from computer science and that refers to the logical division of code in which computer hardware is controlled through software (83). Another way to think about this is in terms of the “constraints that deconstrain”: by having evolved specific protocols and substrates (i.e., constraints), biological systems afford a broad set of a priori unknown adaptations (i.e., de-constrained) in functional expression (84, 85). We hypothesize that the neural-immunological architecture builds on central neural regulation, invoking and involving the same inflammatory cytokines that drive inflammation in peripheral tissues. By this we mean that the brain can express the same inflammatory mediators expressed in inflamed distal organs to build a whole-body map of inflammation; we hypothesize that this “central inflammation map” is a mechanism for central regulation of the inflammatory response. As an example, this hypothesis is supported by research in rodent models showing that intrabronchial administration of a bacterial-derived immunostimulant (Gram-negative endotoxin) in a manner that does not result in the systemic manifestation of inflammatory mediators results in both lung and brain expression of the key inflammatory mediator interleukin (IL)-1β (86). Conversely, injection of IL-1β into the nucleus tractus solitarius in the brainstem is sufficient to replicate the functional derangements in lung physiology induced by endotoxins (87) and the toxin bleomycin (88). Another example is the putative causal link among stress, inflammation, and hippocampus and medial prefrontal cortex dysfunction, which can be seen as the result of multiple neurotoxic processes (89). One of these processes is HPA axis dysfunction, caused by chronic stress and enhanced production of cell-mediated immune cytokines, leading to serotonin depletion, increased glutamate and oxidative stress, and reduced inhibitory control mediated by gamma-amino butyric acid (GABA), in turn resulting in cellular damage and volume reductions (90, 91).

Thus, our hypothesis integrates inflammation into how we sense and appraise reality and shape it through our actions, closing a continuous behavioral feedback loop between the individual and their physical and social world (29). Further, we suggest that this paradigm offers a mechanism whereby stress and chronic inflammation could result in dysregulated cognition and action, which unchecked could, in turn, drive psychological dysfunction at the individual level and, ultimately, drive societal dysfunction. Initial support for this hypothesis comes from the finding that individuals with neuroimmune disorders often experience “brain fog” (92) that can impair their decision-making (93). Additional support comes from the known associations of cognitive derangements with immune dysregulation (29, 30) and the direct relation between stress and memory (94)—which, however, seems robust against “everyday” stress (95).

A related point is that the cytokines expressed in the brain putatively communicate inflammation in distal organs and positively regulate their expression. Continuing with the example of IL-1β, this cytokine can induce further expression of itself (96). We can speculate that repeated bouts of inflammation, coupled with the constant activation of neural sensing and regulatory pathways in response to real or perceived stress, might underlie the reported role of IL-1β in the pathobiology of neurodegenerative conditions such as Alzheimer’s disease (97) and other aspects of cognition (29, 30), and also lead to increased sensitization to stress and inflammation pathways. Once these effects affect action control, they can spill out into the outside world.

Below, we discuss additional factors that impact stress, inflammation, and neural function, how they link individuals to their environment, and how these factors might scale from the individual to the population and back.

Sleep disturbance is one key factor that contributes to the ongoing cycle of inflammation and potential degradation of well-being and cognition. At the individual level, sleep is affected adversely by inflammation, which in turn impairs cognitive function both in the short and long term (98). Restorative sleep is highly dependent upon hormonal cycles, known generally as circadian rhythms. From a global perspective, these cycles represent an alignment of biorhythms with planetary cycles of light and darkness (99); notably, circadian rhythms have been implicated in allostasis and resilience to stress (100). While the pineal hormone, melatonin, is most prominently associated with variations in alertness and somnolence (101), its effects extend to influences on other hormones, including cortisol, catecholamines, growth hormone, leptin, and ghrelin (102, 103). Chronic inflammation is both a cause and effect of hormonal perturbations, notably persistent elevations in cortisol and catecholamines. In turn, these disturbances can foster insulin insensitivity and elevate insulin levels, thereby further impairing endocrine homeostasis (104, 105). Importantly, the disruption of circadian rhythms in the context of the lifestyle noted above is associated with further adverse impacts on metabolism (106). Finally, recent studies also connect cognitive dysfunction induced by sleep deprivation to disturbances of the gut microbiome (107), supporting the aforementioned impact of cumulative stress on communities of organisms (55) through the adverse effects of sleep dysregulation on cognition and action.

Over time, the hormonal repercussions of inflammation tend to disrupt appetite signaling and favor overconsumption of food and the growth (in both size and number) of adipocytes (108). Fatty infiltration of other cells (notably hepatocytes) also occurs, resulting in metabolic dysfunction and further inflammation (109, 110). These pathways sabotage circadian rhythmicity, impairing sleep via a hormonal mechanism (111). Excessive weight gain and adiposity may also impair sleep via a mechanical pathway, since these are linked to laxity in the posterior palate, snoring, and sleep apnea (112). Cumulatively, these factors increase stress on the organism.

