Research on students’ well-being revealed that many experienced loneliness before, during, and after the COVID-19 pandemic, with 29–34% of students reporting high levels of loneliness (Akram et al., 2023; Limarutti et al., 2023; Sivertsen & Johansen, 2022). This sense of loneliness can be enhanced by social exclusion, belittling, and rumours, typical bullying actions among students (Buglass et al., 2021; Cowie & Myers, 2015). Students purposefully left out by peers lose their sense of belongingness (Lund, 2017; Williams & Govan, 2013), adversely affecting their mental health (Moore et al., 2017). Generally, bullying is defined as negative actions carried out by one or more individuals against a weaker peer (Gini & Pozzoli, 2009; Olweus, 1994; Pörhölä et al., 2020). Common key bullying criteria include a power imbalance, repeated aggression, and the victim’s inability to defend themselves.

Bullying can take place both in-person and online: Traditional in-person bullying in HE can be physical, verbal, or relational in nature (Cowie & Myers, 2015). It can be described as intentional social exclusion of an individual from a group and other ways of causing damage to social relationships (Lund, 2017; Pörhölä et al., 2020; Su et al., 2022). Relational bullying in HE to date has been investigated in a few studies and thus appears as a form of bullying that has been neglected in empirical research. Increasing attention has been paid to cyberbullying due to increased online teaching during and after the COVID pandemic, but with universities having returned to teaching on campuses, in-person bullying still is a major threat to student well-being. Cyberbullying takes place in digital spaces, for instance on social media by spreading rumours, sending hurtful messages, or sharing pictures/videos of victims publicly without their consent (Byers & Cerulli, 2021; Fisher et al., 2016). Meaningful relationships with peers provide information-related social capital and social support, positively affecting students’ academic achievement (Mishra, 2020). To create supportive university environments that strengthen students’ social capital and well-being and foster their academic success, research on relational bullying in HE and its underlying social dynamics is critical. With the present study, we aim to contribute knowledge on the field of relational bullying in in-person contexts, as this can offer valuable insights concerning developing positive social-ecological university environments.

Bullying in HE concerns ethical and moral dilemmas bystanders face (Brody & Vangelisti, 2016; Doane et al., 2020; Hayashi & Tahmasbi, 2021). Bullying situations occur not only between bullies and their victims but also often surrounding students (Gini et al., 2021). Bystanders, participants in bullying situations, can side with the bullies, become the victim’s prosocial defenders, or remain outsiders (Bauman et al., 2020; Smokowski & Evans, 2019). Bystanders determine whether the situation qualifies as bullying and gauge its severity to consider appropriate responses (Buglass et al., 2021; Dawes et al., 2022; Harrison et al., 2022; Huang et al., 2023). Forms of bullying which are relational, social, or indirect can be difficult to identify for outsiders, as they overlap in some aspects and can be grounded in a shared history between the involved subjects which bystanders do not have all information about (Allen, 2015). Drama, the intentional involvement in and extension of conflicts spreading across individuals, may be difficult to distinguish from actual conflicts, and both types of interaction can also show characteristics of bullying, as for instance a repeated aggression over time. For outsiders of a social group who are not familiar with the relationships between the group members, verbal bullying is not necessarily clearly distinguishable from banter, that is, provocatively teasing others with the aim of bonding and strengthening the relationship (Dynel, 2008), enhancing the difficulty of assessing a situation accurately (Buglass et al., 2021). Research with adolescents found that bystanders typically recognise the harmfulness of bullying and their moral duty to intervene; however, concerns about their social status and popularity may deter them from acting prosocially towards bullying situations (Pouwels et al., 2019). Their desire for belongingness and a secure status within the social group precedes shame and guilt from ignoring the bullying (Chen et al., 2016; Salmivalli, 2010; Strindberg et al., 2020; Thornberg et al., 2018). In research with HE students, the likelihood of supporting a victim was found to increase with the perceived severity of the incident and decrease with higher levels of social distance (Hayashi & Tahmasbi, 2020).

Fostering prosocial behaviours is an important aspect of anti-bullying programmes at universities, which is why research is needed to better understand what factors relate to prosocial behaviours to inform such programmes and increase their effectiveness (Barlett, 2017). Three prosocial bystander actions have been distinguished in research with adolescents (Gini et al., 2021; Jenkins et al., 2023; Lambe & Craig, 2020): Bystanders may comfort the victim by sitting with them or offering support, confront the bullies and ask them to stop, or report the incident to staff at the educational institution. However, to our knowledge, research in this growing field addressing factors that relate to prosocial behaviours in bystanders in HE has not yet differentiated between these different types (Doane et al., 2020; Hayashi & Tahmasbi, 2020, 2021; Huang et al., 2023). The current study aims to address this research gap by investigating the predictive value of predictors of bystander intentions towards in-person relational bullying according to the theory of planned behaviour (TPB) (Ajzen, 1991), as well as the incremental value of empathy and anticipated regret as emotional aspects that relate to bystander intentions (Hayashi & Tahmasbi, 2021).

