The Role of Likeability in Discriminating Between Kindness and Compassion

Kindness and compassion are prosocial constructs aimed at benefiting others, with the former focused on happiness and the latter on suffering. Despite these distinctly different motivations, kindness and compassion are often used interchangeably. If compassion and kindness are different processes, they should respond differently to the same facilitators and inhibitors, with a key moderator being likeability. We used a cross-sectional survey design to examine whether a target that differed in terms of likeability (liked versus disliked) influenced willingness to engage in kind acts compared to compassionate acts, and the emotional patterns experienced. We recruited 150 participants (83 men, 66 women, 1 other; Mage = 27.85, SD = 10.21) using an online survey platform. Participants reported less willingness to engage in acts of kindness compared to acts of compassion regardless of target likeability. However, this reduction in willingness was markedly greater for disliked targets. Compassionate acts towards liked targets were associated with significantly higher levels of negative emotions (e.g., irritation, sadness, anger, anxiety, and disgust) when compared to kind acts. Conversely, compassionate acts towards disliked targets elicited less feelings of irritation and anger compared to kind acts. These findings indicate that kindness and compassion result from separable motivational systems, differing in both the emotions elicited and the willingness to act. Reluctance in helping disliked others is reduced when the action is aimed at reducing suffering.

The motives to be helpful to others differ in terms of function (why) and form (how) (Curry et al., 2018;Eisenberg et al., 2016;Gilbert, 2019;Phillips & Taylor, 2009). Previous studies have found that common expressions of compassion and kindness can vary in terms of function and form, having different associated intentions and emotions . Although compassion and kindness are terms that are often used interchangeably, both student and community populations can readily discriminate between scenarios that are motivated be kindness (e.g., remembering a birthday) and compassion (e.g., donating a kidney; Gilbert et al., 2019). One way to consider the differences between these prosocial motivations is based on the evolutionary origins and regulators of helpful behavior. In a metaanalysis of kindness, Curry et al. (2018) defined kindness as, "actions intended to benefit others" (p. 321), which could be influenced by a range of different motivations, including kin and reciprocal altruism. Helping others is, in evolutionary terms, an expensive resource with built-in biases towards preferentially offering it to kin and potential reciprocating others (Curry et al., 2013). For example, given a choice to feed one's own children or equally needy strangers, one is far more likely to choose the former (Buss, 2014;Curry et al., 2013). Phillips and Taylor (2009), and Ballatt and Campling (2011) also linked kindness to its root, which is "treating others like kin." As such, kindness is the most frequently and easily directed towards family, friends, colleagues, and spouses (Curry et al., 2018).
Examining kindness from a different perspective, in Buddhist and contemplative traditions, kindness forms part of the four immeasurables, which include metta (lovingkindness), karuna (compassion), equanimity, and appreciative joy. Ricard (2015) stated in his review of altruism that loving-kindness, or metta, is focused on the desire for others to flourish and be happy. In contrast, the primary focus of compassion is to alleviate and prevent suffering (Ricard, 2015). This distinction between the underlying motivations behind loving-kindness and compassion is a view shared among many scholars (e.g., Feldmen & Kuyken, 2011;Gilbert, 2019;Goetz et al., 2010;Mascaro et al., 2020). The Dalai Lama (2001) also distinguished kindness (metta) and compassion (karuna): What is compassion? Compassion is the wish that others be free of suffering. ….. Just as compassion is the wish that all sentient beings be free of suffering, loving-kindness is the wish that all may enjoy happiness (p. 96).
Outside of the Buddhist literature, compassion itself can be traced to the Latin root pati, which means to suffer, with the prefix com meaning with, thus compassion or compati means to suffer with (Goetz et al., 2010). In the scientific community, there have been several review papers examining the differing definitions of compassion (Gilbert, 2019;Goetz et al., 2010;Strauss et al., 2016), and although there is no universally agreed definition, all emphasized the importance of (1) engaging with suffering and (2) acting to alleviate it (Kirby, 2016;Mascaro et al., 2020). A definition of compassion that captures these two aspects is provided by Gilbert (2014), who defined compassion as, "the sensitivity to suffering in self and others with a commitment to try alleviate and prevent it" (p. 19).
Compassion is theorized to be helpful for well-being and relationships (Gilbert, 2019), and cross-sectional research has examined compassion using self-report scales and found significant associations between higher levels of compassion and lower levels of loneliness (Best et al., 2021), higher levels of social connectedness (Seppala et al., 2013), and higher levels of psychological well-being (Zessin et al., 2015). Other studies have examined self-reported compassionate behaviors, with one study of 175 newlywed couples finding significant associations for both the giver of compassionate behavior and receiver of compassionate behavior in terms of lower negative affect and higher overall life satisfaction, with the giver also experiencing higher positive affect (Reiss et al., 2017). Given these positive associations, it is unsurprising that several compassion-based interventions have been developed to specifically cultivate compassion to help improve overall mental health and well-being, with a meta-analysis of these interventions finding they significantly improve self-reported levels of compassion, as well as reduce self-reported symptoms of depression, anxiety, and distress .
According to evolutionary models, motives need two core processes to function successfully (Buss, 2014;Huang & Bargh, 2014). In the case of compassion, the first is to be able to detect signals relevant to the motive (e.g., suffering), and the second is to have a response that helps to successfully enact the motive (e.g., an action response; Kirby & Gilbert, 2017). These two processes, signal detection and signal response, are central to all motives be it feeding, sexuality, harm avoidance, or competing for resources (Buss, 2014;Gilbert, 2014;Huang & Bargh, 2014). All motives, whether it be harm avoidance, sexual, or competitive, can be both facilitated and inhibited (Neel et al., 2016). Thus, if compassion and kindness are different motivations, they will potentially respond differently to the same facilitators and inhibitors.
