Dissecting shared pain representations to understand their behavioral and clinical relevance

Accounts of shared representations posit that the experience of pain and pain empathy rely on similar neural mechanisms. Experimental research employing novel analytical and methodological approaches has made significant advances in both the identification and targeted manipulation of such shared experiences and their neural underpinnings. This revealed that painful experiences can be shared on different representational levels, from pain-specific to domain-general features, such as negative affect and its regulation. In view of direct links between such representations and social behaviors such as prosocial behavior, conditions characterized by aberrant pain processing may come along with heavy impairments in the social domain, depending on the affected representational level. This has wide potential implications in light of the high prevalence of pain-related clinical conditions, their management, and the overuse of pain medication. In this review and opinion paper, we aim to chart the path toward a better understanding of the link between shared affect and prosocial behavior.


Why we should care about the way we share
Pain-related conditions are highly prevalent and debilitating disorders, and they frequently come along with impairments in the domain of social functioning.While these social impairments are commonly attributed to certain limitations set by the disorders themselves (such as the inability to attend leisure activities due to chronic pain), one important aspect has been largely overlooked so far: The dysfunctionalities or imbalances of pain-regulatory systems associated with the clinical conditions lead to a wide range of deficits, from basic nociceptive processes up to difficulties in dealing with the affective consequences of painful and aversive experiences.This will affect not only a person's own pain and emotion regulation but also how they perceive and empathize with the affective states of others.The shared representations account of empathy strongly suggests the latter, as it proposes that we engage similar neural representations when experiencing an affective state first-hand and when perceiving and responding to another person experiencing that affective state (de Waal and Preston, 2017;Lamm et al., 2019).However, some debate has been revolving around the specificity and the precise levels at which pain representations are shared (Zaki et al., 2016).This debate results from pain being a multi-dimensional experience that can be decomposed into a range of different features broadly categorized into sensory-discriminative (location, intensity, quality), affective (discomfort, valence, arousal), and cognitive (motivation, learning) dimensions.While the sensory-discriminative features are closely connected to nociception-related processes and thus cover a level that is fairly pain-specific, the affective and cognitive features also recruit domain-general processes, such as arousal and executive functions, that lack that specificity.Importantly, such processes are also relevant for the sharing of affective states other than pain.As a consequence, clinical conditions characterized by aberrant pain processing potentially impact a range of socio-perceptual and socio-behavioral functions, depending on the condition-specific alterations and the associated level at which shared representations may be affected.Put into simple words, these conditions may lead to reduced or less effective prosocial and affiliative behaviors, due to deficits in empathic sharing.This may consequently impact the quality and the extent of social relationships of the persons affected by them and even drive them into a vicious cycle of less pain-buffering social support and, consequently, more extensive pain and related burdening of mental and physical health.This is in line with more recent definitions of pain (Williams and Craig, 2016), which also include a social component and emphasize the inherently social nature of human care systems and pain management (including the communication of distress to others).
In light of the high prevalence of acute and chronic pain conditions in western societies (20 % of the adult US population reported chronic pain (Dahlhamer et al., 2018) and developing countries (18 % (Sa et al., 2019)), as well as the pressing issue of opioid abuse it is thus a major shortcoming that our knowledge on how pain and its treatment may affect empathy and the connected social behaviors is very limited.This review aims to put a spotlight on this issue.It builds up on recent research in social neuroscience that has developed and used novel experimental and analytical approaches to show a) that the different representational levels are discriminable in terms of the associated neural patterns, b) that they can be separately manipulated, and c) that such manipulations affect social behavior in specific, level-dependent ways.
Our main focus is on prosocial behavior, as a specific type of social functioning that has been directly linked to empathy in the sense of affect sharing.Certainly, empathy with the negative state of another must not necessarily lead to a prosocial response, but could also lead to heightened emotional distress, self-protective neglect, or cognitive reappraisal.Also, shared affective states are not the only driving forces behind prosocial behavior: for example, morality and social norms are other important drivers of benefitting others, as is compassion.Gallagher and colleagues recently proposed a pattern theory of compassion, in which they explore the conceptual overlaps and distinctions between empathy in the sense of affect sharing and compassion, which always contains a caring aspect in the sense of being motivated to act beneficially for others (Gallagher et al., 2024; see also Singer and Lamm, 2009, for similar earlier views).That said, there is a clearly established link between empathy and prosociality (Lamm et al., 2019), and we focus on this specific class of behavior in this review for several reasons: First, affective resonance has been proposed to be the core mechanism for acting in a prosocial manner (Decety et al., 2016;Wu and Hong, 2022).Second, as described below, research into measuring and manipulating shared affective representations has greatly evolved.Third, prosocial behavior is overt behavior that can easily be operationalized and measured in an experimental context (in contrast to e.g., social cohesion).Fourth, it is central to other, more covert social functions like affiliation and to healthy human relationships in general: Developmental studies have found that higher peer ratings of children's prosocial behavior are linked to higher levels of social functioning (Eisenberg et al., 1996), and that difficulties in certain prosocial behaviors during childhood predict poorer social functioning in adolescence (Donohue et al., 2020).Prosocial tendencies are further positively related to social competence and self-regulation (Eisenberg et al., 2013), but negatively related to externalizing and internalizing problems (Eisenberg et al., 2024).
In the following subsections, we will first briefly review the state of the art of research on empathy and shared representations, focusing on experimental approaches demonstrating the identifiability and modifiability of different levels of representations.Then, we will discuss how these levels of representations may be linked to behavior, focusing on recent experimental approaches in neurotypical populations as well as work on empathy in conditions characterized by alterations of pain regulatory systems.Finally, we will highlight open questions and pressing research avenues.We chose to conduct a purposive (non-systematic) review, with the aim to reflect broadly upon the theme by integrating knowledge from different fields.We included publications selected based on knowledge generated from our research in this field and related fields, covering neuroscientific, developmental, socialpsychological, clinical and animal research work, spanning the timeperiod from 1996 to 2024.We put a focus on human fMRI findings but also included other neuroscientific methods, backed by seminal papers and comprehensive reviews capturing the state of the art.