Prosocial Bystander Intentions

Prosocial behaviour is a voluntary action meant to benefit somebody else (Eisenberg & Spinrad, 2014). In a qualitative study on experiences of traditional bullying at HE institutions in the UK (Harrison et al., 2022), students reported sexual harassment, active exclusion and isolation, cyberbullying, controlling or mind games, and verbal aggressions as the most common bullying tactics, often motivated by personal or social gains for the bully. Students reported a feeling of maturity, self-confidence, being comfortable with the peer group, and being acquainted with the victim or bully/bullies as reasons to help a victim. Their intervention strategies comprised verbally trying to stop the bullying, but also offering company to socially excluded victims. Although another qualitative study with U.S. HE students (Byers & Cerulli, 2021) was limited to cyberbullying experiences and participants who had at least once tried to help a victim of cyberbullying, its findings can still be valuable in the overall context of bullying, but may not be directly transferable to the UK context and traditional bullying forms. Students in the USA reported that their prosocial intervention in cyberbullying situations was based on shared identities, friendship, the ability to empathise with the victim, and, in some cases, a sense of social justice and the desire to defend it. Both studies point out the importance of the peer group in bystanders’ decision-making (Byers & Cerulli, 2021; Harrison et al., 2022). Theoretically, prosocial bystander behaviour can be based on a perceived personal responsibility to take action on behalf of another person in need (Latané & Darley, 1970) but can also be the result of a cost-reward analysis towards the bystander’s personal outcome (Dovidio & Banfield, 2015). For example, a bystander contemplating whether to intervene in a bullying situation can weigh the potential cost of jeopardising their social standing in relation to the bullies against the potentially rewarding gratitude from the victim.

Theory of Planned Behaviour

The TPB postulates that behavioural intention is the strongest predictor for actual behaviour and, therefore, may be regarded as a proxy where it is impossible to measure actual behaviour (Ajzen, 1991). It assumes the intention is determined by (i) attitudes towards performing the action, (ii) subjective normative influences, and (iii) perceived control over the behaviour.

In the TPB, attitudes comprise both cognitive and emotional evaluations on a favourable to unfavourable continuum. Actions to support another individual in need are shaped by the assessment of costs and rewards and arousals or emotions (Dovidio & Banfield, 2015). Individuals with strong attitudes towards the perceived appropriate behaviour towards bullying within their group may attempt to lead the group in moments of nonconformity (Hogg & Smith, 2007). Hence, cognitive and emotional attitudes are expected to affect bystanders’ intentions towards a group situation in which other members do not comply with their own attitudes and perceived group norms. Prosocial attitudes against bullying in university students are evident (Garland et al., 2017), and previous research in HE found prosocial attitudes to be a significant predictor of intention to help bullying victims (Doane et al., 2020; Hayashi & Tahmasbi, 2021; Sundstrom et al., 2018).

Subjective norms in the TPB, commonly described as social pressure, comprise injunctive and descriptive norms (Ajzen, 1991). Injunctive norms refer to the approval of the behaviour by significant others, while descriptive norms refer to whether significant others would execute the behaviour themselves if they were in the same situation. Norms are socially constructed in societies, binding, and imply both enforcement and generality (Jensen et al., 2014). Norms are enforcing in that they prescribe what types of behaviour should be applied towards whom and when, systematising the social environment, and they are general in that they apply to all individuals within a group. Individuals align themselves with their in-group based on norms, and they apply normative reasons to enforce group alignment by others in moments of transgression. Actions and articulations by individuals in a group communicate subjective norms and thereby affect actions between group members (Dovidio & Banfield, 2015; Hogg & Reid, 2006; Ryan & Deci, 2000; Turner et al., 1987). Furthermore, solidarity with other group members may create a psychological bond with the group and, thus, determine individuals’ group-based activities (Leach et al., 2008) and may consequently determine whether or how a bystander would react to a bullying situation. Subjective norms significantly predicted the intention to help victims of bullying in previous research with HE students (Doane et al., 2020; Hayashi & Tahmasbi, 2021; Sundstrom et al., 2018).

Perceived behavioural control, as the third predictor of intention in the TPB, describes the feasibility of the action considering all resources and potential hindrances (Ajzen, 2002). Prosocial involvement in a bullying situation may appear feasible to bystanders when their perceived behavioural control is high. This is influenced by resources such as the availability of time to begin a conversation with the involved peers, knowledge of whom to report the incident to, capability to comfort the target student, and a lack of expectations of negative impacts of their involvement. Perceived behavioural control was found to influence bystanders’ likeliness to intervene in HE bullying (Doane et al., 2020; Sundstrom et al., 2018).

