Keywords

Introduction: A Broken Leg and a Stranger

You have fallen and broken your leg on the stairs outside your apartment. Your phone is inside and you’re immobile, in agony. In that moment a stranger walks up, seemingly concerned. You look them in the eyes. They seem kind. You take a chance and ask them if they can quickly go up to your apartment, grab your phone from the coffee table, and call the emergency line. They agree to help, and you breathe a sigh of relief. In the face of your limitations in that moment, you have made a very specific wager: that they will not defy your prediction of their behavior, and particularly, that they will not do so in a way that runs counter to your interests. In other words, not only are you wagering that they will not surprise you by donning a helmet and begin singing opera, you are specifically wagering that they will not betray you by taking advantage of the situation, e.g., stealing everything and walking away. You have decided to trust them.

Trust, this bet on predictable benevolence, is a social heuristic (another word for shortcut) that enables us to navigate a world about which we have limited direct information and within which we have limited agency. Trust is a leap of faith across gaps in information, reducing the energy we would otherwise spend independently verifying others’ beliefs, intentions, and actions (Braynov, 2002), and performing all the actions necessary for our survival alone. Trust allows us to plug others’ hypotheses into the gaps in our information and behave in belief and in action as if those hypotheses are accurate. Doing so in no way ensures that the information is accurate, so trust is ultimately a wager, a simulation of the world that we treat as real. This reduces the energy cost associated with uncertainty, facilitating cooperation, community, and group efficiency (Lewis & Weigert, 1985; Luhmann, 1979). Due to its profound material and subjective advantages, it can be considered a form of social capital (Bachmann, 2001; Morgan & Hunt, 1994; Fukuyama, 1996; Zheng et al., 2008). By virtue of these multiple capacities, it is a foundational pillar of human interaction, ranging from pairs of people (i.e., dyads), to families and societies, all the way up to the global web of socioeconomic relations that undergirds our civilization (Misztal, 1996; Zak & Knack, 2001).

In this chapter, we will dissect the principal components of trust, the role of trust in reducing individual prediction error and group complexity (Lewis & Weigert, 1985) in and through social interaction (cf. Bolis & Schilbach, 2020a; Ramstead et al., 2018). More concretely, we will link trust to fundamental properties of predictive processing and homeostasis and, in doing so, formalize and situate our framework within the contemporary theoretical landscape. Last, we will discuss the potential maladaptive outcomes of trust formation and maintenance and possible insights on how to avoid them. This constitutes a novel framework with which to understand and study trust empirically, relating it inwards to phenomenology, cognition and affect, and outwards to informational and energetic properties of groups, within a global framework of homeostasis and free energy minimization (Table 8.1).

Table 8.1 Glossary of terms
Full size table

Conceptions and Components of Trust

Classical research on trust describes its cognitive, affective, and behavioral components, while primarily approaching it via two core accounts: from the psychological perspective, the disposition to trust is conceived as a trait difference dependent on properties of the other, such as honesty, status, benevolence, etc. From the behavioral perspective, trust is modeled as a risky but advantageous wager on future reciprocity, primarily studied through a small set of paradigms such as the prisoner’s dilemma and the trust game (reviewed in Lewis & Weigert, 1985). Trust can reduced to the following: one agent (the truster) engages in a belief about a future outcome that relies on the behavior of the other (the trustee). This may be voluntary (as when you decide to give the keys to your apartment to a friend) or compelled (as in the case of your broken leg) (Bamberger, 2010; Mayer et al., 1995; McKnight & Chervany, 1996).

From the individual perspective, trust is essentially regarded as a belief in the predictability of a future outcome, whether in the cognitive and social sciences (Lewis & Weigert, 1985), in management science and business (Cui et al., 2018; da Rosa Pulga et al., 2019), or in law. Predictability here means you have well estimated the parts you can “model” and the parts that are chaotic or unpredictable, e.g. the explanatory variables and random error. Predictions about others constitute a cardinal process in social interaction on both an intrapersonal and interpersonal level (Frith & Frith, 2012; Timmermans et al., 2012). Conversely, making oneself predictable, and thus facilitating trust formation, one can help increase the chances of continuing to interact with others (Coan, 2015), while collectively decreasing metabolic cost (Theriault et al., 2021). We don’t just trust those we find predictable, we also seek to make ourselves more predictable to those we trust.

However, predictability alone is not sufficient to capture what we mean when we say “I trust you.” I can trust that person X will betray me if I let down my guard. Person X in this account is predictable, but I cannot say that I trust them if I am sure that they will betray me given the opportunity. To say that I trust them (rather than trust that they will do as expected) conveys a belief that they are predictably benevolent. Philosophers have addressed this by distinguishing trust from reliance, where trust can be betrayed while reliance can only be disappointed (Baier, 1986, 235). I may rely on a clock to give the time, but I do not generally feel betrayed when it breaks. Trust is thus different from reliance in that a truster accepts the risk of being betrayed. In contrast, I can feel confident in the competence of another without being invested in their benevolence (Nooteboom, 2017).

