Facial appearance provokes personal evaluative judgments about others’ trustworthiness which is essential in daily human social interactions. Individuals always ask themselves “Should I trust this person?” whenever they encounter a stranger. That is because interpersonal trust requires one’s willingness to put themselves in a vulnerable position upon another individual’s course of action (Filkowski, Anderson, & Haas, 2016). The ability to discriminate whether someone is trustworthy or not is thought to have some survival significance since the existence of early hominids (Marzi, Righi, Ottonello, Cincotta, & Viggiano, 2014; Riedl, Mohr, Kenning, Davis, & Heekeren, 2014) and enable the beneficial engagement in cooperative reciprocal interactions (Hahn et al., 2014). Several studies have demonstrated that trustworthiness judgment occurs unintentionally and very quickly (Marzi et al, 2014; Hahn et al., 2014; Todorov, 2008; Yang, Qi, Ding, & Song, 2011), thus, it is not quite clear of the inner psycho-neural mechanisms occur during such minimal timeframe impression. This literature review is focused on the neural processing when an individual undergoes the evaluation of whether the other individual is trustworthy or not, using studies that employed functional magnetic response imaging (fMRI), event-related potentials (ERPs), electroencephalogram (EEG), Go/No-Go task, the Trust Game, and neuro-endocrinological structures (oxytocin and opioid) to explain such evaluative judgments of trustworthiness.

Evidence from Facial Appearance

When did trust initiate? Hahn et al. (2014) were interested in the initial trust level before an exchange with a stranger occurred by measuring the resting-state brain-electrical connectivity. The research team employed the trust game – which will be discussed later – and EEG data. In this study, Hahn et al. (2014) found that the initial level of trust was determined even before the game/exchange has even started. The authors posited that individual differences in initial trust level could be higher over time and heritable. On the other hand – opposing such notions – the initial trust level could be mostly influenced by an individual’s emotional state, thus, resting-state connectivity might be affected (Hahn et al., 2014). One limitation of the study was EEG could not able to identify which brain region is responsible for such resting-state dynamics in initial trust behaviors. Hahn et al (2014) suggested that the temporoparietal junction (TPJ) might be activated during investment decisions in the trust game, but further studies need to be evaluated.

Time exposure and facial evaluation. To investigate of how rapid facial evaluation on interpersonal trustworthiness is, Todorov (2008) designed a study that showed faces to participants for 100, 500, or 1000ms, and then asked them to evaluate the trustworthiness of those faces. Surprisingly, with just 33ms exposure to faces, participants could quickly identify which face was trustworthy and which one was not. This result suggested that the evaluation of facial trustworthiness is extremely efficient (Todorov, 2008; Marzi et al., 2014; Riedl et al., 2014; Yang et al., 2011).

The research team then did another follow-up with their results to see could their participants detect trustworthiness even faster than 33ms. In this experiment, eight different presentation times – from subliminal (17ms) to unlimited (no time constraint) – of facial presentations were shown to the participants. Judgment of trustworthiness for the subliminal presentation was considered to be difficult to distinguish between trustworthy and untrustworthy faces (Todorov, 2008). However, participants were able to make accurate judgments after 33ms. In the same manner, as the presentation time increased, participants increased their accuracy in making trustworthiness judgments. This pattern of accuracy stopped at 167ms and anything exceeded after this time interval did not improve the correlation (Todorov, 2008).

Time course of facial trustworthiness appraisal. Another study was conducted by Yang et al. (2011) to understand more about the early stages of perceptual processing in trustworthy and untrustworthy faces. Random faces were generated and shown to participants with shape alterations to make faces look either trustworthy or untrustworthy (Yang et al., 2011). The ERP results revealed that C1 amplitudes elicited greater in trustworthy faces in early visual processing at the occipito-parietal region, but not in untrustworthy faces (Yang et al., 2011). Scalp topography of electrophysiological response to trustworthy face was within a 400-600ms time interval (Yang et al., 2011) which indicated that the late positive component (LPC) was presented. LPC tended to demand the allocation of attention and trigger the motivation process of approach and avoidance tendencies. LPC was activated when untrustworthy faces were presented (Yang et al., 2011). The connection between LPC and the detection of untrustworthy faces indicated that an untrustworthy face conveys an avoidance tendency and therefore, neural processing would increase the level of attention than a trustworthy face (Dolcos, Sung, Argo, Flor-Henry, & Dolcos, 2013; Yang et al., 2011). LPC’s finding has indeed reconfirmed previous studies of fMRI and untrustworthiness. In specific, the amygdala (AMY) response level increased as the level of the untrustworthiness of the faces increased (Todorov, 2008).

Handshake and evaluative judgments in social interactions. Dolcos et al. (2012) were interested in the power of a handshake accompanying with approach versus avoidance of social interactions. Such evaluations are based on body language which could be influenced interpersonal impressions (Dolcos et al., 2012) and trustworthiness of whether the person is approachable or not (Todorov, 2008; Yang et al., 2011). To conduct the experiment, event-related fMRI and skin conductance analysis were both recorded while participants were watching a brief video presentation. This presentation consisted of a handshake scene followed by approach behaviors, no handshake scene followed by avoidance behaviors, or a control situation in which a cardboard of an interactor was shown (Dolcos et al., 2012).

