Introduction

Like other moral judgments, responsibility attributions heavily depend on individual benchmarks and expectations. Before the launch of the SpaceX mission in May 2020, Elon Musk said during a TV interview, “I’m the chief engineer of this thing, so I’d just like to say that if it goes right, it’s credit to the SpaceX-NASA team. If it goes wrong, it’s my fault.” (Ghosh, 2020, para. 2). His statement acknowledges that individuals consider the social context and social status while ascribing responsibility to others. Moreover, Musk suggests that negative and positive consequences of the same action lead to different responsibility attributions, with negative consequences implying stricter evaluative standards for agents with higher status.

Recent research extends Musk’s observation to side effects of profit-oriented company decisions (Kaspar et al., 2016; Willemsen et al., 2018). Social context variables like social roles implicate a set of specific norms and expectations, setting standards for moral behavior (Hamilton & Sanders, 1981). Similarly, social groups also shape social judgments (Brauer, 2001; McGarty & Penny, 1988) and expectations (Hertel et al., 2002; Song et al., 2018). Consequently, deviations from moral standards should be evaluated differently depending on the agent’s social characteristics. However, research on normative expectations and responsibility attribution focuses on cognitive explanations (e.g., Uttich & Lombrozo, 2010; Wright & Bengson, 2009), but it neglects associated affective mechanisms. The present study aims to extend research on moral and causal responsibility attributions by including contributions of social context and affective reactions into current models of blame and praise attribution.

This study examines the affective mechanisms underlying responsibility attribution patterns by means of the following scenario: A manager and an employee of a start-up company, either young or middle-aged, cooperatively make a decision which increases their company’s profit but leads to a secondary outcome either harming or helping the environment. Start-ups provide an ideal frame of reference for the present study, as they are increasingly being initiated not only by older people but also by very young entrepreneurs, which is reinforced, among other things, by corresponding start-up offerings at universities (Shirokova et al., 2018). In the present scenario, the agents’ social role, their age group, and the valence of the secondary outcome were systematically varied. Subsequently, the responsibility attributions were examined with regard to participants’ affective state. The main research question was: How do the valence of the side effect, the agents’ social role, the agents’ age group, and participants’ affective state influence the attribution of moral and causal responsibility after the company workers jointly decided to harm or help the environment? Given the increasingly damaging consequences of the climate crisis (Challinor et al., 2014), it is critical to understand under which circumstances people hold companies accountable for environmental impacts caused by companies’ decisions.

Influence of the Valence of a Decision’s Side Effect on Responsibility Attributions

Over the past two decades, Knobe has laid the groundwork for analyzing blame and praise attributions as a function of the valence of secondary outcomes (Knobe, 2003, 2010; Nichols & Ulatowski, 2007). For negative side effects, perceived intentionality and blame attributions are usually higher than intentionality and praise attributions for positive side effects of the same action. This finding has been replicated by numerous studies (e.g., Kaspar et al., 2016; Willemsen et al., 2018; Wright & Bengson, 2009) and with different methods (e.g., Cushman et al., 2008; Guglielmo & Malle, 2010; Mazzocco et al., 2004). Knobe’s original vignette shows the chairman of a company whose decision leads to a beneficial primary outcome for the company, but also to either good or bad environmental side effects that the chairman is indifferent to (Knobe, 2003). The resulting asymmetry in intentionality judgments could arise because norm-violating behavior provides more information about the underlying mental state than norm-congruent behavior (Uttich & Lombrozo, 2010). Another common explanation is the bi-directional relation between intentionality attributions and moral evaluations, influencing each other (Wright & Bengson, 2009; Knobe, 2010). It is not enough to be causally responsible for a good outcome (e.g., helping the environment), but the reasons for doing so must be morally right in order to be positively evaluated, otherwise praiseworthiness and intentionality judgments remain unaffected (Nichols & Ulatowski, 2007; Wright & Bengson, 2009). However, failure to avoid a negative outcome when one knows better is judged to be morally wrong, leading to an amplification of blame and intentionality attributions. Furthermore, helping the environment should be considered a standard behavior, whereas harming the environment violates dominant social norms and surpasses a certain “default point” of moral behavior (Kaspar et al., 2016; Knobe, 2010). Liao et al. (2018) extended this explanation by showing that the degree of caring for the outcome of an action mediates intentionality attributions, especially for negative outcomes. Another determining factor could be the differing probability of the outcomes, as negative outcomes were deemed more likely than positive outcomes (Nakamura, 2018). Given this large body of evidence, we expected to replicate the side effect asymmetry for moral responsibility and intentionality judgments:

  • H1a. When evaluating the manager, negative side effects are perceived as intentional, whereas positive side effects do not cause the same intentionality attribution.

  • H1b. When evaluating the manager, negative side effects are perceived as more blameworthy, whereas positive side effects do not appear as particularly praiseworthy.

Influence of the Social Context on Responsibility Attributions: Social Hierarchy and Social Groups

While research has shown that the valence of secondary outcomes influences the amount of responsibility attributed to someone, responsibility judgments are affected by further factors (Alicke et al., 2011). In particular, they also depend on the social context in which they occur (Hamilton, 1978). Indeed, the importance of social roles and associated normative expectations for moral responsibility attributions had already been discussed by Hamilton and Sanders (1981) and were empirically supported by recent experiments (Kaspar et al., 2016; Willemsen et al., 2018). Hamilton (1978) emphasized the need to include social roles in models of responsibility attribution, arguing that social roles set the normative context within which decisions and actions are evaluated. Since high power roles are expected to anticipate harmful events and supervise others, stricter rules are used to derive their responsibilities, compared to the responsibilities of subordinate roles.

Using a modified version of Knobe’ original vignette, Kaspar et al. (2016) as well as Willemsen et al. (2018) revealed a second asymmetry in responsibility attribution: Within companies, moral responsibility attributions in terms of blame or praise varied with the social role of a decision maker. Participants ascribed more blame to a company’s boss than to a subordinate employee when their joint decision negatively impacted the environment (Willemsen et al., 2018) or the customers (Kaspar et al., 2016), but more praise to the employee than to the boss when the decision caused positive side effects. The strength of this social role effect depended on participants’ cultural background but it remained substantial overall (Kaspar et al., 2016), and it persisted even when the employee made the final decision (Willemsen et al., 2018). These results underscore the robustness of the effect which was replicated for German participants (Kaspar et al., 2016), participants from the United Arab Emirates (Kaspar et al., 2016), and participants from the USA (Willemsen et al., 2018), given substantial differences in hierarchical thinking between cultures in terms of horizontal and vertical individualism and collectivism (cf. Singelis et al., 1995; Triandis & Gelfand, 1998). Thus, the hierarchical distribution of power within organizations manifests itself in the moral evaluation of certain social roles and leads to a less favorable evaluation of roles that involve more power and control.

The importance of social context variables for responsibility attributions can be derived from the culpable control model (Alicke, 2000; Alicke et al., 2011), supported by multiple experimental studies (e.g., Alicke et al., 2008; Lagnado & Channon, 2008; Menaker & Miller, 2013). Certain social roles may be associated with certain expectations and evaluations of personal control and agency. The culpable control model refers to these evaluations as the non-motivational factors that drive attributions of moral and causal responsibility (Alicke et al., 2011), because they are derived from analytic evaluations of control. Personal control here is reflected by the hierarchical order within companies, which means that in decision-making processes, the boss is normatively expected to exercise more control over the company’s production processes than the employee (Kaspar et al., 2016). Therefore, we expected different social roles to elicit different attributions of blame, praise, and causal responsibility, similar to what has been reported by Kaspar et al. (2016) and Willemsen et al. (2018):

  • H2a. In case of a negative side effect, participants ascribe more blame to the manager than to the employee.

  • H2b. In case of a positive side effect, participants ascribe more praise to the employee than to the manager.

  • H3a. In case of a negative side effect, participants ascribe more causal responsibility to the manager than to the employee.

  • H3b. In case of a positive side effect, participants ascribe more causal responsibility to the employee than to the manager.