In addition to predictability and benevolence, we argue that trust requires the trustor’s implicit belief that the trustee’s conception of benevolence and perception of reality are compatible with their own. I can find a person’s behavior both predictable and a clear expression of their benevolence according to their worldview but still not trust them because that expression of benevolence is incompatible with mine. Take the case of a fanatic who offers to ceremoniously sacrifice me in order to reunite me with my creator. I may believe that they are sincerely benevolent, but I cannot say that I trust them, because human sacrifice is not consistent with my concept of benevolence. In addition, benevolence is often expressed through the transfer of accurate information. Trust requires that we be able to generally believe another’s assertions about the world. Take the case of a person or group of persons significantly deviating from an intersubjective consensus in a given spatiotemporal and, thus, sociocultural context, e.g., a schizophrenic or a person under the influence of strong hallucinogens. In addition to qualms about their benevolence, I cannot use the information they provide me about the world, unless I have access to their perspective of reality. Thus, trust in our estimation entails predictability of benevolence within an interpersonally compatible conception of reality, including one’s values within that reality. In this light, trust is fundamentally a dialectical relationship between individual properties (cf. predictable benevolence) and collective properties (cf. shared reality and values).

The formation of social bonds whether they be romantic, friendly, or transactional relies on the formation and building of trust. Not surprisingly, evidence suggests that the ability to trust and the state of trusting are not only beneficial to the survival of an organism, they are also important for subjective well-being (DeNeve & Harris, 1998). To feel trusted and to trust others feels good, fosters calm, and is important for healthy, warm relationships (DeNeve, 1999). At the other extreme, the betrayal of trust is deeply traumatic and difficult to recover from, whether in relationships or societies (Lewis & Weigert, 1985). In Dante’s Inferno, the lowest circle of hell was reserved for traitors. For what greater sin exists within the human mind and human mythology than betrayal? The emotional experience of being betrayed is deeply traumatic and difficult to recover from, as it constitutes a massive prediction error in the human psyche, a defiance not only of current predictions but of an entire history of belief. Indeed, the powerful subjective experience of trust and its betrayal, and this experience’s ubiquity and power within human art and mythology, points to its pivotal role as a group-level homeostatic function.

Despite manifesting as a compelling component of individual experience, trust is an inherently social construct (Lewis & Weigert, 1985; Searle, 1995), much like language, power, surveillance, and accountability (Gerck, 1998a, b). Note here that this extends to trust in one’s concept of self as the third person, e.g., “I don’t trust myself to drive drunk.” Trust acts as a group-level adaptive mechanism that makes social life more predictable and less dangerous, thereby facilitating coordination (Lewis & Weigert, 1985; Tomasello, 2014). Inversely, the absence of trust inhibits the formation of community, impedes cooperation and in doing so increases instability and entropy within the group, reducing its efficiency (Braynov & Sandholm, 2002; Lewis and Wiegert, 1985; Zak & Knack, 2001). Along similar lines of thought, modern economics considers trust as an economic lubricant, reducing the cost of transactions between parties, minimizing risk and uncertainty, as well as enabling new forms of cooperation and, at the macro level, generally facilitating business, with hypothesized macroeconomic effects even on indices such as GDP and inflation (Morgan & Hunt, 1994; Singh, 2012; Zheng et al., 2008).

Let us imagine an illustrative example of trust’s role in groups (cf. Tomasello, 2014): In a primitive world where relatively small prey was abundant, each individual anthropoid could hunt and take care of their satiation needs independently. Imagine now an abrupt ecological shift after which small prey (e.g., chickens) has been substituted by significantly bigger prey (e.g., bears), which is impossible to hunt individually. In this case, individuals have two options: either they continue going independently for the limited small prey or learn to coordinate in order to effectively catch the bigger prey. One such strategy could have been found in an individual scaring a bear which, trying to escape, falls in the trap of several other anthropoids. Here, crucially, the first anthropoid in our hypothetical scenario should trust the group of the others will provide a fair share of the food at the end of the day.

At a certain point of evolutionary history, humanoids were potentially presented with a fundamental dilemma of trust: I either act independently, risking the unavoidable case of running out of suitable prey; or trust and coordinate with others in order to survive both as an individual and as a group, risking being potentially betrayed. However, the heuristics we use to establish trust, such as perceived similarity, group affiliation, or charismatic persuasion, can result in emergent, maladaptive outcomes. We may trust groups whose behavior, while coordinated, may be ultimately detrimental to ourselves and others or groups whose own internal dynamics may be ultimately self-destructive. Indeed, throughout the evolution of organisms and superorganisms (Kesebir, 2012) of greater and greater complexity, there must have existed a dynamic balance between the opposing needs to minimize prediction error by trust and, on the other hand, to withhold and restrict trust because of the high risk it implies.

The Bayesian perspective construes the brain as an organ that calculates and maintains expectations about subsequent events in the environment or within the body, by combining prior experience (priors for short) and newly sensed or posterior information. Crucially, the more confidence (i.e., precision) is placed on the validity of existing prior expectations/beliefs; the less these are updated in the face of new incoming information (i.e., evidence). Trust in this light operates primarily upon gaps in information or points of uncertainty, allowing us to place high confidence in one’s (or trusted others’) priors in the absence of data. After all, if one possessed absolute knowledge, trust would be unnecessary.