One key moderator of prosocial motivations is likeability. However, there is little empirical evidence supporting the view that compassionate and kind motivations differ based on whether the target for such actions is liked. Evolutionary models posit that prosocial behavior, that is helping to benefit others, is "costly" and not equally provided based on need alone (Colqhoun et al., 2020). Individuals are treated preferentially, such that we are more likely to perform prosocial acts for those whom we like than those we dislike regardless of need (Schreuders et al., 2018). One model to describe how likeability influences prosocial intention and behavior is provided by Loewenstein and Small (2007), who suggested we act prosocially based on deliberative or emotional processes.
The contributions of emotional and deliberative processes to prosocial decision-making can be partly inferred from the emotions and meaningfulness associated with performing said actions. Engaging in compassionate acts is associated with higher reported emotional intensity for disgust, anger, sadness, and fear when compared to kind acts, which are associated with greater joy . Similar patterns of emotional experiences are seen when comparing meditative practices focused on compassion and kindness. In a study of 201 university students, Sirotina and Shchebetenko (2020) randomized participants to undertake either a 15-min single session of loving-kindness meditation, compassion meditation, or a control task. Assessing how these prosocial meditations impacted reported emotional experiences, loving-kindness meditation led to significantly higher happiness and positive emotions than the compassion meditation.
The distinction between these emotional patterns has been verified by functional brain imaging (Cutler & Campbell-Meiklejohn, 2019;. Kindness typically elicits greater activation in the dorsolateral prefrontal cortex and orbito-frontal cortex (Weng et al., 2018), whereas compassion tends to elicit increased activation of threat processing regions including the anterior cingulate and anterior insula Kim, Parker, et al., 2020). Given the primary focus of compassion is on suffering (Gilbert, 2014), it is unsurprising that compassion is associated with increased threat processing and self-reported emotional threat responses (e.g., fear, disgust). In contrast, kindness need not involve suffering but rather an anticipation of what would make somebody happy, flourish, and bring joy (Ricard, 2015). Although positive emotions such as joy and happiness are typically experienced to a lesser extent for compassionate compared to kind acts, compassionate acts have been reported as more meaningful . This pattern may reflect a greater emphasis on deliberative cognitive empathy in the evaluative process. If so, this supports the view that such deliberative processes may buffer against the negative bias associated with acting prosocially towards disliked others.
The proposed advantages of compassionate motivations over kind motives for disliked targets are likely mediated by an individual's beliefs about compassion and compassionate acts. Resistance to compassion is not uncommon, with some believing it is a form of pity, indulgence, or a sign of weakness (Kirby, Day, et al., 2019;Kirby, Seppälä, et al., 2019). Gilbert et al. (2011) coined the term "fears of compassion" to refer to the avoidance of fear responses individuals can have towards expressing compassion. For example, individuals can fear expressing compassion towards others because their compassionate efforts might be seen as incompetent or unhelpful, and may be rejected (Gilbert & Mascaro, 2017). A reluctance to engage in compassion can also stem from the fear one might suffer too much personal distress and be overwhelmed by those in need of compassion (Vitaliano et al., 2003) or the fear that compassion will be perceived by others as a form of manipulative self-interest (Gilbert & Mascaro, 2017). An additional fear of being compassionate may derive from the cost of effort involved and the possibility of resource loss Gilbert et al., 2011).
A recent paper examined fears of compassion, as well as empathy, mindfulness, and compassion, to determine how each individually predicted a morally expansive mindset. A morally expansive mindset is a psychological construct which refers to how an individual determines others (e.g., human and non-human entities) are worthy of moral concern and treatment (Crimston et al., 2016). It is based on Singer's (1981) conceptualization of the moral boundary, which is a demarcation that distinguishes between those individuals who are worthy of our moral concern and help and those that are not. For example, those we dislike are often considered outside of our immediate moral circle and on the fringes of our moral concern, whereas people we do like are at the center of our moral circle and concern (Crimston et al., 2016). Across three studies comprising 749 participants, Crimston et al. (2022) found empathy and mindfulness significantly predicted moral expansiveness, but were no longer significant predictors when compassion and fears of compassion were entered in hierarchical regressions. Both compassion and fears of compassion predicted significant unique variance in moral expansiveness, with fears of compassion to others being the strongest predictor of a more restricted moral circle. As compassion is proposed to draw upon deliberation and cognitive empathic pathways, relying to a greater extent on moral value than emotional responses, this may explain why trait compassion is a greater predictor of moral expansiveness than mindfulness and empathy. The authors concluded that if we wish to encourage greater moral expansiveness, compassion could be the most effective pathway to have this realized. However, another pathway not examined was kindness.
Here, we used a cross-sectional survey design to examine whether the likeability of a target affects the willingness to perform compassionate and kind acts, and the emotional experiences associated with each act. We hypothesize that likability will have a significant impact on preparedness to do acts of kindness but less on acts of compassion when the target is disliked. We also hypothesize that compassionate acts will be associated with greater negative emotional experiences than kind acts. Finally, we hypothesize that fears of compassion will be negatively associated with willingness to act.