Results showed that there were higher ratings for approach than avoidance behaviors, and a positive impact of handshake on social evaluations (Dolcos et al., 2012). Handshake was therefore perceived as optimistic behavior. It enhanced evaluative judgments in business settings and future decisions, which increased the perception of trust and feeling of security (Dolcos et al., 2012). Brain imaging results revealed a strong activation of the social cognitive network in both approach and avoidance behaviors. In specific, the middle temporal gyrus (MTG), temporoparietal junction (TPJ), extrastriate body area (EBA), medial prefrontal cortex (mPFC), lateral prefrontal cortex (lPFC), inferior frontal gyrus (IFG), and right amygdala (rAMY) (Dolcos et al., 2012).

These social cognitive brain regions functioned to perceive interpersonal stimuli, observe still and moving physical movements, allow constructing evaluative judgment of another individual’s behavior, and connect emotion with non-verbal behaviors in social settings (Dolcos et al., 2012). Furthermore, increasing activity of AMY and posterior superior temporal sulcus (pSTS) were both correlated with positive evaluations of a handshake and approach behaviors (Dolcos et al., 2012). AMY has been implied in many studies of its vigilant attention in increasing physiological arousal and state of preparedness, and in this case, in approaching strangers (Dolcos et al., 2012; Yang et al., 2011). Since approach behaviors and handshake were both evaluated as positive, reduce physiological arousal in the skin conductance analysis was predictable (Dolcos et al., 2012). pSTS, on the other hand, was associated with haptic stimulation in handshakes. The comforting experience of being touched increased the perception of trust which was followed by higher ratings for approach behaviors (Dolcos et al., 2012). Such higher ratings opened a channel in the nucleus accumbens (NACC) which activated the reward pathway (Dolcos et al., 2012).

Trait judgment and facial trustworthiness. Todorov (2008) provided evidence of the traits which were correlated with facial trustworthiness. His team first identified trait dimensions that were correlated with faces and then, asked participants to rate those traits corresponding with faces on the computer screen. These ratings were then analyzed and they found that being attractive, intelligent, emotionally stable, responsible, sociable, caring, and confident were positively correlated with facial trustworthiness.

On the other hand, weird, mean, aggressive, unhappy, and dominant were negatively correlated (Todorov, 2008). Such valance evaluations pervaded into daily life social interaction and interpersonal impressions and demonstrated how trustworthiness encompasses all other trait judgments (Todorov, 2008, pp. 210). In addition, when trustworthiness judgment was conceptualized from a situation in which no context was given, it indicated the inferences about the positivity or negativity of the face (Todorov, 2008). Okubo, Ishitawa, and Kobayashi (2013) found that lower trustworthiness increased the intensity of perceived anger, whereas higher trustworthiness increased the intensity of perceived happiness (pp. 181). This led to the behavioral determination that whether an individual should avoid or approach a stranger (Todorov, 2008).

A brief overview of the Trust Game. Before the paper proceeds to other findings, it would be helpful to elaborate on what is the Trust Game first since many studies on the topic of trustworthiness are using this stimulus material as a center of their studies. The idea behind this game is to encourage an investor to earn as much money as possible through trades with a trustee. The investor is given an amount of initial endowment and it is up to the investor to keep or send it to the trustee. A reward is dependent on the choices that both investor and trustee did. (1) If both the investor and the trustee agreed to conduct the investment, both of them will receive a monetary reward ($$$) for the trade.

However, (2) if the investor decided to cancel the trade, both of them would receive less reward ($$). In the case of (3) the investor wanted to make the investment but canceled the transaction later on, the investor would receive only ($) reward while the trustee would receive ($$$$) reward (Hahn et al., 2014; Klackl, Pfundmair, Agroskin, & Jonas, 2013; Riedl et al, 2014; Schweiger, Stemmler, Burgdorf, & Wacker, 2014). The decision of the investor to send their endowment or not is considered a behavioral measure of trust, whereas the decision of the trustee to whether or not return the money is acted as a behavioral measurement of trustworthiness (Hahn et al., 2014; Klackl et al., 2013; Riedl et al, 2014; Schweiger et al., 2014). Besides, this game could be designed for a longitudinal study by increasing the rounds (Riedl et al., 2008) since coming to trust a stranger is a long process based on the other person’s behaviors during past exchanges (Hahn et al., 2014).