The TPB’s utility in predicting behavioural intentions is evident across contexts (Armitage & Conner, 2001; Cooke & Sheeran, 2004; Godin et al., 2005). However, proportions of explained variance in bystander intentions vary in previous research, and specific additional predictors may offer an increased incremental value (Doane et al., 2020; Hayashi & Tahmasbi, 2021; Rivis et al., 2009; Vlaanderen et al., 2020). According to Ajzen (1991), the TPB may be extended with additional predictors, given that these increase the proportion of explained variance in the intention of interest (Rivis et al., 2009; Sandberg & Conner, 2008; Sandberg et al., 2016). Thus, researchers argue that bystanders’ emotional responses could explain additional variance in bystander intentions in HE students beyond the previously investigated TPB factors (Hayashi & Tahmasbi, 2021).

Emotional Factors Predicting Prosocial Bystander Intentions: Empathic Concern and Anticipated Regret

Bystanders of bullying are emotionally affected by the situations they witness, and emotional appraisals can help explain their intentions towards bullying situations. Interviewed HE students mentioned friendship, feeling mature and comfortable in the peer group, and the ability to empathise with the victim as reasons to prosocially react towards bullying (Byers & Cerulli, 2021; Harrison et al., 2022). As bystanders tend to be aware of others’ emotional responses to interpersonal rejection, such as loneliness, embarrassment, or sadness (Leary, 2015), they will likely care for the target student’s welfare to improve their situation (Eisenberg, 2005; Yagmurlu & Sen, 2015). Cuff et al. (2016) reviewed a range of proposed definitions of empathy and concluded that all included the perception and understanding of others’ emotions under the recognition that they are experienced by the other. Davis (1980) defined empathic concern as ‘feelings of warmth, compassion and concern for others’ (p. 2). Identifying this concern and caring for others as the distinguishing aspect between the two concepts, Jensen et al. (2014) claimed that this concern and care are motives leading to prosocial behaviour and that empathy can be regarded as a necessary but insufficient determinant of prosocial behaviour. Nevertheless, previous research found associations between prosocial bystander intentions and empathy as well as empathic concern for different age groups, countries, bullying contexts, and prosocial bystander behaviours. For instance, adolescents’ empathic concern was positively associated with defending behaviour (Pozzoli et al., 2017) and prosocial activity towards excluded characters in a digital game environment (Vrijhof et al., 2016). Adolescents’ cognitive empathy was associated with emotional intervention and reporting to adults for both known and unknown victims (Changlani et al., 2023). Furthermore, adolescents’ empathy was associated with comforting the victim, seeking adult support, and countering verbal bullying (Wachs et al., 2023). In addition, HE students’ prosocial intention towards bullying was positively associated with empathy in high and low bullying severity conditions (Huang et al., 2023) and in cyberbullying (Hayashi & Tahmasbi, 2021).

Furthermore, Hayashi and Tahmasbi (2021) found a positive relationship between the intention to help cyberbullying victims and the anticipated regret in bystanders over not helping the victims. Anticipated regret describes an aversive cognitive emotion from the imagined realisation that a situation would have been better had another course of action been chosen (Brewer et al., 2016). Individuals who do not intend to involve themselves in bullying can develop negative feelings towards their behavioural choice, feeling sorry, shame, or guilt. Anticipated regret relates to normative aspects of peers’ expectations and one’s own future emotional evaluation in the case of an action that is nonconform with group norms (Hogg & Reid, 2006). Consequently, a prosocial behavioural intention meant to improve the victim’s emotional situation may develop, and individuals may conclude that supporting the target student will be a good behavioural choice—be it for altruistic reasons or for egoistic purposes to reduce their own negative emotions (Dovidio & Banfield, 2015; Padilla-Walker & Carlo, 2014; Yagmurlu & Sen, 2015).

The Present Study

This study aimed to contribute knowledge to the existing research on factors related to bystander intentions towards in-person relational bullying based on the TPB. To explore additional insights into bystander intentions, empathic concern and anticipated regret as emotional components were examined. We conceptually followed Hayashi and Tahmasbi (2021), who examined relationships between the traditional TPB predictors, empathy, and anticipated regret and the intention to help cyberbullying victims as one outcome variable in a college-aged sample. As Wachs et al. (2023) found positive associations between empathy and comforting victims, confronting bullies, and reporting bullying in a study with adolescents, we sought to determine whether these findings could be replicated in a university student sample. We aimed to answer the following research questions:

RQ1: To what extent do the traditional TPB predictors attitude, subjective norms, and perceived behavioural control predict the intentions to comfort the victim, confront the bullies, or report the bullying?

RQ2: To what extent do empathy and anticipated regret predict the intentions to comfort the victim, confront the bullies, or report the bullying beyond the traditional TPB antecedents?