Participants
A total of 150 participants (83 men, 66 women, 1 other; mean age = 27.85, SD = 10.21) were recruited on the online survey platform Prolific (www. proli fic. co). A power analysis was conducted based on the effects found in the previous study by Gilbert et al. (2019), which indicated we needed at least 138 participants to ensure a 95% chance of detecting a medium effect size. The study was pre-registered with the data set available on the Open Science Framework: https:// osf. io/ cmfu9/. The study received ethics approval from The University of Derby Ethics Committee (ethical clearance number: 5 61-16/17). There were no eligibility criteria for participation except being over 18 years of age and the ability to understand written English.

Procedures
All research participants who wished to participate were provided an information sheet with an explanation of what the study involved, and they all confirmed they had read and understood this and provided informed consent by completing the consent form. Participants were asked to bring to mind a person they know and either like or dislike. Next, they were asked to write down that person's name, rate their level of liking for them from 1 (do not like at all) to 10 (extreme liking), and write a sentence explaining why they like or dislike the chosen person. Participants were then informed they would be completing a series of questions which asked about their willingness to do certain acts for a target person (the Kindness and Compassion Scenarios Scale), which would be the person they wrote about liking or disliking. Participants were also asked to complete the reactions to Kindness and Compassion Scale for each scenario. The order of target type was counterbalanced between participants. Participants were then asked to complete the fears of compassion to other subscale. The only demographic data collected were gender and age.