Shared representations in pain empathy and social behavior
According to simulation theory, we rely on our own cognitive capacities and mechanisms to understand others' internal states in a fast and cognitively parsimonious way (Goldman, 2006).Especially in the domain of empathy, scholars have strived to provide neuroscientific evidence for shared representations allowing for simulation of others' states.
Twenty-five years ago, Hutchison et al. for the first time demonstrated single-neuron responses of pain-responsive neurons in the anterior cingulate cortex to the observation of pain in others in cingulotomy patients (Hutchison et al., 1999).A few years later, two seminal fMRI studies (Jackson et al., 2005;Singer et al., 2004) demonstrated overlapping brain activation as measured using mass-univariate fMRI for self-experienced pain and observing pain in others.This pointed for the first time to the phenomenon that we recruit our own bodily pain system when perceiving pain and suffering in others (see also Wicker et al., 2003, for similar fMRI findings in the domain of disgust).This claim was later substantiated by first meta-analyses of follow-up studies that indeed revealed a high consistency in the engagement of shared networks and representations (Fan et al., 2011;Lamm et al., 2011).
The large number of studies employing pain empathy paradigms that have appeared since then has led to refined activation-likelihood estimation meta-analyses (Fallon et al., 2020;Jauniaux et al., 2019;Kogler et al., 2020;Timmers et al., 2018).They add to the consensus regarding the originally established "core pain empathy network" (Fan et al., 2011;Lamm et al., 2011), comprising mainly anterior insular (aIns; extending to inferior frontal gyrus, IFG) and anterior midcingulate (aMCC; extending to supplementary motor area, SMA) cortices, but also extend this network by adding the post-central gyrus (primary somatosensory cortex, S1; and supramarginal gyrus; SMG).While these regions are active during empathy for non-painful negative affective states as well (though to a smaller extent), empathy for pain appears to specifically recruit greater portions of the aMCC, as well as the precuneus and regions involved in somatosensory processing, i.e., bilateral mid-insula and S1 (Timmers et al., 2018).Depending on specific characteristics of the employed tasks, additional regions are recruited to a certain extent (Jauniaux et al., 2019), pointing to context-and stimulus-dependent recruitment of different features.There is also evidence that empathy does not cover all aspects and details of the other person's pain experience, and that this is highly context-and stimulus-dependent.For instance, early studies employing transcranial magnetic stimulation (TMS) demonstrated the modulation of activity of motor-evoked potentials during the empathic observation of other's hands and feet being either penetrated by a needle or touched by a Q-tip (Avenanti et al., 2005).They observed muscle-specific modulations in the amplitude of the recorded motor-evoked potentials, and also showed that this modulation was present only when the muscles were deeply penetrated, as opposed to being merely pinpricked (Avenanti et al., 2006).Furthermore, the degree of these modulatory effects was shown to depend on personality traits of the observer (Avenanti et al., 2009).While another more recent TMS study demonstrated the empathic tracking of the sensory-discriminative aspect of others' pain in a prosocial helping context (Gallo et al., 2018), two studies (Hartmann et al., 2021a(Hartmann et al., , 2021b) ) showed that while somatosensory activity during self-experienced pain was down-regulated via placebo analgesia in a body location-specific manner (left vs. right hand), no location-specific effects were found in the empathy condition.This matches the results of a mouse model of empathy study (Smith et al., 2021): while mice undergoing injections that led to long-lasting pain in the affected paw developed location-specific hypersensitivity to pain, bystander mice expressed acquired socially transferred pain in a location-independent manner (i.e.affecting both tested paws).This suggests that across species, sharing the sensory-discriminative features of pain (here, its bodily location) may be of lesser importance than the affective component, or that it only plays a role under specific circumstances (e.g. if the somatosensory cues are essential to extract information on the quality and quantity of the other person's pain, or when they are relevant for prosocial behavior; see also (Keysers et al., 2010;Riečanský and Lamm, 2019) for reviews on the role of somatosensation in social perception).
Thus, similar to the multi-faceted experience of pain, the empathic perception of others' pain relies on the recruitment of various neural mechanisms and a wide-spread network of brain regions (Fallon et al., 2020;Fan et al., 2011;Jauniaux et al., 2019;Kogler et al., 2020;Lamm et al., 2011;Timmers et al., 2018).While the large body of research on shared activations has made highly relevant contributions to the scientific understanding of pain empathy, these studies are somewhat restricted regarding their explanatory power.This is due to the coarse spatial resolution and the prevalent analysis approaches of contemporary human fMRI, where mass-univariate analyses indicate activations that, while overlapping analytically, in reality may result from distinct neural computations and representations (see Lamm et al., 2016;Zaki et al., 2016, for reviews).