Besides social hierarchy, the culpable control model suggests that further social context variables contribute to group-specific norms and expectations, such as the sense of belonging to a social group (e.g., Hertel et al., 2002; Song et al., 2018). Recent findings regarding generational identity demonstrated that categorization into age groups can trigger group identification processes in the same way than other social groups, leading to ingroup and outgroup effects (Van Rossem, 2019; Ross et al., 2019). By explicitly stating the age of the agents in the vignettes of the present study and thus making age a salient factor, social categorization theory (Turner et al., 1987) postulates that young participants will either identify with the young agents or distance themselves from the middle-aged agents. Indeed, age group is considered a critical social category in the context of environmental issues. Older generations are expected to do greater harm to the environment than younger generations (Chazan & Baldwin, 2019), whereas younger generations are expected to act more pro-environmental (Gray et al., 2019). Indeed, some studies indicate that young people show more environmental concerns than older people do (Casey & Scott, 2006). In light of these findings, the question arises as to whether the specific expectations that individuals have regarding the environmental behavior of different age groups affect the attribution of moral responsibility to a company’s agents. We hypothesized:

  • H4. Age group (ingroup versus outgroup) produces a main effect in moral responsibility judgements for negative and positive side effects.

Outgroup Homogeneity, Ingroup Differentiation, and Collective Responsibility

In this section, we will further elaborate on the role of social group in responsibility attribution. A large body of research has shown that people perceive ingroup members differently than outgroup members (e.g., Bente et al., 2016; Brauer, 2001; McGarty & Penny, 1988). The outgroup homogeneity effect (Judd & Kervyn, 2010; Mullen & Hu, 1989) is a well-supported phenomenon for various domains of intergroup perception, which has already been demonstrated for age groups (Linville et al., 1989). It refers to the observation that people are less sensitive to differences between outgroup members than to differences between ingroup members (e.g., Bente et al., 2016). Furthermore, people distinguish more between subgroups within their ingroup (Mullen & Hu, 1989; Park et al., 1992). Availability heuristics and memory biases offer partial explanations for outgroup homogeneity effects (Messick & Mackie, 1989). First, in most intergroup contexts, there are more opportunities to interact with ingroup members than with outgroup members, and to encounter more diverse exemplars in the ingroup, which enhances perceived ingroup variability (Mullen & Hu, 1989). Second, people might pay less attention to differentiating attributes among outgroup members than ingroup members, creating a more complex representation of the ingroup (Judd & Kervyn, 2010; Messick & Mackie, 1989). However, when ingroup norms are salient or when familiarity with outgroup members is high, ingroup homogeneity effects may also occur (Rubin & Badea, 2007). The outgroup homogeneity effect increases with ingroup identification (De Cremer, 2001; Lickel et al., 2006), intergroup threat (Rothgerber, 1997), and stereotypicality of outgroup attributes (Simon & Pettigrew, 1990). Overall, outgroup homogeneity persistently influences how people perceive and evaluate outgroup members.

Assimilating outgroup members further reinforces collective responsibility perceptions like collective blame. In that case, outgroup members are deemed causally and morally responsible independently of their actual causal contribution but simply because they are members of a particular group (Lickel et al., 2006). Indirect contributions like failing to prevent a harmful outcome then justify the ascription of responsibility to all outgroup members, even to those not directly responsible for the outcome (Lickel et al., 2003). As a result, people target outgroups as a whole instead of focusing on individual contributions.

Said outgroup homogeneity and collective responsibility processes can bias responsibility judgments of outgroup members. Given these processes, we assumed that homogeneity and collective responsibility effects can partially overwrite the effect of social role, attenuating the influence of social roles on moral and causal responsibility attribution. However, with respect to the ingroup, individual differences should be emphasized (Mullen & Hu, 1989; Park et al., 1992), increasing the status-associated responsibility differences between the manager and the employee.

  • H5a. In case of a negative side effect, participants in the ingroup condition show a larger difference between the attributed blame to the manager versus the employee, compared to participants in the outgroup condition.

  • H5b. In case of a positive side effect, participants in the ingroup condition show a larger difference between the attributed praise to the manager versus the employee, compared to participants in the outgroup condition.