The stressful effects of inflammation on sleep are reciprocal. Impaired sleep further disrupts circadian patterns, exacerbating hormonal disturbances. Increased chronic pain and psychological duress impair sleep and increase inflammatory responses, while sleep deprivation is associated with reductions in pain tolerance and psychological stress tolerance that potentiate the effects of both pain and stress. The result of these interactions is an adverse feedback loop, with inflammation disturbing sleep and sleep deprivation amplifying inflammation (104, 105, 113), setting in motion coupled multi-scale symptom pathways that amplify stress.

Our hypothesis is based on the intertwined nature of stress, inflammation, and neural dysfunction, and the propagation of these individual stressors both directly and indirectly within a population to ultimately impact societal function ( Figure 1 ). As a first step, we have designed a high-level mathematical model that incorporates the following key variables: stress (S), inflammation (M), neural control (C), healing/restoration of function (H), and intervention (I). We note that, at present, the model is not calibrated to data and is simulated using arbitrary units (AU), given that massive, multiscale data acquisition and computational analysis across all these processes is needed to test these hypotheses and understand the complex dynamics underlying this stress-inflammation system. Figure 2 depicts the basic interactions among these variables, which are translated into the mathematical model (detailed in the Appendix ). Notably, this model is relevant to two different interpretations of the interplay of stress and inflammation: an individual interpretation and a societal interpretation. Further, this framework can account for both eustress and distress.

The stress variable can comprise an unlimited number of internal and external stresses on the system. For example, internal stressors could include an unhealthy diet, physical inactivity (i.e., sedentary lifestyle), sleep deficiency, addictions, chronic illness, dysbiosis, and mental health conditions (e.g., depression, anxiety, and associated medications), or a lack of positive psychology/social connection (i.e., social isolation) (44). External stressors could include microbial infection, climate and natural environment factors, political upheaval, overcrowding, socioeconomic status, direct and digital social interactions, or other sources of digital stress (e.g., constant and easily accessible social media feeds and the news cycle). The model allows for any combination of such stress inputs and their weighting according to empirical data.

In the model, exposure to stressors triggers a varying degree of inflammatory response. At the individual level, an increase in the body’s inflammatory response causes more stress on the body, yielding a self-sustaining cycle of stress → inflammation → stress (i.e., leading to distress). Applying a societal interpretation, societal stressors trigger unrest that leads to a state of alarm and panic (i.e., “inflamed state” of a population) where societal rules are not obeyed. Further, we would argue that societal inflammation causes increased environmental stress because stressed, inflamed, and consequently cognitively impaired humans are more likely to make harmful decisions that contribute to environmental degradation, yielding a similar self-sustaining stress and inflammation cycle as generated on an individual level.

Since the CNS partially regulates the inflammatory response, the controller variable in the model is conceived as neural control at the individual level. At the societal level, the controller is a proxy of acquired societal norms and the organizations and other mechanisms that sustain these norms. The controller works to reduce and prevent an overwhelming inflammatory state abstracted at the whole-person level, just as acquired norms help reduce societal unrest ( Figure 2 ). A highly inflamed state degrades the ability of the CNS to further regulate stress and inflammation, just as citizen unrest leads to actions that degrade adherence to societal norms, rules, and laws. Inflammation promotes healing in the model, since the primary purpose of inflammation is to heal the body (e.g., from infection) and to re-establish stability that is within allostatic bounds. The process of healing provides feedback to reduce inflammation. The healing variable in the model can be interpreted as a measure of deviation from homeostatic balance. On an individual level, the healing variable in the model is linked to the level of damage or DAMPs in the system; at a societal level, as stability reemerges, the healing variable could be linked to the relaxation or removal of laws and regulations put in place to address societal disorder.

Finally, the model can also account for the effects of outside interventions that inhibit stress and inflammation, promote healing, and restore control in a healthy and sustainable range. For example, these could include changes in diet, exercise, sleep, or medication at the individual level and societal trends, governmental subsidies, and rules at the societal level.

When assessing the outcomes of all model simulations, comparing the steady state levels of stress, inflammation, controller, and healing levels with their respective baseline levels (dark blue) at time 0 will provide insight into whether stress added to the system causes detrimental outcomes (e.g., uncontrolled inflammation) in the system or if baseline is reestablished, regardless of all model variables being simulated in AUs. Once a dataset becomes available, inflammatory mediator levels and neurological function biomarkers could be used to calibrate the baseline initial conditions used in the system for a moderate level of input stress; those specific variables would be represented in appropriate units.