Identifying shared representations via neural patterns
Multivariate pattern analysis (MVPA) approaches allow to some extent to overcome such limitations by targeting patterns of neural activation across many simultaneously analysed voxels.In the domain of empathy research, this yielded evidence for shared representations on different levels of specificity.For instance, using closely matched firsthand and empathy conditions, Corradi-dell'Acqua and colleagues provided initial evidence for shared neural patterns between selfexperienced pain and pain empathy (Corradi-Dell'Acqua et al., 2011) in the left anterior insula (aIns) and aMCC.A follow-up study of the same group (Corradi-Dell'Acqua et al., 2016) found that these patterns generalized to first-hand and empathic experiences of disgust, as well as to unfair offers made in an ultimatum game.The authors concluded that this shared pattern reflected the processing of negative affect more generally rather than coding for pain-specific processes.Importantly, though, they also found modality-specific patterns, indicating an important aspect that we will return to later: some aspects of brain responses, within a single study, will indicate domain-general shared representations, while others can be pain-specific.While these studies used a so-called search-light MVPA approach, analysing for local patterns, research looking at patterns of activation across the whole brain expanded our understanding of shared affect.An earlier study had suggested that pain empathy may engage mentalizing processes rather than empathic pain (Krishnan et al., 2016), but their findings may be interpreted by an "empathy condition" that in fact required imagery of first-hand pain, and thus processes not connected to empathy.A more recent follow-up study (Zhou et al., 2020) using a similar analysis approach but employing a passive viewing paradigm (of people in pain) in a large sample (N=252) provided compelling evidence for sharing of pain-specific representations.This study revealed a vicarious pain pattern (with a central role of the mid-insula) that was not only predictive of (empathy for) facial expressions of pain and limbs in painful situations, but also predicted levels of intensity of first-hand thermal pain.Moreover, this pattern was not sensitive to arousal, and performed significantly worse when predicting negative affect.This is an important finding in the present context, as it assesses distinct levels of representations and suggests rather specifically shared pain-related representations.This conclusion has recently been corroborated by a related study (Li et al., 2024) employing the neurological pain signature (NPS; a multivariate brain signature developed for tracking acute somatic pain (Wager et al., 2013)).This study revealed that using the NPS allowed to successfully distinguish passive viewing of painful vs. non-painful pictures.Again, this was mainly driven by responses in the right anterior insula and the mid-cingulate cortex.The findings reviewed in this subsection also indicate that the outcomes of such studies can be strongly driven by the exact nature of the task and the accompanying instructions, control conditions, statistical power, and the analytical approach (search for common vs. distinct representations).Careful experimentation and analysis strategies, including the application of complementary analysis approaches, and different patterns/signatures (such as testing for mere picture-induced negative affect (Chang et al., 2015), arousal (Zhou et al., 2020) or empathic distress (Ashar et al., 2017)) are necessary in order to come to robust and convergent conclusions.But in essence, accumulating evidence reveals that we now have the means to identify shared neural processes at different representational levels, using multivariate neuroimaging, and that this indicates we may share the pain of others both in domain-general ("modality-unspecific", related to e.g.negative affect and arousal) and experience-specific neural circuits.

Modulating shared representations via causal manipulations
Even the most sophisticated fMRI analysis approaches still only yield insights into neural correlates, which is why complementary approaches that causally manipulate the neural, cognitive and affective states of interest are key.Such approaches have revealed insights into how the various representational levels are related to the experience of empathy, based on the premise of a "joint modulation" logic: if pain and pain empathy indeed should rely on shared neural processes, then manipulating the first-hand experience of pain will lead to a similar effect on the empathic experience of painsuch as a down-regulation of pain empathy in the presence of a down-regulation of first-hand pain by means of an administered painkiller.In a research program employing such an approach it was repeatedly demonstrated that placebo analgesia leads to joint down-regulation of first-hand pain and pain empathy, both in terms of self-reported empathy and neural processing (Rütgen et al., 2015a(Rütgen et al., , 2015b).An opioid receptor antagonist abolished these effects both on the self-report level (Rütgen et al., 2015a), and on the level of event-related potentials (Rütgen et al., 2018) (see critical review by Coll, 2018 on event-related potential studies of empathy).Crucially, a follow-up study demonstrated that different levels of shared representations could be targeted via distinct mechanisms: while opioid receptor blockade affected pain and pain empathy, it had no significant effect on first-hand and empathic representations of another affective state, namely unpleasant affective touch (Rütgen et al., 2021).In contrast to these pain-specific effects of the opioid antagonist, placebo analgesia without opioidergic blockade had a down-regulating effect on first-hand unpleasant touch and empathy for such touch, but no effect whatsoever on pleasant touch.These findings thus highlight the modifiability of shared representations on two different levels: first, representations of negative affective states, evidenced by common regulation of pain and unpleasant touch and their empathic experiences by placebo analgesia.Second, pain-specific representations, demonstrated by the specific regulation of pain and pain empathy by opioid receptor blockade.Importantly, neuroimaging of placebo analgesia effects on pain and unpleasant touch revealed that different neural networks in anterior insular and mid-cingulate cortices may underpin the distinct effects.In conclusion, the effects of placebo analgesia on pain and pain empathy appear to be partly pain-specific, but also modulate negative affect processing on a more general level, possibly via expectancy and learning (Atlas, 2021).A similar approach was taken by Mischkowski and colleagues, who conducted two well-powered studies on the effects of the widely used non-opioidergic painkiller acetaminophen (Mischkowski et al., 2016(Mischkowski et al., , 2019)).They found that single administrations of acetaminophen led to reduced empathy for pain, but also reduced empathy for positive events and social pain.In line with shared representations accounts, effects of the substance on empathy for painful noise blasts were mediated by effects on unpleasantness during self-experienced noise blasts.Lastly, they also found reduced empathy for social rejection.An earlier study demonstrated a similar dampening effect of 3-week intake of acetaminophen on the first-hand experience of social rejection, accompanied by reduced neural responses in the anterior insula and ACC (DeWall et al., 2010;see Eisenberger, 2012 for review).This speaks for a non-pain-specific effect of acetaminophen on various aspects of social cognition, or put differently, for an effect on multiple levels of representations.Despite being an efficient and widely consumed painkiller, the mechanism of action of acetaminophen (which is the most commonly used pain medication in the United States and Europe) is not fully understood: it may exert its effects via reducing central cyclooxygenase, and via modulating the endogenous cannabinoid as well as the serotonergic system (Ohashi and Kohno, 2020), which may explain the broad range of observed effects.Pharmacological agents targeting the opioid system may exert more focused effects: Wager et al. established effects of the opioid agonist remifentanil on the level of pain-specific representations via reduced signature responses of the NPS (Wager et al., 2013).Yet, to our knowledge, effects on neural signatures specific to other levels of representations were not tested for opioidergic agents.One way to specifically modulate the representational level of arousal may be propranolol, which has been shown to influence emotional processing and downregulate amygdala responses to emotional stimuli irrespective of valence (Hurlemann et al., 2010;see Terbeck et al., 2016 for review).This may be a promising avenue to test more domain-general, arousal-related representations.