Kindness and Compassion Scenarios Scale
The Kindness and Compassion Scenarios Scale was developed by Gilbert et al. (2019), who used the definitions of compassion as being the alleviating and preventing of suffering, and kindness as being promoting flourishing and well-being. The original scale comprises 18 scenarios, with 10 scenarios focused on kindness and eight on compassion. For this study, 12 were selected as they were applicable to a liked and disliked target. The other six scenarios were removed as they required the target to be married or have children, and the desire was for each scenario to be relevant to any individual. The scale was also adapted to ask participants to indicate their willingness to engage in a prosocial act towards a person they personally knew that they liked or disliked in relation to each scenario (see "4" for further details). Six scenarios assessed willingness to engage in acts of kindness (e.g., "Doing a favor for them that takes up your time") and six assessed acts of compassion (e.g., "Donating a kidney to save them"). Responses were rated on a scale from 1 (not at all willing) to 10 (extremely willing). Final scores for kindness and compassion were calculated as the mean score across the respective scenarios. The kindness and compassion scenarios for both target types had acceptable to good internal reliability (ω = 0.70-0.84).

The Reactions to Kindness and Compassion Scale
This scale was used by Gilbert et al., (2019) to examine the different reactions participants had to each scenario of compassion and kindness in the Kindness and Compassion Scenarios Scale. In the Gilbert et al.'s (2019) study, the scale was titled "The Emotions of Kindness and Compassion Scale." However, the scale comprises more than emotions as it also asks participants how meaningful they feel it would be to act either compassionately or kindly in each scenario. Participants are asked to rate the extent to which they felt the following emotions: anger, anxiety, disgust, irritation, sad, and joy. Each emotion was rated on a scale from 1 (not at all) to 9 (very much). Participants were also asked how meaningful it would be to do the act on the same response scale. The scores for each reaction were then averaged across the kindness and compassion scenarios, respectively. Thus, 14 summary scores were calculated, seven for the kindness scenarios and seven for the compassion scenarios. The

Fears of Compassion to Others
The Fears of Compassion to Others subscale (Gilbert et al., 2011) measures the fears people have about being compassionate to others. Individuals were asked to rate the extent of their agreement from 0 (don't agree at all) to 4 (completely agree) for 15 statements (e.g., "people will take advantage of me if they see me as too compassionate"). Final scores were calculated as the mean across items. The scale had good internal reliability (ω = 0.85).

Data Analyses-increase font size to same as Participants, Procedure, Measures
To test for the role likeability (liked vs. disliked) had on willingness to act in kindness and compassionate scenarios, we performed a series of within-participant analysis of variances (ANOVAs). We also conducted a series of paired samples t-tests to examine the difference in reactions (emotions and meaningfulness) to these scenarios, and applied Bonferroni adjustments to control for multiple comparisons. Finally, we conducted linear regressions to determine whether fears of compassion negatively predicted willingness to behave prosocially.
To analyze the emotional profiles associated across target and scenario types, the intensity of each emotion felt was averaged for kindness and compassion scenarios separately. Difference scores in the level of emotion felt for the scenario types were then calculated by subtracting the average level of emotion felt for disliked targets from the average level of emotion felt for liked targets. A series of paired samples t-tests were conducted on these difference scores to determine if the discrepancy in strength of emotions felt towards liked versus disliked targets differed as a function of whether the scenario was kindness-or compassion-based (see Table 1).
As predicted, the general pattern in emotional responses replicated Gilbert et al.'s (2019) findings (see Fig. 2). Compared to kindness scenarios, compassion scenarios elicited higher levels of negative-based emotions: irritation, anger, disgust, anxiety, sadness. In contrast, compared to the compassion scenarios, the kindness scenarios elicited higher levels of joy. Furthermore, individuals felt higher irritation, anger, anxiety, and disgust about behaving prosocially towards disliked compared to liked others. This discrepancy was significantly stronger for the kindness scenarios. Contrasting this, individuals felt higher levels of sadness towards liked others in the compassion scenarios while higher levels of sadness were reported towards disliked others in the kindness scenarios. Individuals felt more joy and meaning about behaving prosocially towards liked compared to disliked others. Again, this discrepancy was significantly stronger for the kindness compared to compassion scenarios.
To determine whether fears of compassion negatively predicted willingness to behave prosocially, two linear regression models were hypothesized with fears of compassion to others as the predictor and willingness to behave prosocially towards liked others and disliked others as outcome variables. No other variables were included in the regression.