Prior research with HE students (Doane et al., 2020; Hayashi & Tahmasbi, 2021) has shown that the traditional TPB predictors are relevant for prosocial bystander behaviour intentions in general but did not consider different behaviour types (comforting the victim, confronting the bullies, reporting the bullying), which is why we consider the three outcomes exploratively. We expect that cognitive attitude, emotional attitude, injunctive norm, descriptive norm, and perceived behavioural control will be positively associated with the intentions to comfort the victim, confront the bullies, and report the bullying. According to prior empirical research with HE students (Hayashi & Tahmasbi, 2021) and theoretical work (Jensen et al., 2014), we expect that empathic concern and anticipated regret will be most closely associated with the intention to comfort the victim, compared to the intentions to confront the bullies and report the bullying. Our hypotheses are illustrated in Fig. 1.

Fig. 1
figure 1

Hypothetical model of all independent variables predicting comforting, confronting, and reporting as outcome variables

Full size image

Methods

Participants and Procedure

Prolific, a widely used platform for scientific research participant recruitment (Newman et al., 2021; Palan & Schitter, 2018), was utilised to collect cross-sectional data from undergraduate students at universities in the UK using a digital questionnaire (Brehmer, 2023). Participants from 18 to 35 years old were invited and received a small monetary compensation. The questionnaire was created using SurveyXact, and data collection occurred in October 2022. A gender-balanced pilot sample initially comprised 50 participants; however, one participant encountered a technical issue and was excluded upon detection of their missing data. The final sample included the pilot sample and consisted of N = 419 undergraduate students (MAge = 22.76 years, SDAge = 4.02 years; 47.0% male, 50.6% female, 2.4% other). Within the participant group, 86.6% were British, and 34 nationalities were reported. Simplified ethnicity data were provided by Prolific: Participants self-identified as 16.7% Asian, 4.8% Black, 3.8% mixed ethnicity, 73.3% White, and 1.4% other ethnic categories. Participants had self-reported high English proficiency and indicated being enrolled as undergraduate students at a university in the UK. The dataset contained no missing data due to enforced responses. Note that the respondents were assured that their data would be treated confidentially and were at all points of the data collection process given the opportunity to withdraw from their participation. The study received approval from the Norwegian Agency for Shared Services in Education and Research (SIKT; formerly NSD), ensuring adherence to methodological, ethical, and data processing standards.

Instrument

We applied a vignette-based design presenting respondents with a fictive group work situation, followed by measures for the different behavioural intentions towards the situation, TPB measures, empathic concern, and anticipated regret (Supplement 1) (Brehmer, 2023). On an introductory page, participants received the following instruction: ‘The questions are about what you think or how you feel in situations that happen between students at the university. Please answer honestly and tick in the answer that you first think of. There are no right or wrong answers!’. They then proceeded to the next page with the displayed vignette and items to measure the dependent variables.

Vignette

Participants were instructed, ‘Please read the text below and look at the picture. Afterwards, answer the questions’. The vignette (Supplement 2) began with the prompt, ‘Imagine the following situation’, and participants then read a short text (167 words) about being assigned to graded group work, with the group consisting of themselves and three other unknown students, of whom two are close friends (Brehmer, 2023). While the respondent is treated neutrally by the two friends and invited for lunch afterwards, the two friends treat the other student in several negative ways by degrading or questioning their constructive comments and suggestions, ignoring them, delaying the provision of requested materials, and not inviting them to join for lunch as the only person in this scenario. Following previous research (Brody & Vangelisti, 2016; Luo & Bussey, 2019), we attempted to prevent group effects between genders to a feasible extent by presenting respondents with a vignette representing students of their self-reported gender, with illustrations either indicating a situation between male students (Fig. 2) or female students (Fig. 3). Participants who identified as ‘other’ were presented with one of the two versions randomly. Underneath the vignette, participants were presented with nine items to assess the dependent variables to indicate their intentions towards the bullying situation. The text and illustration were shown consistently, with every scale referring to the same scenario.

Fig. 2
figure 2

Illustration indicating male group members ©Storyboard That

Full size image
Fig. 3
figure 3

Illustration indicating female group members ©Storyboard That

Full size image

Measures

Dependent Variables–Prosocial Intentions Towards Bullying

Three possible ways of reacting to a social exclusion situation were adapted from Lambe and Craig’s (2020) Defending Behaviours Scale. Following the prompt ‘How likely are you to do the following?’, respondents indicated their likelihood on a five-point Likert scale ranging from 1 (very unlikely) to 5 (very likely). Three items each assessed the intention to comfort the victim (e.g. ‘I try to cheer up C afterwards’; Cronbach’s α = .87), confront the bullies (e.g. ‘I start a conversation with A, B and C to find a solution’; α = .91), and report the incident (e.g. ‘I tell our lecturer about what happened’; α = .90). This design allowed participants to score high on more than one outcome at a time, so they did not have to choose between outcomes. All nine items were presented in the same random order on the same page to all participants.