  • H6a. In case of a negative side effect, participants in the ingroup condition show a larger difference between the attributed causal responsibility to the manager versus the employee, compared to participants in the outgroup condition.

  • H6b. In case of a positive side effect, participants in the ingroup condition show a larger difference between the attributed causal responsibility to the manager versus the employee, compared to participants in the outgroup condition.

Affective Processes as Driving Factors of Responsibility Attribution

In addition to the role of social hierarchy and social group in attributing responsibility, we now add a perspective on affective processes. Morality assessments are strongly associated with affective information processing (Tetlock et al., 2007). Although responsibility attribution models imply a driving role of affective processes, past research on the social role effect and the Knobe effect has not yet fully unscrambled the contributions of affective mechanisms. According to the culpable control model, causal and moral evaluations are susceptible to affective biases linked to motivational factors of control evaluations (Alicke et al., 2015). While non-motivational factors are assessed through analytical evaluations of control, motivational factors are relatively unconscious and spontaneous assessments influencing control judgments (Alicke, 2000). Spontaneous evaluations affect people’s evidential standards, perceptions of evidence and the amount of control over the outcome, and information search (Alicke et al., 2011). Therefore, spontaneous evaluations contribute to the activation of a blame-validation mode, directing blame to the agent who evokes the most negative affect (Alicke, 2000). This appeals to people’s need to hold others responsible, with stronger affect enhancing blame-validation processes (Alicke et al., 2015; Alicke et al., 2011). As such, negative affective states can mediate the relationship between harm and blame attributions (Quigley & Tedeschi, 1996). Nichols and Knobe (2007) showed that stronger affective responses, initiated by morally bad behavior, enhanced moral responsibility ascriptions even when participants did not believe the agent was responsible based on logical considerations. Furthermore, people high in alexithymia (i.e., having difficulties in emotional processing) displayed a reduced Knobe effect, confirming the driving role of affect for intentionality attributions (Zucchelli et al., 2019). Also, the culpable control model suggests that positive evaluative reactions initiate a desire to praise, enhancing responsibility attributions for positive consequences (Alicke et al., 2011; Moir, 2014).

The fair-but-biased-yet-correctible model (Tetlock, 2002; Tetlock et al., 2007) further clarifies the role of affective mechanisms underlying responsibility judgments. It is based on the intuitive prosecutor approach which assumes that people are inclined to ascribe responsibility and morality to others, and that negative affect evoked by agents serves as cue for these judgments. According to Tetlock (2002), people usually try to be fair while evaluating the responsibility for immoral actions. However, affective cues can bias these judgments by increasing or decreasing the attributed responsibility (Tetlock et al., 2007). Hence, the fair-but-biased-yet-correctible model reinforces the predictions of the culpable control model according to which responsibility is evaluated with regard to moral standards and influenced by affective evaluations. These empirical and theoretical evidences suggest that affective reactions could exert a powerful influence on responsibility judgments. Consequently, we expected to find significant positive correlations between negative affect and blame attribution, and between positive affect and praise attribution:

  • H7a. In case of a negative side effect, the amount of attributed blame correlates positively with negative affect.

  • H7b. In case of a positive side effect, the amount of attributed praise correlates positively with positive affect.

Effects of the Perceiver’s Attributes on Responsibility Attributions

Finally, while social context variables relate to normative expectations tied to the agents’ social characteristics, moral standards also vary with the individual expectations tied to the perceiver (Capraro & Sippel, 2017; Staley & Lapidus, 1997). Individual characteristics could influence the perceiver’s blame or praise ascriptions. For instance, there is evidence for gender differences in overall moral reasoning (Capraro & Sippel, 2017; Rothbart et al., 1986) and moral responsibility attribution in particular (Jamshed & Kamal, 2019; Staley & Lapidus, 1997). As the present study focuses on evaluations of environmental side effects, it appears also worthwhile to analyze environmental protection motivation, which might determine the standards for environmentally responsible behavior. For instance, Fielding and Head (2012) found that responsibility attributions were related to participants’ environmental intentions. Hence, we examined in an exploratory way how participants’ gender and environmental protection motivation are related to their moral responsibility judgments:

  • H8. There are differences between women’s and men’s moral responsibility attributions.