Discussion
The aim of this study was to determine whether the willingness to act with kindness or compassion differs between liked and disliked targets. The key finding from our study was that likeability plays a greater role in the willingness to act kindly compared to acting compassionately. Specifically, having a disliked target diminishes willingness to act prosocially more for kindness scenarios than for compassion scenarios, supporting our hypothesis. In other words, the reduction in willingness that occurs due to disliking the target is far greater for kind acts than compassionate acts. This finding suggests kind acts are more likely reserved for those that are liked and held positively in mind when compared to compassionate acts. Overall, participants were significantly more likely to act compassionately than kindly for all targets, although this finding was smaller for liked targets.
This result has important implications when considering moral expansiveness and effectively expanding our moral circle to others beyond our typical in-group (Singer, 1981), as compassionate acts were more likely than kind acts to be extended to those targets we dislike. Our finding also lends support to the previous work by Crimston et al. (2022) who found compassion was more important than mindfulness and empathy at predicting moral expansiveness. Our study found the willingness to act compassionately towards a disliked target was not as significantly impacted as it was for kindness. Although compassion and kindness are both critical in facilitating prosocial behavior, it is important to consider how these words may be used interchangeably in everyday life. For example, if there is an element of dislikeability between people and attempts are made to improve relationships where there is tension and conflict, emphasizing compassion might be the more likely successful pathway to prosociality and improved relationships compared to kindness.
A second key finding is that participants reported different emotional experiences when considering helping a liked compared to disliked target. Helping a disliked target with kind acts was associated with significantly higher intensity of negative emotions such as irritation, anxiety, anger, and disgust compared to acts of compassion. This is interesting as the kind scenarios did not include aspects of suffering, whereas the compassionate scenarios did. Research has found that engaging in compassionate acts is associated with negative emotions due to the suffering encountered . Thus, the thought of doing kind acts for disliked targets was experienced as more emotionally negative than the emotional experience that occurs when acting in scenarios of compassion where suffering is encountered. There were also greater levels of sadness experienced for participants when acting with kindness for the disliked target, whereas the opposite pattern was found for compassion. This may indicate that when acting compassionately towards liked others, we tend to feel a sense of sympathy or sorrow that the other is suffering. Many definitions and conceptualizations of compassion include sympathy, a feeling of concern, or sorrow as part of its definition (Strauss et al., 2016).
Confirming past research, we found acting kindly to liked targets was associated with more joy than compassion Sirotina & Shchebetenko, 2020). However, sense of meaning was higher for compassionate acts for both liked and disliked targets when compared to kindness. Intriguingly, when it came to disliked targets, higher levels of joy were associated for compassionate acts rather than kind acts. Although the intention of goodwill and happiness is at the core of kindness, when this intention is directed to disliked targets, joy appears to be diminished considerably. Indeed, reported joy decreased to floor levels when directed towards a disliked target (see Table 1, M = 1.84; SD = 1.04), indicating the significant impact likeability has on this prosocial motive. To an extent, this can be linked to the concepts of hedonic and eudaemonic well-being. Hedonia is focused on pleasure, including experiencing feelings of happiness and joy, whereas eudaimonia is focused on meaning and purpose as the sources of well-being (Ryan & Deci, 2001). When it comes to extending prosociality towards those we dislike, focusing on the sense of purpose and meaning it affords might reduce the barriers to extending our prosociality to others typically outside of our immediate moral circle such as those we dislike. This links to the deliberative pathway of Loewenstein and Small's model (2007) whereby in contexts of suffering, compassion is more readily able to consider both the emotional and deliberative components when helping someone who needs it. In contrast, kindness could be relying more on the emotional pathway, particularly given kindness is most often reserved for kin or potential reciprocating others where deliberation is less needed (Curry et al., 2013). Put another way, when targets are disliked, this "turns down" the capacity for those to draw upon deliberative processes (e.g., cognitive empathy) to enact kindness compared to compassion. Future work manipulating factors that can inhibit or facilitate kindness and compassion, such as cost, similarity, or blameworthiness (Loewensetin & Small, 2007), could provide further insights into the pragmatic mechanisms of these two prosocial motivations.
These distinctions suggest a range of implications not only for the study of prosocial behavior but also for areas of work where compassion and kindness are central. For example, in caring professions, staff may try to behave professionally in addressing suffering but may be "less kind" to those patients they dislike. It is also possible that kindness is reflected in friendliness, such that staff appear more friendly to liked rather than disliked clients. When developing compassion programs or "empathy training" programs for staff, the dimensions of likability should incorporate skills on how to empathize with "the disliked." Taking a slightly different perspective and moving from other focused to self-focused, many with mental health difficulties present to therapy about aspects of themselves they dislike. An emerging area of therapeutic intervention is the use of Loving-Kindness Meditations (Galante et al., 2014) and compassion-focused therapies (Gilbert, 2010) to help alleviate client suffering. Given many with mental health difficulties have high levels of shame and self-reported selfdislike or even hate, clients may experience interventions that are aimed to generate self-kindness differently when compared to self-compassion. For example, helping individuals who dislike themselves to be "kind" to themselves can produce resistance and negative affect in the same way that being kind to disliked people does. However, if they can recognize compassion is different to kindness and compassion is about developing the courage and wisdom to address suffering, they might be more likely to engage with the interventions (Gilbert & Simos, 2022). Developing self-kindness is still important but may come at a later stage after the clients have worked through the nature and origins of their self-dislike.