Independent Standard TPB Variables

The respondents’ attitudes (adapted from Hayashi & Tahmasbi, 2021) towards their involvement were assessed on a six-point Likert scale through an involvement being perceived as 1 (completely worthless) to 6 (completely valuable) for cognitive attitude (α = .84) or 1 (completely frightening) to 6 (completely safe) for emotional attitude (α = .72).

Subjective norms (adapted from Hayashi & Tahmasbi, 2021) were assessed on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree) with two items for injunctive norms (e.g. ‘People who are important to me expect me to do something about a situation as shown above’; Spearman’s ρ = .74, p < .01) and two items for descriptive norms (e.g. ‘People who are important to me would do something if they had been in a situation as shown above’; Spearman’s ρ = .60, p < .01).

Perceived behavioural control (adapted from Hayashi & Tahmasbi, 2021) was measured on a six-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree) with three items following the prompt, ‘How much do you agree with the following?’ (e.g. ‘It would be easy to involve myself in a situation as shown above’) (α = .68).

Empathic Concern and Anticipated Regret as Extensions to the TPB

Empathic concern was assessed with items of the Personal Reactivity Index (Davis, 1980) on a five-point Likert scale ranging from 1 (does not describe me well) to 5 (describes me very well); e.g. ‘I often have tender, concerned feelings for people less fortunate than me’; α = .82). Anticipated regret was assessed with two items on a seven-point Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree) adapted from Hayashi and Tahmasbi (2021); e.g. ‘If I did not involve myself in a situation as shown above, I would feel sorry’; Spearman’s ρ = .79, p < .001).

Gender, Age, and Socially Desirable Responding as Covariates

Gender was assessed using the item ‘Which gender describes you best?’, and participants allocated 1 = male, 2 = female, and 3 = other. In addition, the respondents’ ages were retrieved from Prolific. To control for moralistic bias as part of socially desirable responding, a short scale from Kemper et al. (2012) was employed. It comprises two subscales for the exaggeration of positive qualities (e.g. ‘Even if I am feeling stressed, I am always friendly and polite to others’; α = .64) and understatement of negative qualities (e.g. ‘Sometimes I only help people if I expect to get something in return’; α = .58). As researchers suggest Cronbach’s alpha of 0.70 (Clark et al., 2021) or 0.60 (Taber, 2018) may be regarded as an acceptable indicator of internal consistency, we acknowledge that the internal consistency measured here was low and may be regarded as unsatisfactory. Researchers however previously accepted low internal consistency of these scales, for instance due to the low number of items (Sindermann et al., 2021). Another reason for low internal consistency could be the heterogeneity of scales measuring several distinct behaviours within one construct.

To determine how related this study was to students’ lives, we asked, ‘How often have you experienced that someone was singled out by others in your everyday study?’, and participants indicated whether they had witnessed social exclusion ‘never’, ‘seldomly’, ‘sometimes’, ‘1–2 times a week’, or ‘daily’. In our sample, 86.4% of the students had witnessed social exclusion among peers in the past, about half of them sometimes or more often. This demonstrates that most of the participants could imagine the situation realistically based on their experiences. Those who had witnessed social exclusion were presented with the follow-up question, ‘When you experienced that a student was singled out by one or more others, what did you do?’ and responded on a five-point Likert scale ranging from 1 (never) to 5 (always), with 0 = ‘I do not know’, to the voluntary items (N = 359/360/361) ‘I took care of the student who was singled out afterwards’, ‘I talked to the students to stop the situation’, and ‘I reported what happened to an employee at the university’.

Analysis

We examined two models in Mplus 8.0 (https://www.statmodel.com/HTML_UG/introV8.htm), applying maximum likelihood estimation. The first model included comforting, confronting, and reporting as outcome variables and cognitive and emotional attitudes, descriptive and injunctive norms, and perceived behavioural control as the traditional TPB predictor variables. In the second model, anticipated regret and empathic concern were added to the first model as additional independent variables predicting the three outcome variables. In both models, we controlled for age, gender, and the two subscales of socially desirable responding (exaggeration of positive qualities and understatement of negative qualities). To determine our models’ goodness of fit, we followed commonly applied thresholds (Kline, 2016; Schermelleh-Engel & Moosbrugger, 2003): A good fit is assumed when χ2 /df is between 0 and 2.0, CFI and TLI are between 0.97 and 1.0, and RMSEA and SRMR are between 0.00 and 0.05; an acceptable fit is assumed when χ2 /df is between 2.0 and 3.0, CFI and TLI are between 0.95 and 0.97, and RMSEA and SRMR are between 0.05 and 0.10.