Brain activity during interpersonal trust attitude change. The willingness to put oneself in a vulnerable position and stay dependent on another person’s actions (Filkowski et al., 2016) involved a large scale of pre-judged trustworthiness. In a study, Filkowski et al. (2016) wanted to tie in the neural component when it came to interpersonal trustworthy attitude formation and emotional regulation. They employed fMRI to examine how humans attempt to control their attitudes of trust versus distrust. In both the trust and distrust conditions, activation activities were prominent in the temporoparietal junction (TPJ), medial prefrontal cortex (mPFC), insula, and inferior and lateral frontal cortices (Filkowski et al., 2016).

These brain regions encompassed the mentalizing system. In other words, mentalizing refers to the human ability to read the inner states of another individual in order to predict their emotional state, intention, and behavior, such as the theory of mind or perspective-taking (Filkowski et al., 2016; Riedl et al., 2014). Individual differences in change-of-trust scores were correlated with the activation of the precuneus (PrC) activity, meanwhile, individual differences in change-of-distrust scores were correlated with the activation of the inferior frontal gyrus (IFG) activity.

Activation in this region implied an engagement of inner speech in the cognitive control in distrust evaluations (Filkowski et al., 2016). AMY was not observed in terms of its activation activity in this study (Filkowski et al., 2016), but has been mentioned in many other previous studies (Doallo et al., 2012; Todorov, 2008). In addition, during the experiment before participants proceeded to their fMRI scan, they were instructed to read either a trust or distrust instruction. It was later found that faces paired with distrust instruction were eventually rated as untrustworthy (Filkowski et al., 2016). This validated the initial-negativity hypothesis.

The hypothesis described that distrust might account for a more salient negative social cue which is processed more automatically, rapidly, and strongly than trust cues (Filkowski et al., 2016). It is consistent with the evolutionary explanation that even early hominids prioritized negative over positive social information for the increased likelihood of survival (Filkowski et al., 2016; Riedl et al., 2014)

Cheaters, and hemispheric and hemifacial asymmetries of emotional expressions. Photographs of cheaters when they expressed anger were rated on a lower scale of interpersonal trustworthiness (Okubo et al., 2013) and this led to an avoidance approach (Todorov, 2008). However, the detection of cheaters became difficult when they posed a smile (Okubo et al., 2013). To further enrich such understandings, Okubo et al. (2013) investigated whether cheaters would use far more of their left hemiface – which is connected to the right hemisphere that controls the emotional side of the brain – to conceal their anti-social attitudes by posing a smile.

To test such inquiry, the trust game was used as the manipulation of the study. Consistent with the proposed hypothesis, left-hemifacial bias for trustworthiness was greater in rating cheaters than co-operators, even when both posed a smile. Ratings from the participants also revealed that the left hemiface bias for trustworthiness judgment when posing a smile looked happier than the right. This demonstrated that cheaters effectively co-operated the emotional side of the brain to manipulate others and conceal their anti-social attitudes (Okubo et al., 2013). It was also found that women – compared to men – were highly sensitive to detecting facial emotional expressions of cheaters (Okubo et al., 2013). Nevertheless, all of these findings of left-face bias became invalid when cheaters expressed anger on their faces (Okubo et al., 2013).

Response inhibition and emotional devaluation of faces. Distractor devaluation was selectively inhibited during visual selection inducing more negative affective judgments (Doallo et al., 2012). Doallo et al. (2012) evaluated fMRI scans to broaden the understanding of response inhibition and how it is related to interpersonal trustworthiness. Experimental materials included the Go/No-Go (GNG) task to novel faces and the rating of facial trustworthiness. Results confirmed the hypothesis that top-down motor and somatosensory controls executed with No-Go faces in which they impacted the negative evaluations of such faces. The top-down control was indeed controlled by the left middle frontal gyrus (lMFG), right post-central gyrus(r-pCG), insula, right thalamus, and left cerebellum for guiding actions (Doallo et al., 2012). In particular, inhibitory signals were encrypted with the representation of faces.

Then, it was used to lessen the values connected with such faces which stemmed the response inhibition of No-Go untrustworthy faces (Doallo et al., 2012). In addition to those motor-related and somatosensory areas, modulation of the limbic system – predominantly the rAMY – to No-Go stimuli during the trustworthy evaluation task was also significant. This showed the important role of this brain region in complex social judgments – trustworthiness in this case (Doallo et al., 2012). Examination of the orbitofrontal cortex (OFC) showed its activation in both of the Go/No-Go task and the latter facial evaluation of trustworthiness. The right lateral OFC was significantly activated the most (Doallo et al., 2012).

This area is associated with executive control mediating suppression of inappropriate responses, spatial and temporal orienting tasks using invalid cues, decision-making, and postponed non-matched samples (Doallo et al., 2012). Connecting back to AMY, there was a pattern of activation during the facial evaluation tasks in which, the larger the activation of the OFC, the higher the response of the AMY to No-Go-Low-Trust (Doallo et al., 2012). Such correlation indicated that OFC acted as top-down inhibitory processing on AMY’s emotional responses to untrustworthy faces (Doallo et al., 2012).