Animal models of empathy, and in particular lesion studies allow to move beyond some of the limitations of research in humans (see (Paradiso et al., 2021) for a recent in-depth review of rodent and non-human primate models of empathy).Early demonstrations of social modulation of pain as evidence for empathy in mice (e.g., Langford et al., 2006), paved the way for rodents being established as suitable models for behavioral and neural research in the domain of empathy (see Panksepp and Panksepp, 2013 for review).Recent neuroscientific work in rodents even showed the crucial involvement of particular brain areas such as the cingulate cortex in modulating both pain and pain empathy.Sakaguchi and colleagues found that pain-responsive neurons in the mouse ACC are similarly recruited during the observation of conspecifics in pain (Sakaguchi et al., 2018).Carrillo and colleagues presented multi-method converging evidence for shared pain representations in the rat ACC (Carrillo et al., 2019) by employing single-cell recordings during a shock observation paradigm.Their findings and methodological approach largely align with those of the placebo analgesia studies in humans.First, their study documents the activation of pain-responsive neurons during the observation of another rat's pain.Importantly, this effect seemed specific to pain, as the majority of these neurons did not respond to another unpleasant affective experience, namely fear.Thus, their approach identified both pain-selective neurons and neurons that responded to more general features of negative affective experiences, suggesting that pain can be regulated in isolation or in concert with other states.In addition, a neural pattern trained on observation of another rat's pain could accurately predict the intensity of a rat's self-experienced pain, which closely matches the abovementioned insights from human experiments (Zhou et al., 2020).Finally, pharmacologically deactivating the ACC via muscimol microinjections led to a specific reduction of freezing in response to pain observation, but not to differences in conditioned fear responses.A very recent study (Zhang et al., 2024) in mice further extended these findings.First, they corroborated the conclusions of the above-mentioned rodent studies by demonstrating shared pain-specific representations (vs.general responses to stress) in the mouse ACC when observing conspecifics in pain and when being in pain themselves.On top, the authors investigated targeted helping behaviour (allolicking at injury sites) and were able to show that observer mice who underwent chemogenetic inhibition of the ACC exhibited significantly reduced prosocial behavior (as compared to saline-injected observer mice).They also demonstrated enhancement of prosocial behaviours through augmenting ACC activity.These effects were not related to differences in general sociability and were specific to this kind of targeted prosocial behaviour (vs.grooming behaviour).Thereby, these findings corroborate and extend findings of an early seminal study (Bartal et al., 2011) of prosocial behavior in rats, in which it was demonstrated that rats liberated trapped cage mates in response to their distress (see Keysers et al., 2022;Meyza et al., 2017 for reviews).

Shared representations and their relevance for social behaviors
Empathy is key in social interactions because it allows us to better understand and represent the inner states of others.It is even more important for social functioning, though, via its direct impact on motivating (pro)social behaviors.If our ability to simulate and thereby understand another person's internal state is limited due to an impairment at any representational level, we will also be less efficient in adequately responding to them (e.g., by prosocial actions).It is thus surprising that knowledge is scant on whether and how precisely shared representations influence social behavior in humans.In fact, there is an obvious imbalance between the extensive debate on how we represent and share others' affective states, and investigations of how and in what specific ways these representations are relevant for how we behave and interact with others.This is a major limitation when it comes to understanding the real-life impact of empathy and its dysfunctions; one major aim of the present review is, thus, to call for further research that will help fill this gap of knowledge.
The degree to which social behaviors may be affected by changes in shared representations will in the first place be determined by the position of these changes on a continuum between highly specific and domain-general impairments.As the impairments move from highly specific to domain-general levels, they will also act on an increasingly broader range of behaviors.Highly specific impairments (such as sensory deficits in congenital insensitivity to pain) will most likely have a rather subtle impact on everyday social behavior, while domain-general problems (such as systemic imbalances of affective-hedonic and motivational-reward related brain systems in prolonged opioid use disorder) will affect a range of behaviors, from affiliative behavior to helping others.
Individual consequences of conditions such as chronic pain and opioid use disorder depend on a variety of different individual and contextual factors (e.g.comorbidities, duration of use/disease, medication/other drugs, or age of onset), and differences in shared representations due to imbalances of pain regulatory systems are just one of them.But some of the strongest protective factors dampening adverse effects and reducing disease burden in chronic pain pertain to the social domain, such as social support, well-functioning social relationships, and social network size.In the biopsychosocial model of pain, such factors are acknowledged not only as outcomes of chronic pain, but also as potential promotors of healing (Cohen et al., 2021).These factors will indirectly suffer from impaired social abilities (such as acting in a prosocial manner), which could be influenced even by subtle impairments of social cognition.Thus, intact affective representations could be viewed as precursors to protective factors in the social domain.
The reliance on different levels of representations of affective states, and their potential impairments could thereby lead to a variety of social interaction problems, depending on whether the affected individual is the observer or origin of an affective state (in most social interactions, individuals will take on these roles alternatingly).Being unable to adequately and fully represent the affective state of another person as an observer will naturally lead to misunderstanding, disappointment or conflict.In a professional medical context, such an observer would potentially fail to provide adequate treatment or information.Being impaired in expressing one's own affective state will similarly lead to negative outcomes, as it would make it much harder for the interaction partner to form an adequate representation of one's affective state and react accordingly.