Results

Pearson’s Correlation Coefficients

Pearson’s correlation coefficients for all variables are presented in Supplement 3. We consider coefficients .10 small or weak, .30 moderate, and .50 large or strong following the recommendations by Cohen (1988) (e.g. Parker et al., 2022); however, we acknowledge that researchers have been expressing needs for more differentiated effect size benchmarks (e. g. Bakker et al., 2019; Bosco et al., 2015). It is worth mentioning that respondents’ past behaviours and respective intentions correlated moderately to strongly for comforting (r = 0.48, p < 0.01) and strongly for confronting (r = 0.58, p < 0.01) and reporting (r = 0.64, p < 0.01), suggesting, to a certain degree, accordance between respondents’ indicated behavioural intentions in this study and their self-reported past behaviour in similar bullying situations.

Results of the Traditional TPB Model

The results from our structural equation modelling are presented in Supplement 4 for both the traditional and extended TPB models. The findings of the traditional TPB model are illustrated in Figs. 4, 5, and 6. Both models indicated an acceptable to good fit (Supplement 4). Perceived behavioural control significantly predicted comforting (β = 0.310, p = 0.000). However, none of the other predictors was significantly associated with comforting. Cognitive attitude (β = 0.192, p = 0.005), descriptive norms (β = 0.168, p = 0.027), and perceived behavioural control (β = 0.297, p = 0.000) significantly predicted confronting. Cognitive attitude (β = 0.139, p = 0.047) and descriptive norms (β = 0.156, p = 0.043) significantly predicted reporting. The predictors entered in the traditional TPB model explained 31.5% of the variance in the intention to comfort the victim, 36.50% of the variance in the intention to confront the bullies, and 10.7% of the variance in the intention to report the bullying. The outcomes of our explorative approach suggest that the traditional TPB predictors were more relevant to the intention to confront the bullies, as measured by the number of significant coefficients in the model, compared to the other outcomes.

Fig. 4
figure 4

Results traditional TPB model - Intention to comfort

Full size image
Fig. 5
figure 5

Results traditional TPB model - Intention to confront

Full size image
Fig. 6
figure 6

Results traditional TPB model - Intention to report

Full size image

Results of the Extended TPB Model

The findings of the extended TPB model with empathic concern and anticipated regret are illustrated in Figs. 7, 8, and 9. In line with our expectations regarding the additional emotional factors, perceived behavioural control (β = 0.171, p = 0.012) and the two additional factors empathic concern (β = 0.162, p = 0.032) and anticipated regret (β = 0.356, p = 0.000) significantly predicted comforting. Cognitive attitude (β = 0.181, p = 0.011), descriptive norms (β = 0.161, p = 0.039), and perceived behavioural control (β = 0.270, p = 0.000) significantly predicted confronting. Perceived behavioural control (β = −0.144, p = 0.029) significantly predicted reporting negatively in this model. The predictors entered in the extended TPB model explained 41.2% (ΔR2 = 0.097) of the variance in the intention to comfort the target student, 36.7% (ΔR2 = 0.002) of the variance in the intention to confront the bullies, and 12.3% (ΔR2 = 0.019) of the variance in the intention to report the bullying.

Fig. 7
figure 7

Results extended TPB model - Intention to comfort

Full size image
Fig. 8
figure 8

Results extended TPB model - Intention to confront

Full size image
Fig. 9
figure 9

Results extended TPB model - Intention to report

Full size image

Discussion

The present study investigated predictors of bystander intentions according to an extended TPB framework (Ajzen, 1991; Hayashi & Tahmasbi, 2021), adding insights from the university context to the body of literature on bystander intentions towards relational in-person bullying. We focused on the differential associations of the factors described by the TPB and emotional aspects (empathic concern, anticipated regret) with comforting the victim, confronting the bullies, and reporting the bullying as distinct types of bystander intentions towards relational in-person bullying. The study had two main findings. First, we found that the factors described in the TPB predicted cognitive aspects of prosocial behaviour. Confronting the bullies was found to be a likely intention in respondents who held a positive cognitive attitude, who perceived social pressure but also a high capability to involve themselves in the situation. The intention to comfort the victim was positively associated with perceived behavioural control, but neither with respondents’ attitudes towards involvement, nor their subjective norms. Lastly, the intention to report the bullying was positively associated with a positive cognitive attitude towards involvement, and normative pressure, but not the perception of behavioural control. Thus, the TPB offered a framework which was mostly applicable to predicting the intention to confront bullies, with aspects of attitudes and subjective norms and perceived behavioural control being associated with this specific cognitive intention. This result matches with research indicating that confrontation involves cognitive and strategic engagement (Dovidio & Banfield, 2015), while descriptive norms guide bystanders behaviours towards other group members (Hogg & Reid, 2006; Jensen et al., 2014; Ryan & Deci, 2000; Turner et al., 1987). When involvement in a bullying situation aligns with group norms, and no major obstacles impede bystanders, confrontation may appear as a reasonable behavioural decision. Moreover, higher perceived behavioural control also predicted confronting, indicating that individuals intend to confront bullies when they perceive no hindrances and have the necessary resources available (Ajzen, 2002).