Evidence from Neuroendocrine Mechanisms

Opioid receptor blockade and warmth-liking. Opioidergic neurotransmission has been approached in various past studies and has been indicated in its affiliative goal, gratifying pleasure, and warm-liking, rather than the traditional view of dopaminergic neurotransmission (Schweiger et al., 2014). Those surprising findings have led Schweiger et al. (2014) to investigate the role of opioid receptor blockade (Naltrexone) versus placebo on their inducement of either a neutral emotional state or warm-liking. To quantify the study, the team used the trust game and then performed an EEG recording to clarify the valance hypothesis of frontal asymmetry.

The hypothesis claims that greater activation in the left frontal region for both positive and negative emotions, than in the right (Schweiger et al., 2014). In addition, the emotion/motivation-dependent lateralization model along with approach/withdrawal tendency was also postulated with left/right frontal asymmetrical brain activation (Schweiger et al., 2014). Results confirmed the impact of opioidergic neurotransmission controlled the positive emotion of warm-liking – which acted as an interpersonal trust behavior (Schweiger et al., 2014). Introducing inhibitory µ-opioid receptor – Naltrexone – reduced or even eliminated the experience of warm-liking and trustful subjective behaviors. This effect was demonstrated in the trust game where an investor reduced their investment in the real estate property, but not increased (Schweiger et al., 2014).

On the other hand, assuming that the investor increased their investment, it was not equivalent to the reduction of Naltrexone to participants’ high risk-taking decision to increase their investments (Schweiger et al., 2014). In fact, it was the impact of oxytocin on trustful behavior (Schweiger et al., 2014). In a nutshell, Naltrexone proved its modulatory effect of opioid receptor blockade on EEG frontal asymmetry and trust behavior (Schweiger et al., 2014). Interestingly enough, a frontal EEG scan provided that a low alpha band (8-10 Hz) was associated with a reduction of left frontal activity, after being introduced to positive emotions like warm-liking accompanied by decreased opioidergic tone (Schweiger et al., 2014). To be more precise, the reduction of activity on the left frontal region was associated with a reduction in approach motivation. Taken everything together, opioids played a role in warm-liking/pleasure consummatory processes (Schweiger et al., 2014).

Oxytocin promotes non-personalistic attributions in response to a trust betrayal. Numerous studies indeed investigated the effect of oxytocin (OT) administrated intranasally and revealed that this neuropeptide played a critical role in social bonding (Ross & Young, 2009; Klackl et al., 2013). To further observe the role of OT, Klackl et al. (2013) decided to conduct a study to explore whether OT promoted the use of non-personalistic attributions after the trust was betrayed.

The results confirmed the initial hypothesis that OT was highly correlated with angry rumination and non-personalistic attributions (Klackl et al., 2013). OT interestingly had no effect on making participants less angry. However, since OT generally played the role in social bonding – empathy, perspective taking, mentalizing, and self-other overlap – thus, that explained why OT reduced the likelihood of personalistic attributions in the OT group (Klackl et al., 2013).

Future Implications

The findings above have enriched the understanding of how humans came to trust each other through the cognitive, affective, and neuroscience lenses. Despite such achievements, there are still many questions that remain unsolved. First, it is a question regards on the psycho-neural learning mechanisms of interpersonal trustworthiness. The paper discussed the time interval it takes for an individual to make judgments on facial trustworthiness, but it did not provide detailed information on how it came to the conscious level. Trustworthy evaluative judgment might have arisen from the unconscious level of processing, but this idea would need to be further examined.

An unconscious level of processing might be needed to explore further. Second, most of the studies on trustworthiness have been conducted in laboratories and using dependent variables such as the trust game or showing participants with faces. Since interpersonal trustworthiness is done in many social situations, researchers should move away from artificial environments to real-life conditions. Third, using fMRI could reveal the source of brain activation in the initial interpersonal rating of trustworthiness. So far, it has been EEG and ERP are being employed in trying to answer the initial social interaction, and the brain processing mechanisms remain a mysterious answer. Last but not least, it has been suggested that the ability to trust others was deeply embedded since the existence of early hominids (Reidl et al., 2014), thus, it would be fascinating to see how such socio-cognitive behavior is encoded inside human genotypes.

Conclusion

This literature review provided evidence of the mechanism of one of the most fundamental social interpersonal behavior – trustworthiness. Studies are classified into two sub-categories of research, which are facial appearance and neuroendocrine mechanisms. Using vast methods manipulated from brain-social-computer interface – facial appearance, the trust game, ERP, EEG, fMRI, and neuro-hormones – research has found many different aspects of how humans come to trust one another. Research within the facial appearance approach has investigated the time frame of initial interaction, traits related to trustworthiness, detection of cheaters based on the left hemiface when posing a smile, brain activation during trust attitudes change, and devaluation of untrustworthy faces. The roles of opioids, opioid receptor blockade, and oxytocin on warmth-liking and social bonding have also been discussed. Several future implications for research are suggested. Such findings supported the affective and neurobiological perspectives gleaned from basic neuroscience work to enhance our understanding of interpersonal trustworthiness.