So far, research on the relationship between shared representations and social behaviors has focused on correlational approaches, linking personality traits or self-report measures of state empathy and prosocial behavior (Cameron et al., 2022).However, in a sample of extraordinary altruists who had previously donated a kidney to a stranger, enhanced sharing of multivariate pain representations was demonstrated, especially in the left aIns and aMCC (O'Connell et al., 2019).This confirmed previously reported results of overlapping aIns activations in the same sample of extraordinary altruists (Brethel-Haurwitz et al., 2018).In contrast, psychopaths, known for their reduced altruism and prosociality, showed decreased multivoxel similarity of activity in the aIns during pain and pain empathy, which downscaled with increasing cold-heartedness, psychopathy, and self-centred impulsivity (Berluti et al., 2020).This suggests that increased sharing of other people's pain is somehow related to altruistic traits, but does not yet allow to link them precisely to a specific representational level.Again, studies employing causal manipulations went one step further by demonstrating that targeted down-regulations of pain representations can lead to highly specific behavioral outcomes.In a combined EEG-TMS study (Gallo et al., 2018), participants could donate money in order to alleviate someone else's pain receiving painful slaps on their hand with a belt.In that study, activity in the primary somatosensory cortex during pain observation explained variation in prosocial helping.Disrupting activity in the hand area of the somatosensory cortex via TMS led to reduced prosocial helping, specifically when viewing only the hand undergoing the painful slap (as compared to also showing the target's painful facial expressions).This speaks for a mechanistic link between shared somatosensory pain representations and pain-related location-dependent prosocial behavior.Further, it highlights that while empathy and prosocial behavior may often be closely linked, they are by no means identical phenomena, are underpinned by different mechanisms, and may be influenced by contextual factors in various ways.In the present case, information on the amount of pain of the target and thus also the need for prosocial help could only be extracted from the trajectory of the slapping belt, while in the abovementioned placebo studies as well as in the facial expression condition, information on the amount of pain was provided by abstract cues or could be extracted from the pain display.
A similar causal approach was taken in a recent preregistered behavioral study employing computational modelling.Here, downregulation of the pain system by means of placebo analgesia led to reduced effortful helping when preventing others from receiving painful electrical shocks (Hartmann et al., 2022).Crucially, a mediation analysis in this study showed that unpleasant affect associated with experiencing the other's pain explained the effect of placebo analgesia on prosocial behavior, which points to reliance on shared representations.These findings demonstrate how a causal manipulation of first-hand pain impacts a type of prosocial behavior that is highly ecologically valid, as real-life prosocial helping is most often associated with effort.
Further hints regarding the link between shared representations and prosocial behavior come from the non-clinical phenomenon of mirrorpain synaesthesia (Fitzgibbon et al., 2010) which is characterized by elevated feelings of vicarious pain.Individuals experience bodily sensations that are qualitatively equivalent to being in pain themselves when witnessing others in pain.On the level of neural differences, the phenomenon has been linked to grey matter volume increases in the insula and the somatosensory cortex, along with volume decreases in the right temporo-parietal junction (TPJ; Grice- Jackson et al., 2017a), which is involved in self-other distinction (see review; Lamm et al., 2019).In an fMRI study on functional connectivity, increased connectivity between the bilateral aIns and the right TPJ (Grice- Jackson et al., 2017b) was found for mirror-pain synesthetes.Regarding prosocial behavior, mirror-pain synaesthesia has been associated with stronger helping attitudes both in a study employing a self-report questionnaire (Botan et al., 2018), and in a more recent fMRI study employing a costly helping task (Fornari et al., 2023).In the latter study, somatosensory brain activity was associated with helping behavior especially when the pain of the other person was conveyed by viewing their pain-receiving hand.This, again, suggests that stronger self-other overlap may motivate prosocial behavior.
The evidence reviewed so far thus highlights that there is no unitary or universal sharing, but that sharing others' pain (and other affective states) takes place at different levels of granularity, ranging from highly pain-specific (or affective state-specific) representations, over more generalized representations of negative affect, to domain-general functions such as arousal.These levels of representations can be identified (as demonstrated by distinct responses of various multivariate patterns of pain, arousal and negative affect (Zhou et al., 2020); see section "Identifying shared representations via neural patterns"), and differentially affected by different kinds of manipulations (e.g., by more or less specific pharmacological manipulations such as opioids or acetaminophen, by cognitive top-down regulation via placebo analgesia, or by TMS; see section 'Modulating shared representations via causal manipulations').
They may therefore be altered by perturbations and imbalances of bodily systems regulating pain and other affective states.Further, variation in these representational levels may lead to level-dependent differences in socio-perceptual and socio-behavioral outcomes: Variations in arousal impact general alertness for others' emotions, while variations in somatosensory pain representations may lead to inaccuracies in interpreting others' pain, potentially even limited to specific body parts.Depending on the context, variations on these levels may have a very different impact on social behavior and interaction.For example, spotting injured persons in a crowd will predominantly rely on the salience of pain expressions, while for a physiotherapist treating pain conditions a targeted focus on the source and impact of nociception and pain will be crucial (see Fig. 1a).
The insights presented underscore that sharing others' pain and affective states is not a monolithic process but occurs across multiple levels of granularity.Practically, this suggests that interventions aimed at modulating empathy or affect sharing need to be tailored to the specific level of representation they target.This has significant implications for clinical practice, particularly in pain management and mental health treatment, where understanding the distinct pathways and representations involved can lead to more precise and effective therapies.On a basic theoretical level these findings and the way we interpret them challenge the notion of a universal mechanism for affective sharing, proposing instead a multi-layered model where different levels of representation are differentially engaged and manipulated.Testing this nuanced view, if confirmed, will enhance our understanding of socio-perceptual and socio-behavioral processes, illustrating how variations in arousal or somatosensory pain representations can influence social interactions and behaviors in context-specific ways.For example, heightened arousal may improve general emotional vigilance in social settings, while specific somatosensory inaccuracies could impact professions requiring precise pain interpretation, such as physiotherapy.Thus, based on these practical and theoretical implications, we propose that recognizing these representational levels enriches our comprehension of the complex dynamics of empathy and social behavior.