  • H9. Moral responsibility attributions correlate with participants’ environmental protection motivation.

Method

Participants

As the social role effect varies with cultural background (Kaspar et al., 2016), the study was restricted to participants living in Germany. Only participants aged between 18 and 30 years were included to recruit a homogenous sample being representative for the young generation and forming the ingroup (vs. outgroup). Participants were recruited via social networking sites and university mailing lists, using cross-sectional voluntary response sampling. Each participant was randomly assigned to one of four scenarios. Answers were recorded anonymously, and participation was voluntary with no incentives provided.

Kaspar et al. (2016) reported social role differences in moral responsibility attribution ranging from d = .39 to d = 1.17 for German participants. We aimed at detecting a small effect (d = 0.3) using two-tailed t-tests for independent samples with a power of .80 and an error probability of α = .05. Hence, the required minimum sample size was 172 participants per condition, 702 participants in total (calculated with G*Power version 3.1.9.2 – Faul et al., 2009). Overall, 1003 participants completed the study. Six participants were excluded due to lack of informed consent, 40 because they were older than 30 years, seven because they gave incorrect responses to all three attention check questions, and 25 because they indicated that they did not participated truthfully and thoughtfully. The final sample comprised n = 925 participants (Mage = 24.59, SD = 3.32). 687 participants were women (74.3%), 220 were men (23.8%), and 18 participants (1.9%) did not identify as male or female. 857 participants (92.7%) had a university entrance qualification or a higher level of education, 68 participants (7.3%) had lower levels of education.

Study Design and Procedure

The study was conducted as an online experiment using a 2 × 2 × 2 (social role × valence of side effect × age group) mixed-measures design. The independent variables were 1) the agent’s social role (within-participants factor; company manager versus subordinate employee), 2) the valence of side effect (between-participants factor; negative versus positive impact on environment), and 3) age group (between-participants factor; start-up with a young team versus start-up with a middle-aged team). Following previous studies (e.g., Knobe, 2003; Kaspar et al., 2016; Willemsen et al., 2018), social role was realized as a within-participant factor in the vignette in order to explicitly represent the differences in hierarchical status between a superior person (manager) and a subordinated person (employee) and to realize communicative aspects of a joint decision-making scenario. The dependent variables were the intentionality, moral responsibility, and causal responsibility attributed to the agents, participants’ positive and negative affective reactions, and their environmental protection motivation.

Initially, participants were informed about the study’s purpose, their rights, and data protection procedures. After providing informed consent, they were randomly assigned to one of four experimental scenarios: young team (ingroup) whose decision has a negative side effect on the environment; young team and positive side effect; middle-aged team (outgroup) and negative side effect; middle-aged team and positive side effect. After reading the corresponding vignette, participants responded to three attention check questions (age of the main agent reported in the vignette, the type of company, and the decision made by the agents). Then, participants indicated how much blame or praise the manager and employee deserved for their decision’s side effect, to what extent they were the cause for the side effect, and whether they intentionally triggered the side effect. Subsequently, participants rated their current affective state, and their environmental protection motivation. Finally, participants provided demographic data (age, gender, highest educational degree) and indicated whether they participated truthfully and thoughtfully.

Materials

Vignette

The vignette was based on Willemsen et al. (2018) and translated into German. Agents with different social roles, differing in their position within the company’s hierarchy, jointly made a decision leading to a beneficial primary outcome for their company, but to either a positive or negative secondary outcome (side effect) for the environment. The present study’s vignette differed from the original vignette in three minor aspects: First, the explicit statement of indifference to the consequences of the decision was replaced by a sentence indicating that the manager and the employee only cared about the company’s profit, making their indifference less explicit. Second, the age of the company workers was varied according to the study’s focus on social group effects, with either a young team (ingroup) or a middle-aged team (outgroup). Third, the nature of the company was specified to increase the authenticity of the setup by stating that the company is a start-up producing video games and game consoles. The vignette was preceded by an explanation stating that the company workers are mostly young (versus middle-aged) people who gain their first experiences in the video game industry. The vignette reads as follows (variations in brackets):