Facial appearance provokes personal evaluative judgments about others’ trustworthiness which is essential in daily human social interactions. Individuals always ask themselves “Should I trust this person?” whenever they encounter a stranger. That is because interpersonal trust requires one’s willingness to put themselves in a vulnerable position upon another individual’s course of action (Filkowski, Anderson, & Haas, 2016). The ability to discriminate whether someone is trustworthy or not is thought to have some survival significance since the existence of early hominids (Marzi, Righi, Ottonello, Cincotta, & Viggiano, 2014; Riedl, Mohr, Kenning, Davis, & Heekeren, 2014) and enable the beneficial engagement in cooperative reciprocal interactions (Hahn et al., 2014). Several studies have demonstrated that trustworthiness judgment occurs unintentionally and very quickly (Marzi et al, 2014; Hahn et al., 2014; Todorov, 2008; Yang, Qi, Ding, & Song, 2011), thus, it is not quite clear of the inner psycho-neural mechanisms occur during such minimal timeframe impression. This literature review is focused on the neural processing when an individual undergoes the evaluation of whether the other individual is trustworthy or not, using studies that employed functional magnetic response imaging (fMRI), event-related potentials (ERPs), electroencephalogram (EEG), Go/No-Go task, the Trust Game, and neuro-endocrinological structures (oxytocin and opioid) to explain such evaluative judgments of trustworthiness.

Evidence from Facial Appearance

When did trust initiate? Hahn et al. (2014) were interested in the initial trust level before an exchange with a stranger occurred by measuring the resting-state brain-electrical connectivity. The research team employed the trust game – which will be discussed later – and EEG data. In this study, Hahn et al. (2014) found that the initial level of trust was determined even before the game/exchange has even started. The authors posited that individual differences in initial trust level could be higher over time and heritable. On the other hand – opposing such notions – the initial trust level could be mostly influenced by an individual’s emotional state, thus, resting-state connectivity might be affected (Hahn et al., 2014). One limitation of the study was EEG could not able to identify which brain region is responsible for such resting-state dynamics in initial trust behaviors. Hahn et al (2014) suggested that the temporoparietal junction (TPJ) might be activated during investment decisions in the trust game, but further studies need to be evaluated.

Time exposure and facial evaluation. To investigate of how rapid facial evaluation on interpersonal trustworthiness is, Todorov (2008) designed a study that showed faces to participants for 100, 500, or 1000ms, and then asked them to evaluate the trustworthiness of those faces. Surprisingly, with just 33ms exposure to faces, participants could quickly identify which face was trustworthy and which one was not. This result suggested that the evaluation of facial trustworthiness is extremely efficient (Todorov, 2008; Marzi et al., 2014; Riedl et al., 2014; Yang et al., 2011).

The research team then did another follow-up with their results to see could their participants detect trustworthiness even faster than 33ms. In this experiment, eight different presentation times – from subliminal (17ms) to unlimited (no time constraint) – of facial presentations were shown to the participants. Judgment of trustworthiness for the subliminal presentation was considered to be difficult to distinguish between trustworthy and untrustworthy faces (Todorov, 2008). However, participants were able to make accurate judgments after 33ms. In the same manner, as the presentation time increased, participants increased their accuracy in making trustworthiness judgments. This pattern of accuracy stopped at 167ms and anything exceeded after this time interval did not improve the correlation (Todorov, 2008).

Time course of facial trustworthiness appraisal. Another study was conducted by Yang et al. (2011) to understand more about the early stages of perceptual processing in trustworthy and untrustworthy faces. Random faces were generated and shown to participants with shape alterations to make faces look either trustworthy or untrustworthy (Yang et al., 2011). The ERP results revealed that C1 amplitudes elicited greater in trustworthy faces in early visual processing at the occipito-parietal region, but not in untrustworthy faces (Yang et al., 2011). Scalp topography of electrophysiological response to trustworthy face was within a 400-600ms time interval (Yang et al., 2011) which indicated that the late positive component (LPC) was presented. LPC tended to demand the allocation of attention and trigger the motivation process of approach and avoidance tendencies. LPC was activated when untrustworthy faces were presented (Yang et al., 2011). The connection between LPC and the detection of untrustworthy faces indicated that an untrustworthy face conveys an avoidance tendency and therefore, neural processing would increase the level of attention than a trustworthy face (Dolcos, Sung, Argo, Flor-Henry, & Dolcos, 2013; Yang et al., 2011). LPC’s finding has indeed reconfirmed previous studies of fMRI and untrustworthiness. In specific, the amygdala (AMY) response level increased as the level of the untrustworthiness of the faces increased (Todorov, 2008).

Handshake and evaluative judgments in social interactions. Dolcos et al. (2012) were interested in the power of a handshake accompanying with approach versus avoidance of social interactions. Such evaluations are based on body language which could be influenced interpersonal impressions (Dolcos et al., 2012) and trustworthiness of whether the person is approachable or not (Todorov, 2008; Yang et al., 2011). To conduct the experiment, event-related fMRI and skin conductance analysis were both recorded while participants were watching a brief video presentation. This presentation consisted of a handshake scene followed by approach behaviors, no handshake scene followed by avoidance behaviors, or a control situation in which a cardboard of an interactor was shown (Dolcos et al., 2012).