Variations and imbalances of pain processing systems may be of particular relevance for pain-related clinical conditions, as pain and disrupted pain-regulatory systems seem to not only affect cognitive skills (see Khera and Rangasamy, 2021 for review), but also appear to affect social cognition due to the reliance of socio-cognitive skills on the aforementioned levels of representations of pain and other affective states.Thus, it is a worthwhile endeavour to investigate and uncover the potential associations between pain-related conditions and deficits in social cognition.

Consequences of altered representations in pain-related clinical conditions
There are different kinds of conditions associated with altered pain regulatory systems, differentiated mainly by their genesis (innate, acquired, substance-induced) and persistence (chronic, transient).Identifying alterations on different representational levels and linking them to social behavior would have epistemic and practical value for such disorders, as it would allow targeted treatments and prevention programs in the social domain.
One particularly interesting condition in this regard is congenital insensitivity to pain, a condition in which affected individuals are strongly impaired in their capacity to feel and correctly interpret physical pain due to an innate deficit at the level of nociception (and a resulting lack of direct experience with the sensory origins of the complex experience of pain).Danziger and colleagues demonstrated that these patients were still able to infer pain from others' facial expressions and verbal descriptions of painful events, but were specifically impaired in assessing painful events in videoclips which lacked explicit painrelated behavioral cues (such as facial or bodily pain expressions) (Danziger et al., 2006).In a follow-up fMRI study, patients showed largely normal neural responses to observed pain in the aMCC and aIns (though smaller clusters of activity were observed in the patient group).Thus, their deficit seemed to be largely restricted to the pain-specific level, while they were still capable of interpreting the emotional significance of the depicted scenarios (Danziger et al., 2009).This suggests that patients were able to correctly interpret painful facial expressions due to their intact representations at the level of negative affect as well as arousal.This fits with the observation that the patients showed a stronger correlation than controls between empathy ratings and activity in brain regions related to mentalizing and cognitive perspective-taking, speaking for the recruitment of compensatory processes and a more abstract-cognitive understanding of the other's pain.To our knowledge, there are no extensive empirical tests of social behavior in patients with this condition.Potentially, these patients would show very limited deficits, for example in types of behaviors that are guided and finetuned by accurate estimation of others' pain expressions or even just nociceptive signals (such as in physiotherapeutic work, or in idiosyncratic settings where they cannot rely on acquired prior knowledge on whether a condition leads to pain and associated negative affect).
Chronic pain conditions are usually acquired later on in life, and are characterized by grey matter volume decreases in bilateral insular and cingulate cortices, and the supplementary motor area (among other regions) (Brandl et al., 2022).While the psychosocial aspects of pain and the need for accordingly adapted treatments of pain have been acknowledged in principle (Darnall et al., 2017), the exact impact of chronic pain conditions on different sub-domains of social cognition and social behavior are currently under-investigated.Clearly, potential deficits in these aspects of cognition and behavior come along with social difficulties.Social relationship quality suffers strongly from chronic pain conditions: chronic pain patients report lower friendship quality, as well as higher perceived hostility and rejection (Philpot et al., 2020).Fig. 1.Overview of relationship between representational levels and potentially affected prosocial behaviors per level.(a) Arousal: encompasses a domain-general response captured e.g. by activation of the sympathetic autonomic nervous system or increased attention to the person in pain; arousal itself may lead to targeted empathy and prosocial behavior, but it may be as unspecific as "this person is in a negative affective state and needs help" (see right column for examples); examples of involved brain areas: networks involved in triggering autonomic arousal, such as the amygdala and hypothalamus, but also downstream areas representing and regulating autonomic responses.Affect: encompasses representing the valenced affect of the other person, such as in feelings of unpleasantness in response to an aversive emotional state; it may lead to more targeted (i.e.state-adequate) empathy and prosocial behavior, on the level of the affective state; brain areas: networks involved in valenced affective processing and affective awareness, such as e.g.ventro-medial prefrontal and striatal valuation, or insular and cingulate areas.Painspecific: encompasses representing the specific characteristics of the other person's pain experience, including sensory aspects such as the location on the body and the quality of the pain; it may lead to very targeted empathy and prosocial behavior, such as acts of support directed at the affected body part; brain areas connected to nociceptive pain, such as those covered in the NPS (Wager et al., 2013).In the left column, we summarize causal manipulations mentioned in the review.Depending on their target, manipulations like TMS and tFUS may exert effects on different representational levels.(b) Presumed degree of impairments in CIP, chronic pain, and opioid use disorder (color coding indicates the main level presumed to be affected).Note that this is just a rough assumption, as chronic pain and opioid use disorder encompass a heterogeneous array of clinical conditions.Abbreviations: CIP = Congenital insensitivity to pain; TMS = transcranial magnetic stimulation; tFUS = transcranial focused ultrasound stimulation.