When empathic concern and anticipated regret were added to the traditional TPB predictors, they were only directly associated with the intention to comfort the victim of bullying, as will be discussed below. The intention to confront the bullies was not affected by empathic concern or anticipated regret: it was still positively associated with the same TPB predictors. The intention to report the bullying was not associated with empathic concern and anticipated regret either; however, their addition to the model affected other predictors of reporting. For instance, cognitive attitude and descriptive norms were no longer associated with the intention to report, which became negatively associated with perceived behavioural control. This suggests that respondents perceiving high behavioural control would become less likely to report the bullying to the university. Considering our respondents’ ages, it appears reasonable that students perceiving high control over their behavioural options would intend to manage the situation themselves rather than involve university staff. This preference for direct support for the victim over reporting bullying, presuming that students feel able to take care of the matter, was also expressed in interviews with students (Byers & Cerulli, 2021).

Second, we provided further evidence on the role of emotional aspects for bystander intentions, specifically by adding empathic concern and anticipated regret into the model. Our results highlight that emotional aspects were mainly related to comforting: We found that the intention to comfort the victim was the only one of three intentions that had positive associations with empathic concern and anticipated regret. Respondents with strong emotional appraisals and high awareness of others’ emotional states may prefer comforting victims to avoid confrontation and further escalation. However, Wachs et al. (2023) found positive associations of empathy with comforting the victim, seeking help at school, and countering verbal bullying in pupils (12–19 years old). In the present study, we did not replicate these findings in a sample of university students. Although the severity of the bullying incident cannot be expected to affect the bystander response (Huang et al., 2023), the different bullying types (verbal bullying and social exclusion) may affect the comparability of results, even though evidence suggests relationships between empathy and prosocial bystander responses across bullying contexts (Changlani et al., 2023; Hayashi & Tahmasbi, 2021; Pozzoli et al., 2017; Vrijhof et al., 2016). Another potential reason for this difference could lie in the interrelationship between empathic concern and subjective norms and how these are developed differently in children, adolescents, and young adults (e.g. Jensen et al., 2014). Perhaps, based on their emotional appraisals and subjective norms and the consequences to be expected in their complex social-ecological environment, adult bystanders become more selective in their choice of prosocial behaviours. The low correlations between empathic concern and injunctive and descriptive norms in our sample (Supplement 3) indicate weak relationships between these predictive factors among university students in this study. Empathy and norms in adults may still interact in similar ways as in children and adolescents; however, empathic students may more likely intend to care for the victim’s welfare rather than intend to enforce alignment by asking the bullies to include the victim in their group work, thus potentially escalating the situation, or reporting the incident and thereby causing increasing negative emotions in the victim. On the contrary, strong descriptive norms were associated with confronting the bullies. Thus, students reporting stronger descriptive norms likely attempt to enforce the bullies’ alignment.

Limitations

This study used self-reported data reflecting participants’ subjective interpretations. To mitigate the potential impact of a moralistic bias in participants’ responses, we controlled for socially desirable responding, making this study one of few to incorporate socially desirable responding (e.g. Xie et al., 2023) into bystander intention research. Notably, this approach revealed that respondents may be less inclined to comfort victims in real-life situations. However, we acknowledge that this finding ought to be interpreted carefully, as the internal consistency of the two employed scales was low, presumably due to scale heterogeneity with items describing actions which strongly differ from each other (e. g. taking advantage of someone, only helping others when expecting reciprocity, throwing away litter in the countryside). Other recent research employing the same scales in different countries found similarly low values for internal consistency (Münscher et al., 2020; Paula Sieverding et al., 2023; Sindermann et al., 2021), but accepted these for instance due to the low number of items per scale (Sindermann et al., 2021). In future research, instruments for assessing socially desirable responding ought to be explored further. Nevertheless, assessing respondents’ past behaviours towards bullying situations was another way to check how respondents may have responded realistically and indicated coherence between the reported past behaviours and behavioural intentions towards in-person bullying. For future research, we recommend assessing respondents’ actual self-reported behaviour in a follow-up study.

This study was conducted via Prolific.co.uk; thus, respondents were likely accustomed to participating in surveys and potentially especially interested in bullying and social exclusion topics. Such an interest might result from their own negative experiences with bullying in various participant roles, potentially leading to stronger expressions in the survey. Different sampling methods in universities could mitigate this bias.

The study’s outcomes would perhaps have varied if the vignette had described the situation differently or provided additional information (Aguinis & Bradley, 2014). For example, qualitative research in the USA found that some White-identified students would be hesitant to involve themselves in racial bullying to avoid speaking on behalf of another ethnicity or to avoid misplacing structural issues onto individuals (Byers & Cerulli, 2021). Correspondingly, students of colour explained that they would feel personally attacked due to shared gender and ethnicity with the victim, influencing their prosocial reaction based on shared identity. Whether these findings are replicable in the UK context or other countries would be an interesting line of inquiry for future research.