Results showed that there were higher ratings for approach than avoidance behaviors, and a positive impact of handshake on social evaluations (Dolcos et al., 2012). Handshake was therefore perceived as optimistic behavior. It enhanced evaluative judgments in business settings and future decisions, which increased the perception of trust and feeling of security (Dolcos et al., 2012). Brain imaging results revealed a strong activation of the social cognitive network in both approach and avoidance behaviors. In specific, the middle temporal gyrus (MTG), temporoparietal junction (TPJ), extrastriate body area (EBA), medial prefrontal cortex (mPFC), lateral prefrontal cortex (lPFC), inferior frontal gyrus (IFG), and right amygdala (rAMY) (Dolcos et al., 2012).

These social cognitive brain regions functioned to perceive interpersonal stimuli, observe still and moving physical movements, allow constructing evaluative judgment of another individual’s behavior, and connect emotion with non-verbal behaviors in social settings (Dolcos et al., 2012). Furthermore, increasing activity of AMY and posterior superior temporal sulcus (pSTS) were both correlated with positive evaluations of a handshake and approach behaviors (Dolcos et al., 2012). AMY has been implied in many studies of its vigilant attention in increasing physiological arousal and state of preparedness, and in this case, in approaching strangers (Dolcos et al., 2012; Yang et al., 2011). Since approach behaviors and handshake were both evaluated as positive, reduce physiological arousal in the skin conductance analysis was predictable (Dolcos et al., 2012). pSTS, on the other hand, was associated with haptic stimulation in handshakes. The comforting experience of being touched increased the perception of trust which was followed by higher ratings for approach behaviors (Dolcos et al., 2012). Such higher ratings opened a channel in the nucleus accumbens (NACC) which activated the reward pathway (Dolcos et al., 2012).

Trait judgment and facial trustworthiness. Todorov (2008) provided evidence of the traits which were correlated with facial trustworthiness. His team first identified trait dimensions that were correlated with faces and then, asked participants to rate those traits corresponding with faces on the computer screen. These ratings were then analyzed and they found that being attractive, intelligent, emotionally stable, responsible, sociable, caring, and confident were positively correlated with facial trustworthiness.

On the other hand, weird, mean, aggressive, unhappy, and dominant were negatively correlated (Todorov, 2008). Such valance evaluations pervaded into daily life social interaction and interpersonal impressions and demonstrated how trustworthiness encompasses all other trait judgments (Todorov, 2008, pp. 210). In addition, when trustworthiness judgment was conceptualized from a situation in which no context was given, it indicated the inferences about the positivity or negativity of the face (Todorov, 2008). Okubo, Ishitawa, and Kobayashi (2013) found that lower trustworthiness increased the intensity of perceived anger, whereas higher trustworthiness increased the intensity of perceived happiness (pp. 181). This led to the behavioral determination that whether an individual should avoid or approach a stranger (Todorov, 2008).

A brief overview of the Trust Game. Before the paper proceeds to other findings, it would be helpful to elaborate on what is the Trust Game first since many studies on the topic of trustworthiness are using this stimulus material as a center of their studies. The idea behind this game is to encourage an investor to earn as much money as possible through trades with a trustee. The investor is given an amount of initial endowment and it is up to the investor to keep or send it to the trustee. A reward is dependent on the choices that both investor and trustee did. (1) If both the investor and the trustee agreed to conduct the investment, both of them will receive a monetary reward ($$$) for the trade.

However, (2) if the investor decided to cancel the trade, both of them would receive less reward ($$). In the case of (3) the investor wanted to make the investment but canceled the transaction later on, the investor would receive only ($) reward while the trustee would receive ($$$$) reward (Hahn et al., 2014; Klackl, Pfundmair, Agroskin, & Jonas, 2013; Riedl et al, 2014; Schweiger, Stemmler, Burgdorf, & Wacker, 2014). The decision of the investor to send their endowment or not is considered a behavioral measure of trust, whereas the decision of the trustee to whether or not return the money is acted as a behavioral measurement of trustworthiness (Hahn et al., 2014; Klackl et al., 2013; Riedl et al, 2014; Schweiger et al., 2014). Besides, this game could be designed for a longitudinal study by increasing the rounds (Riedl et al., 2008) since coming to trust a stranger is a long process based on the other person’s behaviors during past exchanges (Hahn et al., 2014).