Trust reduces the energy used in mitigating uncertainty around incomplete information and lessens the impact of conflicting data. Trust can thus be viewed as a complementary mechanism of attention. Attention putatively allows for reallocation of monitoring through selectively keeping precision (confidence) of incoming information high or low, in order to attend or not, respectively (cf. Mirza et al., 2019; Friston, 2009). Trust acts by reallocating attention as if prediction error was low (Fig. 8.1), through the selective tuning of precision/confidence toward the trustee, resulting in a reduction of energy consumption. Trust is a wager driving informational confidence in prior beliefs about blind or occluded spots in the social and material world—the unseen priors driving others’ observed behavior and our priors about the unseen world that trusted others provide.

Fig. 8.1
An illustration of a human face explains the prediction error and the interpersonal sharing of priors.

Trust minimizes apparent prediction error and facilitates the interpersonal sharing of priors

Full size image

Let us unpack this core idea intuitively. When introducing their Bayesian model of selective attention, Birza and colleagues describe the example of the lost red pen (Mirza et al., 2019). Imagine you are at your office and you have lost a red pen. How will you search for it? Or to put it more technically, how will you sample the environment? A naive robot with high luxury of time might choose to deploy a serial search, sequentially scanning all possible positions in the room until the red pen is found. However, in real life, humans deploy certain heuristics in order to optimize their sensorimotor processes in space and time and eventually the chances of maintaining their own existence. Searching for a red pen, arguably, might not be crucial for one’s own life these days, yet quickly spotting the red apple on a tree and, thus, avoiding eating a poisonous fruit or encountering a wild predator might have been a life-saving process in prehistoric times. In this scheme, attention modulates the expected precision (confidence), so that task-irrelevant observations have less expected information gain, resulting in an agent less motivated to actively seek for them (Mirza et al., 2019).

In this light, the features of the environment employed in searching for the red apple have been suitably modulated in order to fit the purposes of the search. By selectively increasing the precision of features like the redness or the curving shape of the apple, these features will become more salient in the belief updating process, as high precision (confidence) information weighs more in Bayesian information integration.

Now imagine a stranger offers you a fruit that looks like an apple, assuring you it is delicious. Will you eat that fruit? A naive approach implying ample luxury of time may include meticulously researching the stranger’s past and slowly forming an understanding of their feelings, intentions, and beliefs. However, in real life, one is typically forced to make a decision quickly based on incomplete information. If you choose to trust this person, you choose to consider their offer as a predictable, benevolent action from a person sharing similar values with you, and as such not worthy of further scrutiny. You will take the apple and skip the time and energy necessary for verification of its nature. Technically speaking, one is choosing to increase the confidence of the priors about the trustee, allowing one to selectively disregard or weigh less other types of (even contradictory) information in a context-specific manner, regarding the trustee or the information coming from them, as well as the specific goals and contingencies. This wager saves energy but carries the risk of both being incorrect and not learning from it, with potentially disastrous consequences. In summary, we construe trust as a selective increase of prior beliefs’ precision/confidence about the trustee predicated on a wager on the predictability, benevolence and interpersonal similarity of the trustee.

The Experience and Function of Trust: Affect and Homeostasis

Think of someone you trust. What is trusting them like? What does it mean for your relationship with them? Who has not experienced the deep comfort of feeling trust in another human being? Or the heartbreaking, scarring pain that comes from experiencing trust betrayed? Trust’s affective content is undeniable in daily life (Lewis & Weigert, 1985). Classical accounts of trust conceive of trust in terms of individual predispositions toward trusting, judgments based on qualities of the trustee, and as a rational cost-benefit analysis, in which trust constitutes a risky but advantageous wager on future benevolence (Lewis & Weigert, 1985). These accounts, however, largely overlook the subjective experience of trust and do not analyze what the functional role of this affective component may be, or from where it may arise.

Contemporary theories of affect posit that feelings are the extrusion into conscious experience of homeostatic processes that arise from the interaction of the body, brain, and exterior milieu (Damasio, 2018; Damasio & Carvalho, 2013). Homeostasis is here not taken to mean a stable unchanging state but rather a dynamic process aimed at minimizing the prediction error or disconnect between one’s expectations of the world and one’s body, and information derived from the world and one’s body (cf. Seth, Suzuki, & Critchley, 2012). The feelings associated with a process such as trust (especially their valence and intensity) are informative in understanding the importance and role of that process. The intensity of trust’s formation and betrayal suggests from a homeostatic perspective that there is something so profoundly useful about trusting relationships that we are evolved to seek, enjoy, foster, and preserve them. Indeed, the powerful feelings associated with trust (cf. Lewis & Weigert, 1985) drive correspondingly striking patterns of behavior and belief: if you trust someone in the extreme, you assume that they will act in your interests despite ample motivation to do otherwise. You assume that they are not misleading you even if the content of their words is unbelievable and assume that there is a rhyme and reason to their actions even if they appear nonsensical or deeply unethical.