Moreover, we used an online sample of English-speaking university students from the UK. Thus, we cannot speak to the generalisability of our results. Future studies with larger samples could examine group effects based on factors such as gender or ethnicity (Eisenberg & Spinrad, 2014; Mulvey et al., 2020). Potential effects based on ethnicity can be expected, as described above, with students being more inclined to support a victim based on shared ethnicity (Byers & Cerulli, 2021). Further research on group effects may be conducted by applying vignettes, which systematically vary in gender, ethnicity, or other factors of interest (Aguinis & Bradley, 2014). Depicting a situation in which students meet in a private context on campus, such as the cafeteria, gym, or other common areas, or outside of their campus in private settings, such as the park, café, or a party, would be of interest to investigate what role the academic versus private context play in bystanders’ formation of behavioural intentions towards bullying.

Cyberbullying, which was briefly mentioned in the introduction, is another field of research which could be explored further in the context of HE. Some studies have already investigated different issues with regards to cyberbullying in HE (Byers & Cerulli, 2021; Doane et al., 2020; Gahagan et al., 2016; Hayashi & Tahmasbi, 2020, 2021; Huang et al., 2023; Jeong et al., 2022). While there seems to be a trend towards research on cyberbullying at all educational levels, we decided to focus on traditional in-person relational bullying in this study as the research foundation on traditional bullying in HE and the role of bystanders in it is still considered scarce and ought not to be overlooked. However, it would be of great interest to us to find out whether the findings our study provides may be replicated in the cyberbullying context, as online interactions make up a considerable part of HE students’ academic and private interactions with each other.

Regarding methodology, emotional attitudes and injunctive norms did not significantly predict the outcomes. The item prompt ‘Involving myself in a situation as shown above would be…’ might have led respondents to think of involvement as an action similar to confronting the bullies rather than other types of involvement. In future research, the item wording should imply, more specifically, various forms of involvement. However, more differentiated constructs would cause a higher number of variables in explorative research like this. Despite these limitations, the present study offered the first insights into the relationships between the traditional TPB constructs and various prosocial bystander intentions. Future research should target comforting, confronting, and reporting individually to enhance knowledge of their specific determinants.

Implications for Practice

Our study offers several practical implications. Comforting victims and confronting bullies are both desirable prosocial responses with meaningful implications for social group dynamics at the university. Assertive bystanders confronting bullies can stop the bullying (Bauman et al., 2020) and affect group norms over time (Abbott & Cameron, 2014), whereas empathic bystanders comforting victims may help mitigate the negative impacts of bullying by offering victims companionship and emotional support.

Campaigns aiming to encourage students to comfort their peers should focus on strengthening empathic concern, perspective-taking, emotional awareness of others’ situations, and everyone’s collective responsibility for each other’s welfare (Branscum et al., 2023; Brüggemann et al., 2019; Sundstrom et al., 2018). In their experimental study, Nordgren et al. (2011) showed that people tend to underestimate the social pain resulting from social exclusion in others when they have not experienced it themselves. In contrast, participants who had experienced exclusion themselves showed greater acknowledgement of others’ social suffering. Bridging this empathy gap through activities like role plays and discussions can enhance students’ awareness of others’ distress and encourage comforting in bystanders (Cowie & Myers, 2014). Campaigns aiming to encourage students to confront bullies should focus on students’ prosocial attitudes towards being a bystander in bullying. Additionally, enhancing communication about perceived subjective norms and mutual expectations is important (Meriläinen et al., 2015; Myers & Cowie, 2013). Students lacking clear guidance on how to react towards bullying situations in a socially accepted way felt that their hesitance was reduced by discussing ethical issues and courses of action with other students (Byers & Cerulli, 2021). Campaigns aiming to increase reporting bullying rates ought to provide clear definitions of bullying and describe how to recognise it (Harrison et al., 2020; Vaill et al., 2023). Importantly, the results showed the need to communicate the importance of reporting bullying, even if one has already reacted to it, to reduce the number of undetected bullying incidents. Students expressing a preference for directly supporting victims of bullying over reporting the bullying incident to the university (Byers & Cerulli, 2021) demonstrates a need for universities to request bullying reports.

Our investigation of HE students’ prosocial intentions towards in-person relational bullying showed that from the factors described in the TPB framework, subjective norms and perceived behavioural control were the best predictors of students’ intention to confront bullies. Furthermore, empathic concern and anticipated regret were positively associated with students’ intention to comfort bullying victims. Lastly, higher perceived behavioural control was found to make students less likely to report bullying to the university. Regarding practical implications, we cannot assume causality, but found that prioritising emotional-affective responses could be a valuable strategy to enhance students’ intention to comfort victims of bullying, while focusing on cognitive-affective responses may encourage students’ intention to confront bullies. Moreover, HE institutions need to address the importance of reporting bullying in their anti-bullying campaigns to encourage prosocial bystander behaviours and facilitate access to reporting tools.