Brain activity during interpersonal trust attitude change. The willingness to put oneself in a vulnerable position and stay dependent on another person’s actions (Filkowski et al., 2016) involved a large scale of pre-judged trustworthiness. In a study, Filkowski et al. (2016) wanted to tie in the neural component when it came to interpersonal trustworthy attitude formation and emotional regulation. They employed fMRI to examine how humans attempt to control their attitudes of trust versus distrust. In both the trust and distrust conditions, activation activities were prominent in the temporoparietal junction (TPJ), medial prefrontal cortex (mPFC), insula, and inferior and lateral frontal cortices (Filkowski et al., 2016).

These brain regions encompassed the mentalizing system. In other words, mentalizing refers to the human ability to read the inner states of another individual in order to predict their emotional state, intention, and behavior, such as the theory of mind or perspective-taking (Filkowski et al., 2016; Riedl et al., 2014). Individual differences in change-of-trust scores were correlated with the activation of the precuneus (PrC) activity, meanwhile, individual differences in change-of-distrust scores were correlated with the activation of the inferior frontal gyrus (IFG) activity.

Activation in this region implied an engagement of inner speech in the cognitive control in distrust evaluations (Filkowski et al., 2016). AMY was not observed in terms of its activation activity in this study (Filkowski et al., 2016), but has been mentioned in many other previous studies (Doallo et al., 2012; Todorov, 2008). In addition, during the experiment before participants proceeded to their fMRI scan, they were instructed to read either a trust or distrust instruction. It was later found that faces paired with distrust instruction were eventually rated as untrustworthy (Filkowski et al., 2016). This validated the initial-negativity hypothesis.

The hypothesis described that distrust might account for a more salient negative social cue which is processed more automatically, rapidly, and strongly than trust cues (Filkowski et al., 2016). It is consistent with the evolutionary explanation that even early hominids prioritized negative over positive social information for the increased likelihood of survival (Filkowski et al., 2016; Riedl et al., 2014)

Cheaters, and hemispheric and hemifacial asymmetries of emotional expressions. Photographs of cheaters when they expressed anger were rated on a lower scale of interpersonal trustworthiness (Okubo et al., 2013) and this led to an avoidance approach (Todorov, 2008). However, the detection of cheaters became difficult when they posed a smile (Okubo et al., 2013). To further enrich such understandings, Okubo et al. (2013) investigated whether cheaters would use far more of their left hemiface – which is connected to the right hemisphere that controls the emotional side of the brain – to conceal their anti-social attitudes by posing a smile.

To test such inquiry, the trust game was used as the manipulation of the study. Consistent with the proposed hypothesis, left-hemifacial bias for trustworthiness was greater in rating cheaters than co-operators, even when both posed a smile. Ratings from the participants also revealed that the left hemiface bias for trustworthiness judgment when posing a smile looked happier than the right. This demonstrated that cheaters effectively co-operated the emotional side of the brain to manipulate others and conceal their anti-social attitudes (Okubo et al., 2013). It was also found that women – compared to men – were highly sensitive to detecting facial emotional expressions of cheaters (Okubo et al., 2013). Nevertheless, all of these findings of left-face bias became invalid when cheaters expressed anger on their faces (Okubo et al., 2013).

Response inhibition and emotional devaluation of faces. Distractor devaluation was selectively inhibited during visual selection inducing more negative affective judgments (Doallo et al., 2012). Doallo et al. (2012) evaluated fMRI scans to broaden the understanding of response inhibition and how it is related to interpersonal trustworthiness. Experimental materials included the Go/No-Go (GNG) task to novel faces and the rating of facial trustworthiness. Results confirmed the hypothesis that top-down motor and somatosensory controls executed with No-Go faces in which they impacted the negative evaluations of such faces. The top-down control was indeed controlled by the left middle frontal gyrus (lMFG), right post-central gyrus(r-pCG), insula, right thalamus, and left cerebellum for guiding actions (Doallo et al., 2012). In particular, inhibitory signals were encrypted with the representation of faces.

Then, it was used to lessen the values connected with such faces which stemmed the response inhibition of No-Go untrustworthy faces (Doallo et al., 2012). In addition to those motor-related and somatosensory areas, modulation of the limbic system – predominantly the rAMY – to No-Go stimuli during the trustworthy evaluation task was also significant. This showed the important role of this brain region in complex social judgments – trustworthiness in this case (Doallo et al., 2012). Examination of the orbitofrontal cortex (OFC) showed its activation in both of the Go/No-Go task and the latter facial evaluation of trustworthiness. The right lateral OFC was significantly activated the most (Doallo et al., 2012).

This area is associated with executive control mediating suppression of inappropriate responses, spatial and temporal orienting tasks using invalid cues, decision-making, and postponed non-matched samples (Doallo et al., 2012). Connecting back to AMY, there was a pattern of activation during the facial evaluation tasks in which, the larger the activation of the OFC, the higher the response of the AMY to No-Go-Low-Trust (Doallo et al., 2012). Such correlation indicated that OFC acted as top-down inhibitory processing on AMY’s emotional responses to untrustworthy faces (Doallo et al., 2012).