Correspondingly, the neural correlates of trust appear related to affective processing in general. Judgments of trustworthiness involve a broad array of affective decision-making regions such as the anterior cingulate, frontal lobe, caudate, insula, and amygdala (Todorov et al., 2008; Watabe et al., 2011; Winston et al., 2002). The amygdala has been thought to be hyperactive in response to specifically untrustworthy faces (Adolphs et al., 1998; Baron et al., 2011; Santos et al., 2016) although some data suggests the amygdala may actually have a non-linear response to the strength of the trustworthiness judgment (Freeman et al., 2014; Said et al., 2009). The assessment of trustworthiness is associated with amygdalar activity (signaling trust or mistrust), as well as connectivity with and activation of regions associated with top-down control of affect like dorsolateral prefrontal cortex, temporoparietal junction, and ventromedial prefrontal cortex (Bellucci et al., 2019).

The maintenance and formation of unconditional trust relationships is associated with activation in mesolimbic reward systems associated with encouraging and reinforcing optimal behaviors (Krueger et al., 2007). When choosing to trust and interacting with a trustworthy individual, activity heightens in the orbitofrontal cortex and caudate, though this activation decreases with age suggesting that trust and cooperation become more of a given than a novel reward (Decety et al., 2004; Fett et al., 2014; Gromann et al., 2014). It is also thought that affective regions such as the insula and anterior cingulate may be hyper-responsive when there is a possibility of betrayal, reflecting the strong importance of avoiding misplacing trust (Aimone et al., 2014; Fett et al., 2014).

As social animals, we need trust because it is simply impossible to experience the whole of our social and material world directly. Much of our perception of the world relies on others (cf. participatory sense making De Jaegher & Di Paolo, 2007) and dialectical attunement (Bolis & Schilbach, 2020a, Vygotsky, 1978). Hence, from the perspective of a homeostatic organism that seeks to minimize prediction error about the world, what could be better than having an affective marker of the validity of one’s models of others? Individuals within trusting (and hence coordinated) groups expend radically less energy in monitoring the world, getting information and verifying it, predicting the behavior of others, and ensuring their benevolence. Conversely, high unpredictability and a high possibility of malevolence or betrayal within our immediate environment is energy-intensive to monitor and manage, detrimental to homeostasis, highly suboptimal for group efficiency, and correspondingly regarded as both stressful and unpleasant (DeNeve, 1999; DeNeve & Harris, 1998; Singh, 2012).

The affective nature of trust supports a crucial role for it in the formation and maintenance of efficient, coordinated groups, from dyads to civilizations (Lewis & Weigert, 1985; Zak & Knack, 2001). Much as the theory of embodied cognition speaks of the brain’s cognitive processes as situated within the body (Wilson & Foglia, 2011), we cannot disregard that as the brain exists within the body, the body exists and grows within a society in virtually every instance in our evolution. It has been theorized that we “extend” our cognition via external tools (Clark & Chalmers, 1998), such as our phones. Perhaps our cognition, knowledge, and affect are also extended within our trusted groups. Our knowledge of the world, our reactions to it, and our motivations within it are informed by others we trust, as such trusted conspecifics are able to compensate for incomplete data and uncertainty, thus achieving coordinated and potentially optimal knowledge, motivation, and affective states that emerge between individuals. The wager of trust consists in our choice of who to include in this collective entity, and thereby our window to the unseen and occluded world.

If our approach to trust is correct, then the social pain of betrayal may reflect more than individual homeostasis: it reflects “damage” to the collective, a reflection into the individual of group homeostatic signals: When I trust you, I’m in a sense saying “Let’s form a superorganism, a collective entity, together.” This may be why violations of trust are deeply traumatic and have such an enduring impact on subsequent behavior and inference. When betrayed we are saying: “We were a superorganism together. I did not second guess your inferences about the world, or your future behavior towards me, and it was great. As it turns out, I was wrong. My choice of superorganism and the investment it implied was incorrect. What a surprising disappointment.” The intensity and valence of the affective response when trust is betrayed suggests that it constitutes a massive prediction error that must be incorporated into behavior and decision making, and remembered about that person, group or institution, lest we risk future negative consequences. Via this bridge between phenomenology and individual and collective forms of homeostasis, we can draw novel links between the subjective experience of trust and predictive theories surrounding cognition.

Trust as Dialectical Bridge Between Individual Prediction Error and Group Complexity

We posit that trust is a characteristic example of a phenomenon bridging individual accounts of cognition that emphasize the minimization of prediction error (Clark, 2013) and group accounts of efficiency and coordination that emphasize the reduction of complexity (Lewis & Weigert, 1985), unified by the principle of free energy (cf. Bolis & schilbach, 2020a; Ramstead et al., 2018) (Fig. 8.2). Leaning on dialectical (e.g., Vygotsky, 1978) and Bayesian (e.g., Clark, 2013; Friston, 2013) accounts of cognition and action, we also regard trust as a core process of dialectical attunement. Dialectical attunement construes human becoming as the dynamic interplay between (social) internalization and (collective) externalization in and through culturally mediated social interaction (Bolis, 2020; Bolis & Schilbach, 2020a). Here, internalization is thought of as the co-construction of bodily hierarchical models of the (social) world and the organism. Externalization is taken as the collective transformation of the world.