For this in silico exercise, the model is tested with six abstract stress factor inputs (which could correspond to any of the factors listed above) that are equally weighted concerning their impact on the overall system. Simulating the model shows how rising stress increases inflammation while impairing neural control of inflammation ( Figure 3 ). When all stress input factors are set at their reference level (f_i = 0.5), the system is considered to be in a baseline state with moderate stress and inflammation levels (see dark blue curves; units are arbitrary). Figure 3 simulates the upper and lower bounds of the six stress inputs: the red curves correspond to all high stress (all f_i = 1) and the teal curves to low stress (all f_i = 0). Here, the system is simulated in the absence of any intervention.

Combinations of varying degrees of stress factors can also be simulated. For example, an input of f ₁ = 1, with all other factors f₂ = f₃ = f₄ = f₅ = f₆ = 0.5, would show a substantial increase in stress and inflammation to the system (compared with baseline), illustrating how a single component of lifestyle stress could cause severe harm to the body. However, the model is currently predicated on the assumption that high stress in one aspect of a lifestyle or moderate stress in multiple areas will yield a similar inflammatory response. Such a model outcome could be used to show why diet and sleep habits that are perceived as relatively harmless could, in fact, be highly detrimental to an individual (or community) in a highly inflamed state, since these factors interact synergistically to induce high levels of stress.

Crucially, this modelling framework is scalable from the individual to the population. In Figure 4 , the model is used to simulate the effect of stress transmitted by others. The vertical axis plots an individual’s stress level (here varied from 0–10); the horizontal axis shows an additional model input corresponding to the stress transmitted to an individual by others. Since this is a model stress input, it varies from 0 to 1: 0 corresponds to the absence of stress transmission (shown as the region to the left of the dotted lines) and 1 denotes the greatest level of stress that can be induced by a population. In Figure 4A , total stress in the system is indicated by the color map, where red indicates the highest stress levels. The color map in Figure 4B shows the level of controller function in the system; as expected, regions of high-stress lead to regions of controller dysfunction.

An emergent feature of this model is that the controller, which acts to stabilize the individual and population, can transition to a state in which it is not only incapable of dealing with stress, but where, in fact, it drives dysfunction. We interpret negative values for the controller function as indicating a scenario in which the controller is causing a deregulation of the system, which would be manifested in harmful decision-making. The boundary of the blue region in Figure 4B corresponds to the tipping point from reversible to irreversible societal dysfunction. From here, the model can be used to estimate the number of individuals necessary to cause a system to collapse.

The modeling framework can also assess the potential impact of interventions. Currently, the intervention (variable “I”) is modelled in a way that inhibits stress and inflammation and promotes the controller and healing function. This formalism was chosen to test the broad impact of interventions, but we envision modifying this so that a specific intervention only alters one function (e.g., only the controller versus only the inflammatory response). Two different individual-level interventions are introduced in Figure 5 and Figure 6 . In Figure 5 , a lifestyle intervention—e.g., exercising moderately/vigorously a few days per week (44, 129) or cognitive behavioral therapies (24, 44), known to reduce mortality in large epidemiological studies (130)—is included on the assumption that its efficacy is approximately 30%. Figure 6 simulates a medical drug intervention assumed to be 100% effective (a clear over-estimation used merely to test the model) with an onset of therapeutic benefit observed a short time after administration, which would be consistent with the time needed for an antidepressant drug to take full effect (131, 132); other relevant types of drugs (or devices) could also be modeled with differences in kinetics. In these simulations, the drug has a faster and greater effect than the single lifestyle intervention and shows a particularly greater benefit at high baseline stress levels. However, this gap is probably too narrow if one assumes much lower (and more realistic) drug efficacy.

Inherent in our model is the concept of resilience. Figure 7 depicts the impact of increasing the resilience parameter, S_crit , slightly, in the absence of intervention. Model results are shown for the conditions of high stress, although the effect of increasing S_crit is evident at any level of stress. Increasing the resilience threshold from an arbitrary initial value of 6 to 7 reduced the system response from an extremely high level of inflammation to a much lower level (a little higher than baseline) while simultaneously increasing controller function and healing. This increased threshold corresponds to an individual or society that is sufficiently resilient to withstand a higher stress level.

The model is intended to illustrate the complex relationship among various endogenous and exogenous factors involved in stress and inflammation and its propagation in individuals and collectives. It aims to show that using a modeling approach can assist us further in analyzing, monitoring, and predicting the impact of interventions on individuals and the population at large. Given the urgent need to better understand the multi-scale nature of stress and inflammation, because of its large-scale impact on society, we argue that such models must become part of the stress and inflammation control toolbox to inform rational collective decision-making.