Limitations and Future Research Directions
As with all studies using cross-sectional self-report survey designs, intentions do not necessarily provide accurate predictions of actual behavior. Hence, these dimensions require further experimental work, preferably in real-world conditions. Future work could also control for differences in urgency, severity, and importance between the kindness and compassion scenarios. If the key difference is the concern with alleviating suffering compared to promoting well-being, then the effect should still emerge if the kind and compassionate acts are equally serious/costly. We have argued compassionate acts are inherently more costly than kindness acts, but there could be situations where this might not universally hold. For example, paying for someone's cancer treatment (an act of compassion that reduces suffering) versus paying for the person to go to college (an act of kindness that promotes well-being). Although we predict the difference would still emerge (people would be more willing to pay for cancer treatments of those they dislike than pay for their college), this is an empirical question requiring additional work to fully test.
Future work can also continue to examine the psychometric validity of the scales used in the study, specifically The Kindness and Compassion Scenarios Scale, as well The Reactions to Kindness and Compassion Scale, with other psychometrically valid prosocial measures. Finally, all variables in the study were measured through self-report, which may have contributed to common method bias, potentially influencing the results (Podsakoff et al., 2012). Despite these limitations, this follow-up study to Gilbert et al. (2019) confirms likeability is a key factor that inhibits the prosocial motivations of kindness and compassion differently, with willingness to act kindly being inhibited to a significantly greater degree than the willingness to be compassionate to disliked targets. Taken together, these findings confirm that kindness and compassion are tapping different motivational systems involving different emotional patterns, with different levels of willingness to invest time and energy to act.
Author Contributions JNK and PG designed the studies. RG recruited participants and analyzed data. JNK, PG, RG, and CS all contributed to the writing of the manuscript.

Funding Open Access funding enabled and organized by CAUL and its Member Institutions
Data Availability Data are available on Open Science Framework: https:// osf. io/ cmfu9/.

Ethics Approval
The study received ethics approval from The University of Derby Ethics Committee (Ethical clearance number: 5 61-16/17).

Conflict of Interest
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.