Fig. 8.2
An illustration of a human face explains the low trust and high trust group behavior on the left and right, respectively.

At a collective level, trust facilitates efficient group behavior, reducing group complexity and overall energy consumption

Full size image

Our notion of internalization is largely based on predictive processing conceptualizations (Bolis, 2020; cf. Friston, 2013; Clark, 2013). Predictive processing has been defined as a hierarchical bidirectional process through which an organism adjusts itself in order to “optimally” predict environmental and bodily regularities. In brain function, predictions are continuously generated and propagated from higher levels of the neural hierarchy to lower ones in an attempt to explain away prediction errors, i.e., the discrepancy between incoming information and generated predictions. On the other hand, prediction errors are propagated from lower levels of the hierarchy to higher ones in order to suitably readjust the organism. The ultimate goal of such a process is to minimize prediction error as precisely as possible, through processes such as perception and learning. Such hierarchical structures should be considered as collectively shaped. First, we dynamically embody others in and through social interaction, shaping each other’s hierarchical structure (Bolis, 2020; Bolis & Schilbach, 2020a), and second, such structures might even be socially extended into interbodily configurations (Ramstead et al., 2018). The structure and culture of social groups are two possible avenues to achieve such configurations.

Yet, organisms such as humans are not passive observers of reality who merely try to adapt to it (cf. Bolis, 2020; Bolis & Schilbach, 2020a). Organisms continuously interact with their world (including their own body), adjusting it according to their prior expectations (cf. Friston, 2013; Clark, 2013). For instance, the body temperature tends to fall behind expected values in extremely cold environments. Bodily tremor, lighting a fire, or choosing to go in a warm space typically reverses such a bodily temperature decrease, helping keep it within well-defined bounds. Processes of actively controlling the body and the environment in order to minimize prediction error have been described as active inference (Friston, 2013; see also Clark, 2013). However, such processes should not be exclusively attributed to the individual. For instance, “architecture and technology can be viewed as a collective effort for reducing overall uncertainty by transforming the environment according to bodily and interpersonal expectations” (Bolis, 2020). In a nutshell, humans actively co-construct and co-regulate—in interaction with other organisms—their ecosocial niches, with the ultimate aim to facilitate survival of not only the individual but also the social group and the species as a whole (cf. dialectical attunement). Here we suggest that multiscale processes of trust between individuals and groups are cardinal to such an endeavor.

Such processes of prediction error minimization can be thought of as processes of complexity reduction subserving life. According to the free energy principle theory (cf. Friston, 2013), life is thought of as a natural process leading to a restricted number of states. For instance, a human being, conceptualized here as a system, typically inhabits a well-defined range of states across several dimensions, such as temperature, size, and body structure. A human corporeal system maintains an order of a certain extent. Such a process of life implies a tendency to resist the second law of thermodynamics of keeping disorder (entropy) as low as possible. As entropy can be mathematically defined as the mean value of surprise over time, a living system needs to also keep surprise as low as possible. However, the precise calculation of surprise is not accessible to a living system, as it should be aware of the dynamics of all possible states of a given world. Therefore, in practice, an upper bound of surprise (i.e., free energy), as opposed to the exact value of surprise, is kept as low as possible. In turn, free energy minimization can be cast as prediction error minimization under simplifying assumptions. Taken together, according to the free energy theory, a living organism, such as a human being, achieves staying alive via effectively minimizing overall prediction error.

Similarly, trust can be viewed as a multiscale process of dialectical attunement via interpersonal prediction error minimization. Put simply, the trustor outsources a part of their active inference and predictive processing to the trustee. By assuming predictability and benevolence of another, within a compatible set of values and worldview, the trustor gives up control over the actions of the trustee. In so doing, it inextricably links interindividual processes of making sense and controlling the world, allowing for potential complexity reduction in broader scales, as well as within each individual (Fig. 8.2). Of course, trust also entails the risk of severe increase of complexity across various scales (from harm of the trustor to disorder of the group) in the case of betrayal or breakdown of trust.

Trust does not solely allow us to minimize uncertainty about conceptions and models of other people, groups, and institutions, it also mediates the extent to which we reduce uncertainty about the world at large and ourselves, by mediating the extent to which we internalize the predictive models and world views of other individuals, groups, and institutions. Trust mediates not only our view of others but also the extent to which we accept them as windows into a world that we can never see and understand in its entirety for ourselves. In other words, within a group, trust operates as a gate on the intergroup sharing of priors as well as a gate on the level of precision applied to those priors. We wish to emphasize that the increased coordination and reduced complexity afforded by intergroup trust lies separate or orthogonal to the potential maladaptive outcomes of that group coordination. Take for example, the collective suicides of doomsday cults. You may well argue that they are capable of incredibly effective coordinated behavior, but the outcome of that coordination and integration is clearly maladaptive. This has applications to understanding both the phenomenon of groups centered around charismatic leaders as well as emergent networks of trust like those that emerge in social media echo chambers.