Evidence from Neuroendocrine Mechanisms

Opioid receptor blockade and warmth-liking. Opioidergic neurotransmission has been approached in various past studies and has been indicated in its affiliative goal, gratifying pleasure, and warm-liking, rather than the traditional view of dopaminergic neurotransmission (Schweiger et al., 2014). Those surprising findings have led Schweiger et al. (2014) to investigate the role of opioid receptor blockade (Naltrexone) versus placebo on their inducement of either a neutral emotional state or warm-liking. To quantify the study, the team used the trust game and then performed an EEG recording to clarify the valance hypothesis of frontal asymmetry.

The hypothesis claims that greater activation in the left frontal region for both positive and negative emotions, than in the right (Schweiger et al., 2014). In addition, the emotion/motivation-dependent lateralization model along with approach/withdrawal tendency was also postulated with left/right frontal asymmetrical brain activation (Schweiger et al., 2014). Results confirmed the impact of opioidergic neurotransmission controlled the positive emotion of warm-liking – which acted as an interpersonal trust behavior (Schweiger et al., 2014). Introducing inhibitory µ-opioid receptor – Naltrexone – reduced or even eliminated the experience of warm-liking and trustful subjective behaviors. This effect was demonstrated in the trust game where an investor reduced their investment in the real estate property, but not increased (Schweiger et al., 2014).

On the other hand, assuming that the investor increased their investment, it was not equivalent to the reduction of Naltrexone to participants’ high risk-taking decision to increase their investments (Schweiger et al., 2014). In fact, it was the impact of oxytocin on trustful behavior (Schweiger et al., 2014). In a nutshell, Naltrexone proved its modulatory effect of opioid receptor blockade on EEG frontal asymmetry and trust behavior (Schweiger et al., 2014). Interestingly enough, a frontal EEG scan provided that a low alpha band (8-10 Hz) was associated with a reduction of left frontal activity, after being introduced to positive emotions like warm-liking accompanied by decreased opioidergic tone (Schweiger et al., 2014). To be more precise, the reduction of activity on the left frontal region was associated with a reduction in approach motivation. Taken everything together, opioids played a role in warm-liking/pleasure consummatory processes (Schweiger et al., 2014).

Oxytocin promotes non-personalistic attributions in response to a trust betrayal. Numerous studies indeed investigated the effect of oxytocin (OT) administrated intranasally and revealed that this neuropeptide played a critical role in social bonding (Ross & Young, 2009; Klackl et al., 2013). To further observe the role of OT, Klackl et al. (2013) decided to conduct a study to explore whether OT promoted the use of non-personalistic attributions after the trust was betrayed.

The results confirmed the initial hypothesis that OT was highly correlated with angry rumination and non-personalistic attributions (Klackl et al., 2013). OT interestingly had no effect on making participants less angry. However, since OT generally played the role in social bonding – empathy, perspective taking, mentalizing, and self-other overlap – thus, that explained why OT reduced the likelihood of personalistic attributions in the OT group (Klackl et al., 2013).

Future Implications

The findings above have enriched the understanding of how humans came to trust each other through the cognitive, affective, and neuroscience lenses. Despite such achievements, there are still many questions that remain unsolved. First, it is a question regards on the psycho-neural learning mechanisms of interpersonal trustworthiness. The paper discussed the time interval it takes for an individual to make judgments on facial trustworthiness, but it did not provide detailed information on how it came to the conscious level. Trustworthy evaluative judgment might have arisen from the unconscious level of processing, but this idea would need to be further examined.

An unconscious level of processing might be needed to explore further. Second, most of the studies on trustworthiness have been conducted in laboratories and using dependent variables such as the trust game or showing participants with faces. Since interpersonal trustworthiness is done in many social situations, researchers should move away from artificial environments to real-life conditions. Third, using fMRI could reveal the source of brain activation in the initial interpersonal rating of trustworthiness. So far, it has been EEG and ERP are being employed in trying to answer the initial social interaction, and the brain processing mechanisms remain a mysterious answer. Last but not least, it has been suggested that the ability to trust others was deeply embedded since the existence of early hominids (Reidl et al., 2014), thus, it would be fascinating to see how such socio-cognitive behavior is encoded inside human genotypes.

Conclusion

This literature review provided evidence of the mechanism of one of the most fundamental social interpersonal behavior – trustworthiness. Studies are classified into two sub-categories of research, which are facial appearance and neuroendocrine mechanisms. Using vast methods manipulated from brain-social-computer interface – facial appearance, the trust game, ERP, EEG, fMRI, and neuro-hormones – research has found many different aspects of how humans come to trust one another. Research within the facial appearance approach has investigated the time frame of initial interaction, traits related to trustworthiness, detection of cheaters based on the left hemiface when posing a smile, brain activation during trust attitudes change, and devaluation of untrustworthy faces. The roles of opioids, opioid receptor blockade, and oxytocin on warmth-liking and social bonding have also been discussed. Several future implications for research are suggested. Such findings supported the affective and neurobiological perspectives gleaned from basic neuroscience work to enhance our understanding of interpersonal trustworthiness.