Painspecific
Moreover, these conditions are strongly related to difficulties in identifying own emotional states and in regulating positive and negative emotions (Aaron et al., 2020;Koechlin et al., 2018).One such chronic pain condition is fibromyalgia, a complex disorder characterized by widespread chronic musculoskeletal pain that is presumably caused by dysfunction of pain regulatory systems (Häuser et al., 2015).An fMRI study in fibromyalgia patients showed lowered neural responses (activity in thalamus, ACC, DLPFC, SMA) to empathic pain when viewing hands in painful situations, as compared to healthy controls (Lee et al., 2013).A magnetoencephalography study found similar results for somatosensory pain representations: while healthy controls showed clear differentiation between somatosensory brain responses to painful vs non-painful pictures, fibromyalgia patients did not (Goldstein et al., 2019) (see also de Tommaso et al. ( 2019)).Such reduced neural differentiation (particularly in somatosensory cortices) between painful and non-painful states has recently been shown to be predictive of reduced empathic concern for others (Lulla et al., 2024).These findings indicate that the partial dysfunction of patients' pain regulatory systems extends to pain empathy (as shown by the reduced differentiation between painful and non-painful pictures), and affects neural pain representations on the affective and somatosensory levels.A behavioral study suggested that the socio-cognitive alterations may be even more widespread, affecting emotion recognition in the negative domain and theory of mind (Di Tella et al., 2015), highlighting that a broader range of representations and social behaviors may be impacted.Again, causal experimental approaches as well as evidence on putative behavioral implications are currently absent.Judging from the available findings, a wider range of behaviors is expected to be affected, involving situations that require estimation not only of others' pain, but also negative affective states in general, as well as other, more complex mental states.If patients with fibromyalgia are thus less empathic to others' pain, this could have adverse effects on their social interactions, and, consequently, on their pain status by weakening protective factors in the social domain.Admittedly, chronic pain conditions are multi-factorial and typically associated with mood and anxiety comorbidities, which could make it difficult to isolate the specific role of pain representations in social deficits.Depending on other patient characteristics and capacities, compensatory mechanisms could make up for potential deficits in sharing others' affective states.
Finally, the impact of opioid use and abuse is of particular concern and in need of more concerted research efforts in light of the surge of pain-related clinical conditions, their management and the associated opioid crisis (see Volkow and Blanco, 2021 for review).If empathy is indeed grounded in various self-related pain processes and mechanisms, impairments in empathy are also to be expected in long-term abuse of opioidergic drugs, as this leads to an imbalance of the opioid system (Ballantyne, 2018;Toubia and Khalife, 2019) and correspondingly disrupted responses to painful stimuli.In line with that, non-medical prescription opioid users showed reduced ratings of affect sharing and pain evaluation in response to seeing others in pain (Kroll et al., 2021).While the latter study appears to be the only (published) investigation of pain empathy in chronic opioid abuse, there is considerably more evidence on other socio-cognitive functions.Indeed, disrupting the balance of the opioid system via long-term intake of opioid drugs affects other areas of social cognition, as this system is involved not only in pain processing, but also in reward (Laurent et al., 2015) and in regulating stress (Drolet et al., 2001), partly via interactions with other neuromodulatory systems (van Steenbergen et al., 2019).It further shows strong overlap with human emotion circuits and modulates affiliative behaviors (Nummenmaa and Tuominen, 2018).Thus, alterations of a broad array of social functions due to imbalances of this system are conceivable and corresponding evidence has been growing over the course of the last years.Studies employing the multifaceted empathy test (Dziobek et al., 2008) in regular opioid users demonstrated reduced sharing of positive affect (Carlyle et al., 2020) and worse mentalizing performance (Kroll et al., 2018).The latter investigation also showed dose-dependent effects: heavy users (with highest morphine hair concentrations) were worst at mentalizing.Similar results were found in opioid-dependent patients on maintenance therapy, who expressed impaired mentalizing abilities in three studies employing different social inference paradigms (Eidenmueller et al., 2021;Gandolphe et al., 2018;McDonald et al., 2013), and two studies using the reading the mind in the eyes test (RMET; Terrett et al., 2020;Yazgec et al., 2023).In the study by Terrett and colleagues, reduced facial mimicry was also demonstrated, and interpreted as an indication of lower affect sharing.Notably, causal experimental research on many aspects of social cognition (both at the core of this review, but also beyond) in these populations is scarce, and there are no neuroimaging studies explicitly testing shared representations (or their link to social behaviors) in them.We are also unaware of any studies on prosocial behavior in this context.Thus, we can only speculate about potential effects on social behavior and real-life consequences.Of note, though, social interaction difficulties are frequently reported by opioid users (Ashrafioun et al., 2022), and opioid use disorder and social isolation appear to be closely linked (see Christie, 2021, for review).Opioid maintenance patients also displayed higher social rejection sensitivity in the cyberball task, which was furthermore related to smaller gray matter volume in the bilateral insula and inferior frontal gyri (Bach et al., 2019b).In another study employing the same task, a sample of opioid maintenance patients reported more feelings of social exclusion in inclusion trials, and, in contrast to controls, did not show activation differences between exclusion and inclusion trials in the aMCC, insula, or fusiform gyrus (Bach et al., 2019a).Lastly, hyperreactivity of the hypothalamic-pituitary adrenergic stress axis in response to social exclusion was found, suggesting dysfunctional physiological responses to psychosocial stressors (Kroll et al., 2019).
In summary, the abovementioned clinical populations show distinct deficits that are in line with the proposal that deficits at different levels of pain processing systems lead to more or less pronounced deficits in empathic responding (see Fig. 1b).Highly focused deficits such as the nociceptor dysfunction in congenital pain insensitivity appear to have a limited effect, which can be compensated, while systemic imbalances tapping into affective regulation appear to have more widespread effects.In line with the strong link between affect sharing and prosocial behavior, those groups would be expected to be less prosocial to some degree (CIP the least, opioid abusers the most).These deficits would potentially lead to a vicious cycle, as they would make it harder to connect with others, which could further weaken protective social factors and increase the risk for opioid use or symptom severity in chronic pain.In the same vein, differences in social functioning could be an underlying cause or starting point for the development of conditions such as chronic pain or opioid use disorder, instead of being a consequence.Relatedly, such deficits may also be a severe issue for professional pain management, as failing to adequately communicate affective states and failure to solicit empathy by their caregivers due to difficulties in social interaction may impair treatment efficacy of pain patients.The evaluation of a patient's pain in health care professionals depends on a variety of individual, contextual and relational factors (Coll et al., 2011) and may be particularly sensitive to impairments in a patient's social interaction abilities.Of note, healthcare experience has been shown to lead to decreased pain-specific responses to others' pain in the anterior insula (Corradi-Dell'Acqua et al., 2023).