The Formation and Maintenance of Trust: Bottom-Up and Top-Down Interactions

Despite the risk to material and psychological benefits it entails (Lewis & Weigert, 1985), trust is often established quickly and without thorough verification due to the frequent interactions necessary for social life (particularly in large societies). As trust ultimately lies in a probabilistic belief about the other, not in certainty, there is a question of how much we can “trust trust” (Gambetta, 2000). We often have to rely on explicit and implicit heuristics that allow us to quickly attribute trust that emerge in the intersection of top-down and bottom-up processes, all of which hinge on perceived similarity.

From the top-down, explicit direction, the predominant marker of trust is group identity as manifested in markers of group affiliation (Platow et al., 2012) and stereotypes (Foddy et al., 2009). In practice, these provide the quickest markers and indicator of trustworthiness. From the bottom-up perspective, there are both active and passive forms of implicit trust formation. On the passive end, people who are viewed as more similar to oneself are more likely to be empathized with, trusted, and vice versa (DeBruine, 2002, 2005). Conversely, similar neural responses to naturalistic stimuli in audiences predict affiliation (Parkinson et al., 2018). On the active end, enforced or created similarity, either through mimicry or through joint action and cooperation, can also create a sense of affinity that translates into trust, empathy, and subsequent benevolence (Chartrand & Bargh, 1999). Both of these paths (top-down and bottom-up) point to trust as a bridge between individual phenomenology and homeostasis, as well as group formation and maintenance.

This can be observed in the bidirectional interaction between implicit and explicit forms of coordination, whether affective, somatomotor, or behavioral: Studies have found that mimicry of behavior can create a sense of affiliation and trust, and these are in turn associated with increased behavioral, affective, and cognitive coordination (cf. “The Chameleon Effect”, Chartrand & Bargh, 1999). Conversely, neuroscience has found that subjects’ pain centers become active when observing another in pain, a phenomenon which is considered a marker of empathic concern (Reviewed in Lamm, 2011). In fact, not only my perception of the pain of the other, but even my very feeling of pain is socially modulated, being dependent on both embodied social factors and personal attachment styles (Fotopoulou & Tsakiris, 2017). Curiously, when subjects are induced to distrust a mock subject (after observing them engage in traitorous behavior), this vicarious pain response is diminished (Hein & Singer, 2008). This diminished response is also observed when observing members of a perceived out-group experience pain (Hein & Singer, 2008), suggesting that trust is a mediator of coordination, even at a basic somatomotor level, modulating the extent to which the perceived internal states of another will impact my own. Coordination is the key word here: when people trust each other and begin to resonate with each other, they do not necessarily solely mimic/imitate each other. Rather they become part of a coordinated group; they plug into a collective entity linked by common priors and shared mutual confidence, with downstream effects on their internal states. Put simply, you internalize the entities around you, and they internalize you (cf. Bolis & Schilbach, 2020a). Our brains become less independent from one another when we trust each other.

An illustrative example of extremes of trust formation and its manifestation in coordination and optimality is in the training of groups of soldiers into platoons. From the start, they are stripped of other affiliations, allegiances, and markers, and generally without explanation induced to engage in coordinated motor behavior, such as simultaneous and repetitive behavioral exertions and utterances. They also undergo similarly harrowing painful experiences and suffering. These early difficult experiences are often cited as the foundation of the later sense of affiliation and trust to an extent rarely observed in the civilian world. From the top-down, soldiers are encouraged to strip themselves of group affiliations and identities and submit to a common identity and a common goal and unifying mythology, as well as a belief in their complete interdependence and complete reliance on one another for their very survival. The result, when successful, is a degree of coordination, efficiency, and ability to quickly align and pursue common goals that are virtually unparalleled in daily life. In all instances, there is a convergence of shared experience, coordinated behavior, repeated interaction, as well as a shared worldview and identity.

Whether emerging from explicit or implicit sources, trust either consist of—or at least uses as—its prime heuristic, similarity. Perceived similarity is correlated with trust when controlling for other variables, and engaging in behavioral synchrony or coordination with others can foster a sense of affiliation and trust (DeBruine, 2002, 2005; Chartrand & Bargh, 1999). Similarity here acts as a heuristic element. By its virtue, I can make fewer extraneous assumptions about behavior, as well as develop models of others using my own priors. It should be clear that similarity in this case does not simply refer to the static, superficial identification of similarity, as in a photograph, but to the whole action set of a person, their behavioral cadence and the idiosyncrasies of their movements and expressions, which are reflective of their internal drives, preferences, and values. Perhaps through the practices of institutionalized as well as spontaneous mimicry, coordination, and joint attention, there is also fostered a sense of a shared worldview. This manifests at the cognitive level but more importantly, at the level of values, underlying drives, points of attention, reaction patterns, affective trajectories, and also less independent internal states by virtue of self-other resonance (e.g., mirroring, contagion, mimicry).

Theories on attachment revolve around trust. Whether the caregivers are mostly predictable in meeting needs leads to either trusting relationships later on or something maladaptive like the need to seek absolute security, etc. (Mikulincer, 1998). Deficits of trustfulness and the ability to form trusting relationships imply deficits in forming parts of coordinated groups, because one’s somatomotor and affective integration into group state might be inefficient. The possible exception to this trend is in affiliation with large-scale group identity in which there is less implicit face to face, bottom-up somatomotor/affective interaction, e.g., ideological camps from a computer screen. This implies that people with emotional, affective problems and trust issues interpersonally may still be able to form part of mass movements. Hence, the authoritarian loner archetype described in works such as Eric Hoffers’ The True Believer: disaffected loners or self-perceived losers/exiles who are fertile terrain for recruitment into fanatical ideologies and associated tribal identities.