A 24-(54)-year-old employee of a start-up company consisting of young university graduates (middle-aged entrepreneurs), which produces video games and consoles, presents the following idea to the 26-(56)-year-old company manager: “The economic situation of our company is difficult. We need to check where we can reduce costs. One option might be to improve the manufacturing process. I have heard about this new module for the game console that lasts longer and is much cheaper. At the same time, the manufacturing process is associated with higher (lower) levels of CO2 emissions and therefore harms (helps) the environment. But for me, reducing costs has the highest priority. I also see potential for economization in other divisions of the company. It’s on you to decide whether we switch to the new module or not.”

The manager responds: “From a technical perspective, the module will work just as good as the more expensive one and it will certainly allow us to save a lot of money. I therefore suggest installing the new module. For me, too, reducing costs has the highest priority.” After the manager’s decision, the cheap, longer lasting module is installed, and the environment is harmed (helped) by higher (lower) CO2-emissions. The company successfully manages to save money.

Intentionality, Moral Responsibility, and Causal Responsibility

To assess the moral responsibility (deserved blame / praise) attributed to the manager and employee, participants responded to the question “How much blame (praise) do you think the manager (the employee) deserves for harming (helping) the environment?”. Moral responsibility was assessed on a seven-point rating scale ranging from 1 (no blame / no praise) to 7 (high blame / high praise). To assess the causal responsibility attributed to the manager and employee, participants responded to the question “To what extent do you agree with the statement that the manager (employee) was the cause of harming (helping) the environment?”. Finally, to assess the intentionality attributed to the manager and employee, participants responded to the question “To what extent do you agree with the statement that the manager (the employee) intentionally harmed (helped) the environment?” Causal responsibility and intentionality attributions were assessed on a seven-point rating scale ranging from 1 (do not agree at all) to 7 (totally agree).

Affect and Environmental Protection Motivation

Negative and positive state affect were measured using the German version of the Positive and Negative Affect Schedule (PANAS; Krohne et al., 1996; original scale by Watson et al., 1988). This questionnaire comprises 10 positively (e.g., proud, determined, enthusiastic) and 10 negatively (e.g., upset, guilty, hostile) valenced items. Participants indicated to what extent the adjectives describe their current affective state on a five-point scale ranging from 1 (not at all) to 5 (very much). Both scales showed a good internal consistency, with Cronbach’s α = .85 for positive affect and α = .90 for negative affect. Finally, participants indicated whether they agreed with the statement that humans should do everything in their power to protect the environment, rated on a seven-point scale ranging from 1 (do not agree at all) to 7 (totally agree).

Results

All analyses were run with IBM SPSS 27. Initially, data screening showed that there were no outliers exceeding three interquartile ranges. Although some dependent variables were not normally distributed, the sample size was sufficiently large for the analyses to be robust against non-normality (Rasch & Guiard, 2004).

Participants’ environmental protection motivation was high (M = 5.91, SD = 1.12). Four participants (0.4%) gave only one correct response to the attention check questions, 65 participants (7%) gave two correct responses, and 856 participants (92.5%) answered all questions correctly. Table 1 shows descriptive statistics for the responsibility and intentionality attributions for each vignette.

Table 1 Descriptive statistics for moral responsibility in terms of blame (negative side effect) and praise (positive side effect), causal responsibility, and intentionality attributed to the manager and employee depending on the vignette
Full size table

H1: Influence of the Valence of the Side Effect (Knobe Effect)

With respect to intentionality (H1a) and moral responsibility (H1b) attributed to the manager, we expected to replicate the side effect asymmetry. Supporting H1a, negative side effects were perceived as more intentional (M = 4.50, SD = 2.09) than positive side effects (M = 2.34, SD = 1.43), t(803.42) = 18.30, p < .001, d = 1.21. Supporting H1b, negative side effects were also perceived as more blameworthy (M = 4.97, SD = 1.49) than positive side effects were perceived as praiseworthy (M = 4.28, SD = 1.82), t(896.97) = 6.31, p < .001, d = 0.41.