The wide-open research gaps we identified in this section need to be filled in order to make progress on two important epistemic as well as practical issues: understanding the link between shared representations and social behavior in general, and understanding the interpersonal and societal impact of pain conditions and their pharmacological management.Research on patients with clinical conditions comes along with some difficulties, of which comorbidities are among the greatest issues, especially in drug use disorders.In the concluding section we focus on how we can advance our understanding of how shared representations inform and shape (pro)social behavior.

Future research avenues and implications
Our selective review of the literature shows that research has met several prerequisites for investigating and better understanding graded effects of different levels of shared representations on prosocial behavior.Researchers have recently operationalized and disentangled the different levels underlying painful/affective experiences, and also demonstrated that features on these pain-specific and domain-general levels can be modulated by different manipulations.An important next step would be to demonstrate more specifically and explicitly how differences in shared neural representations (brought about either by clinical conditions or experimental manipulations) influence (pro)social behavior.Our review also revealed a number of open questions and research avenues, which should be addressed in order to refine our knowledge of shared representational levels and their relationship with social behavior.Answering these questions will yield important mechanistic insights of relevance for basic research and the above-mentioned clinical conditions alike.
First and foremost, we need systematic research that connects the levels of shared representations, from pain over affect to arousal, to tasks probing different social behaviors in within-participants designs and in the same experimental sessions.This would explain how these levels drive different social behaviors, for example by showing whether and to which extent each of these levels contribute to overt behaviors such as pain-related helping, prosocial effort, or obtaining rewards for others.This could be achieved by standardized batteries of representational paradigms (involving self-and other-directed conditions for all tested levels/functions) and experimental measures of (pro)social behavior.
Such an approach would allow to go beyond merely linking empathic traits (or extreme forms such as extraordinary altruism or psychopathy) to differences in shared representations, by e.g.relating objective multivariate patterns (covering nociception, pain, arousal (Zhou et al., 2020), negative affect (Chang et al., 2015), or empathic distress (Ashar et al., 2017)) instead of subjective self-reports to experimental tests of social behavior, ideally by applying computational modelling approaches and techniques such as reinforcement learning.Recently, initial attempts of such approaches have been used, e.g. in more ecologically valid tasks that investigate social behaviors such as benefitting others (Lockwood et al., 2017), effortful helping to reduce someone else's pain (Hartmann et al., 2021b), resolving moral conflicts (Fornari et al., 2023), or incorporating social feedback in reciprocal interactions (Shamay-Tsoory and Hertz, 2022).Applying computational modelling to these tasks and the related behaviors will also offer the opportunity of achieving a more mechanistic or at least process-level understanding of helping (and other) behaviors.Further, modelling allows to relate certain parameters of behavior to self-report and neural activity acquired in representational paradigms, via techniques such as model-based fMRI (O'Doherty et al., 2007).Also, combinations of experimentally controlled approaches with research under real-life conditions, using experience-sampling methods, seem particularly promising to explore how shared representations inform and shape social behaviors (Stijovic et al., 2023).These approaches may be especially well suited for chronic pain conditions to monitor daily fluctuations in pain intensity and experiences in social interactions.
Second, research on prosocial behaviors is especially scarce in clinical populations, although they could be easily implemented in behavioral studies as a first step.Their relation to other protective factors in the social domain (e.g.social network size) would be particularly interesting.As summarized above, there are no studies explicitly testing prosocial behavior in the reviewed clinical populations (congenital pain insensitivity, fibromyalgia and opioid users).Yet, each of these populations offers a different angle on the connectedness between shared representations and prosocial behavior, as they differ with regard to the affected representational levels and potential consequences associated with them.For instance, research into congenital pain insensitivity could shed light on how sensory impairments impact prosocial tendencies in environments requiring precise assessments of localized pain responses, elucidating the extent to which reliance on somatosensory information is indispensable in such contexts.The subgrouping of opioid users, spanning from recreational users to those with opioid use disorder, represents yet another broad spectrum of scientific scrutiny.Key questions pertain to the extent to which prolonged opioid use or maintenance precipitates specific deficits, the potential for recovery to normative socio-affective capacities among former users, and the ramifications of current intoxication.Insights into the consequences of prolonged opioid use and intoxication extend beyond the immediate concerns of opioid use disorder, offering implications for broader questions concerning the influence of various pain medications, such as ubiquitous analgesics like acetaminophen, on empathic responses, prosocial conduct, and societal dynamics.Tentative evidence linking indiscriminate analgesic consumption to diminished empathic concern and altruistic behaviors has emerged from recent survey research (Banwinkler et al., 2023).
Apart from clinical populations, sub-clinical conditions such as mirror-pain synaesthesia may constitute promising fields of research, along with more research into the relationship between first-hand pain sensitivity and sensitivity for the pain of others.
Finally, targeted causal interventions should be combined with the abovementioned approaches.Novel deep brain stimulation techniques such as transcranial focused ultrasound stimulation hold great potential in this respect, as they allow modulation of activity in brain structures such as the insula (Penton et al., 2022) that are not easily accessible to other brain stimulation methods, and due to the causal inferences they allow us to draw.In the domain of self-experienced pain, López-Solà and colleagues have provided an elegant example of testing the impact of a causal intervention on multivariate patterns (López-Solà et al., 2022): they tested the effects of an anti-inflammatory drug on three different signatures, and could demonstrate pain-specific effects on the NPS in knee osteoarthritis patients versus no modulation of signatures specific to negative affect (Chang et al., 2015) or distress (Ashar et al., 2017).
In conclusion, the field of social neuroscience has generated novel and increasingly refined as well as more mechanistic insights into the neural underpinnings of empathy and affect sharing.However, as this review has shown, we are still a long way from fully understanding if and how precisely the various levels of empathy and affect sharing influence our behavior towards others.Focusing future research on these key gaps of knowledge will not only advance our basic understanding of this central social skill, but also inform interventions targeting empathyrelated deficits in health and disease.

Declaration of Competing Interest
None.