Interpersonal traumas and breakdowns in relationships can be reduced to or at least related to a violation or erosion of trust (Lewis & Weigert, 1985). A current danger relevant to our subject is the breakdown in shared realities in large-scale societies, which has occurred in large part through social media algorithms through which people are “siloed” into only viewing content that already conforms and encourages their pre-existing beliefs, encouraging disparate but more importantly incommensurable realities, thus irrevocably destroying the capacity for trust. If I perceive someone to have an incompatible concept of benevolence—one so incompatible that it affects my perception of their predictability—then I am incapable of trusting them and hence incapable of forming part of coordinating groups with them, a phenomenon which you can see crystallized in the disruption of work forces, in teams, on ideological grounds and at a larger scale in the fragmentation of societies in periods of civil war.

We should not solely focus on the loss of the positive effects of trust within groups: the loss of coordination, the loss of empathy, or the loss of affiliation. We must also examine the gain of negative properties and outcomes, beyond dislike or disagreement, down to a dysfunction in empathy and associated coordination at the somatic motor and affective level. The loss of trust and affiliation may modulate our subconscious affective and somatomotor processing such that when confronted with members of a distrusted group, their suffering has a reduced power to move us and motivate sympathy and compassion. This could hypothetically make it more likely for harm to be inflicted by otherwise typical people. For instance, dehumanization of the other (an ultimate form of trust disruption) through language coupled with reinforcement of in-group coordination (through rallies, synchronized behaviors, and symbols) has been deliberately deployed for enabling subsequent mass atrocities (cf. Scarry, 1985).

Empirical Implications and Future Experiments

Due to conceptual and methodological constraints, research of trust has largely focused on either the individual or the sociological (reviewed in Lewis & Weigert, 1985). Here, we emphasize the importance of studying intrapersonal and interpersonal processes in their inherent interrelation, as they unfold during social interactions and beyond (Bolis & Schilbach, 2017, 2020a, b; Bolis, 2020; De Jeagher & Di Paolo, 2007; Dumas et al., 2014, 2020). The concept of trust as a bridge between individual prediction error and group complexity (Fig. 8.2) is intended to spur novel work. To this end, we suggest empirically studying the links between the phenomenology of trust, the behavioral and neural correlates of minimization of prediction error, as well as the complexity and efficiency at the group level. As we discussed throughout this chapter, trust can be thought of as lying at the dynamic intersection of the individual and the collective, entailing both bottom-up (somatomotor and affective forms of coordination, on one hand) and top-down (contextual and reputation-based factors) processes. Combined, trust can thus be studied as a single interconnected construct. Here, we envision a research line which will elucidate the links between trust processes across scales, extending from implicit behavior and contagion, all the way up to conscious phenomenology and further up to patterns of collective coordination in brain and behavior. In what follows, we describe an experimental framework, namely collective psychophysiology, as well as an analysis scheme, namely multi-level analysis of intersubjectivity that could help us do so (cf. Bolis & Schilbach, 2020a).

Traditionally, psychophysiology as a research paradigm has enabled the empirical investigation of the interrelation between physiological and psychological processes, offering important insights about the mechanisms at the level of the individual. However informative this kind of endeavor may have been, the inherent dynamics of social constructs, such as trust, will remain largely unexplored until dynamic interpersonal processes are systematically considered, as, for instance, (social) cognition might be fundamentally different when we really interact with others (Schilbach et al., 2013). In fact, it has been argued that the most important experience of the other stems from the archetypal situation of face-to-face social interaction, while all other cases remain mere products of it (Berger & Luckmann, 1967).

Building upon empirical frameworks of interpersonal research (e.g., Bolis & Schilbach, 2020a; Dumas, 2011; Froese et al., 2015; Koike et al., 2016; Montague et al., 2001; Schilbach et al., 2013), the paradigm of collective psychophysiology allows for the empirical investigation and systematic manipulation of real-time social interaction, across various modalities and temporal scales. To give a concrete example of such a framework, in the special case of two-person psychophysiology (Bolis & Schilbach, 2018), study participants sit opposite each other, working on tasks either individually or collectively, while being able to interact via a micro-camera communication system. Such a two-person framework allows for the monitoring and systematic manipulation of processes that lie in different levels of organization, from psychophysiology to culture. In fact, via systematically controlling the diversity of the interacting individuals across various dimensions, such as age, culture, social class, or even psychological condition, core interpersonal processes of trust can be put to the test: emerging contextual and interpersonal differences and similarities in social interactions might prove equally, or even more important than individual traits in building and maintenance relationships of trust (cf. the dialectical misattunement hypothesis; Bolis et al., 2017). For instance, it has been shown that it is the interpersonal similarity of (autistic) traits that primarily predicts friendship quality in the general population and not the traits per se (Bolis et al., 2020).