H2: Effects of Social Role on Moral Responsibility Attribution

Supporting H2a, in case of a negative side effect, participants ascribed more blame to the manager (M = 4.97, SD = 1.49) than to the employee (M = 4.06, SD = 1.74), t(455) = 14.74, p < .001, d = 0.69. Supporting H2b, in case of a positive side effect, participants ascribed more praise to the employee (M = 5.21, SD = 1.64) than to the manager (M = 4.28, SD = 1.82), t(468) = 16.84, p < .001, d = 0.78. Results are visualized in Fig. 1.

Fig. 1
figure 1

Moral responsibility (H2; negative side effect = blame; positive side effect = praise) and causal responsibility (H3) attributed to the manager versus the employee depending on the valence of side effect. Vertical bars indicate the standard error of the mean. *** p < .001

Full size image

An exploratory 2 × 2 × 2 (social role × age group × valence of side effect) mixed-measures analysis of variance (ANOVA) with moral responsibility attribution as dependent variable supported the findings of the confirmatory analyses, revealing a large-sized and dominating interaction effect between social role and valence of side effect, F(1, 921) = 496.77, p < .001, ηp2 = .350. This interaction was additionally qualified by age group as indicated by a significant but smaller three-way interaction, F(1, 921) = 8.68, p = .003, ηp2 = .009, because the difference in attributed blame between manager and employee was greater for young agents (ingroup) than middle-aged agents (outgroup) (see also below). There was also a small main effect of valence of side effect, F(1, 921) = 4.95, p = .026, ηp2 = .005, with an overall higher praise than blame score. Also, a small two-way interaction between age group and valence of side effect was found, F(1, 921) = 4.04, p = .045, ηp2 = .004, as blame attributions differed more between young and middle-aged agents than it was the case for praise attributions. No further main or interaction effect reached statistical significance, all Fs(1, 921) ≤ 3.69, all ps ≥ .055, all ηp2 ≤ .004.

H3: Effects of Social Role on Causal Responsibility Attribution

Supporting H3a, in case of a negative side effect, participants ascribed more causal responsibility to the manager (M = 5.02, SD = 1.56) than to the employee (M = 3.95, SD = 1.67), t(455) = 11.55, p < .001, d = 0.54. Supporting H3b, in case of a positive side effect, participants ascribed more causal responsibility to the employee (M = 5.00, SD = 1.67) than to the manager (M = 3.14, SD = 1.57), t(468) = 22.05, p < .001, d = 1.02. Results are visualized in Fig. 1.

Again, an exploratory 2 × 2 × 2 (social role × age group × valence of side effect) ANOVA validated the dominating interaction effect between social role and valence of side effect, F(1, 921) = 550.93, p < .001, ηp2 = .374. This interaction was additionally qualified by age group as indicated by a significant but smaller three-way interaction, F(1, 921) = 7.92, p = .005, ηp2 = .009, because the differences in attributed causal responsibility between manager and employee were greater for young agents (ingroup) than middle-aged agents (outgroup) (see also below). Also, we found a main effect of social role, F(1, 921) = 41.95, p < .001, ηp2 = .044, as more causal responsibility was attributed to the employee (versus manager) when averaging across side effect conditions, and we found a main effect of side effect valence, F(1, 921) = 23.35, p < .001, ηp2 = .025, with an overall higher causal responsibility attributed to the agents in case of a negative (versus positive) side effect of their joint decision. No further main or interaction effect was found, all Fs(1, 921) ≤ 1.80, all ps ≥ .180, all ηp2 ≤ .002.

H4: Main Effect of Age Group on Moral Responsibility Attribution

Participants ascribed more blame to middle-aged outgroup members (M = 4.69, SD = 1.53) than to young ingroup members (M = 4.35, SD = 1.41) in case of a negative side effect, t(454) = 2.46, p = .014, d = 0.23. However, there was no difference in praise attribution between outgroup members (M = 4.71, SD = 1.62) and ingroup members (M = 4.78, SD = 1.63) in case of a positive side effect, t(467) = −0.48, p = .635, d = 0.04. Hence, young participants blamed the middle-aged agents more for harming the environment than they blamed their peers for the same behavior. With regard to praise attributions, however, they did not distinguish between ingroup and outgroup members.