While laboratory studies typically offer excellent experimental control, collective psychophysiology of trust can (and should) be eventually examined beyond the laboratory walls, where it manifests itself, in real-world social life. Two paradigmatic cases of such scenarios could be found in pedagogy and psychotherapy. In fact, in line with pedagogical and clinical insights (cf. Bolis, 2020; Bolis & Schilbach, 2020a; Hendren & Kumagai, 2019; Koole & Tschacher, 2016; Lee, 2007; Ramseyer & Tschacher, 2011; Terrell & Terrell, 1984; Thompson et al., 2004), we suggest that the formation of trust and interpersonal attunement between the educator and the student, as well as the therapist and the patient, might be a first step of pivotal importance to an eventual educational and therapeutic success. Here, multipersonal neuroimaging and motion tracking could be deployed in order to capture the interpersonal mechanisms of real-time social interactions in classrooms and psychotherapeutic settings (cf. Bolis et al., 2017; Dikker et al., 2017; Lahnakoski et al., 2020; Tschacher et al., 2014), being complemented by digital phenotyping and interactive self-reports (cf. Bolis et al., 2020; Insel, 2017).

It may be fruitful as well to explore phenomena that exemplify our framework and its predictions in action: one possible example is placebo, in which individuals seemingly internalize not just the abstract beliefs about the properties of a drug or procedure/ritual but also the internal states that would be implied by the belief. One could possibly interpret this as the internalization of affective and homeostatic priors, which powerfully suggests that the mitigating factors in the placebo effect are the degree of trust in the physician/clinician/healer/belief system surrounding the treatment, as well as trustfulness as a trait in the patient. This merits study.

Parts of this chapter discussed trust in terms of interpersonal predictive processing and active inference. Here, we suggest moving from exclusively focusing on the isolated individual, toward a multilevel understanding of intersubjectivity and psychopathology (cf. Bolis & Schilbach, 2017, 2020a). It would be also interesting to observe the ways and extent to which the dynamic formation of trust within a group and its neural and behavioral correlates, as well as group-level measures such as efficiency and complexity. Not only observing formed groups but rather observing the formation of groups and their maintenance, and how the success or failure of this process is modulated by bottom-up and top-down processes, what their respective contributions are, and also how these factors co-vary with the emergence of group schisms and reformations, induced and spontaneous. This could be fruitfully applied in specific instances where trust mediates information flow, such as in the classroom setting. Taken together, collective psychophysiology, we suggest, appears as a promising empirical framework for studying trust, enabling both great experimental control and ecological validity (cf. Bolis, 2020; Bolis & Schilbach, 2020a).

Conclusion

To conclude, we define trust at its core as a belief and behavior in accordance with predictable benevolence in another within a compatible worldview. Classical theories and studies on trust still bear the mark of behaviorism, with little regard for social interaction dynamics and the informational depth provided by subjective experience, particularly affect, and lacking any computational account of how these relate to brain function and cognition, due also to limitations in experimental methods and conceptual commitments at the time.

Our primary contribution may be to tie the group level reduction of complexity not only to extant concepts of trust at the individual level, mainly individual differences in trustfulness, or game theoretical conceptions of trust, but deeper, to fundamental theories about cognition, specifically predictive processing and active inference. In relating levels of analysis and inquiry that encompass a conceptual space bridging cortical hierarchies, upwards/outwards to hierarchies of individuals within groups and groups within civilization, we have a tentative framework with which to potentially examine trust at all these levels separately, in relation, and simultaneously. Thus, we provide a novel synthesis and a framework which is capable of empirical implementation and moreover carries with it novel domains in which to study the phenomenon of trust and its manifestations in daily life.

How does the novel synthesis presented in this framework help us? What does it add to our knowledge beyond description? How can we perhaps mitigate the maladaptive outcomes of the heuristics we use for the formation of trust using this model and understand them better? Broadly, our model provides a unified framework and common language with which to address a massive swath of human experience, with putative scale-free properties. More specifically, it may focus awareness on the extent to which the perception of similarity or dissimilarity among people can create coordination that extends far beyond superficial measures of worldview or declarations of goodwill, down to coordinations between somatomotor and affective states.

This highlights the tremendous importance of trust for maintaining and facilitating the wellbeing of society, and the tremendous risk that comes when, due to shortsighted ambitions, leaders, and institutions undergo actions that create distrust between the components of society (such as local population and immigrants), as this goes beyond damaging goodwill to possibly forming a material, concrete antecedent to schisms in relationships, groups, societies, and civilizations. Unveiling the mechanisms of trust formation and breakdown across various domains of human life, ranging from relationships and pedagogy to marketing and politics, may help facilitate group coordination and individual well-being, but also boost immunity against the misuses of trust. Taken together, we hope this work ultimately makes starkly apparent that trust’s role and importance is difficult to overestimate.

From a broken leg and a kind stranger, we have attempted to take the reader inwards to the predictive and homeostatic processes underlying cognition and affect, and outwards, to the formation and maintenance of collective entities. In doing so, we hope we have conveyed that trust is a root fundament of the structure of human life in every form of interaction, from dyads to civilizations.