H5: Effects of Age Group on the Asymmetry in Moral Responsibility Attribution

Initially, we calculated role difference scores for each participant by subtracting the moral responsibility attributed to the employee from the moral responsibility attributed to the manager. The use of this difference score as dependent variable allows for a more economic analysis design and facilitates the interpretation of simple main effects, given the significant three-way interaction reported above. A higher difference score indicates a stronger social role effect. We first tested whether the social role effect was actually larger in the ingroup condition (young company workers) compared to the outgroup condition (middle-aged company workers). Supporting H5a, in case of a negative side effect, participants in the ingroup condition (M = 1.11, SD = 1.35) showed a larger difference in the attributed blame to the manager versus the employee, compared to participants in the outgroup condition (M = 0.71, SD = 1.26), t(454) = 3.27, p = .001, d = 0.31. However, contradicting H5b, in case of a positive side effect, participants in the ingroup condition (M = −0.97, SD = 1.14) did not show a larger difference in the attributed praise to the manager versus the employee, compared to participants in the outgroup condition (M = −0.89, SD = 1.24), t(467) = −0.77, p = .444, d = −0.07. Results are visualized in Fig. 2. Therefore, in case of a negative side effect, the social role effect (manager versus employee) for blame attribution was larger when young participants evaluated their peers compared to when they evaluated middle-aged agents. For positive side effects, however, the social role effect for praise attribution was not significantly larger when evaluating ingroup members compared to outgroup members.

Fig. 2
figure 2

Size of the social role effect (ratings for manager minus employee) for attributed moral responsibility (H5) and causal responsibility (H6) depending on the valence of side effect. * p < .05, *** p = .001

Full size image

H6: Effects of Age Group on the Asymmetry in Causal Responsibility Attribution

Again, we initially calculated role difference scores (dependent variable) by subtracting the causal responsibility attributed to the employee from the causal responsibility attributed to the manager. Supporting H6a, in case of a negative side effect, participants in the ingroup condition (M = 1.27, SD = 2.11) showed a higher difference in the attributed causal responsibility to the manager versus the employee, compared to participants in the outgroup condition (M = 0.85, SD = 1.79), t(449.56) = 2.33, p = .020, d = 0.22. However, contradicting H6b, in case of a positive side effect, participants in the ingroup condition (M = −2.00, SD = 1.79) did not show a larger difference in the attributed causal responsibility to the manager versus the employee, compared to participants in the outgroup-condition (M = −1.73, SD = 1.86), t(467) = −1.63, p = .104, d = −0.15. Results are visualized in Fig. 2. Therefore, in case of a negative side effect, the social role effect (manager versus employee) for causal responsibility attribution was larger when the participants evaluated agents belonging to their own age group compared to when they evaluated middle-aged agents, but this was not the case for positive side effects.

H7: The Role of Positive and Negative Affect

To analyze the relationship between affect experienced by the participants and their moral responsibility attributions, we calculated the correlations between blame attribution and negative affect (H7a), and praise attribution and positive affect (H7b). Supporting H7a, in case of a negative side effect, the amount of attributed blame correlated positively with negative affect, rs = .328, p < .001. Supporting H7b, in case of a positive side effect, praise attribution correlated positively with positive affect, rs = .158, p = .001. For exploratory reasons, these correlations were also calculated separately for each social role. Negative affect correlated positively with blame attributed to the manager, rs = .346, p < .001, and with blame attributed to the employee, rs = .265, p < .001. Positive affect correlated positively with praise attributed to the manager, rs = .132, p = .004, and with praise attributed to the employee, rs = .176, p < .001. Hence, negative affect was actually related to blameworthiness, and positive affect was actually related to praiseworthiness, whereas the correlation between negative affect and blame was stronger than the correlation between positive affect and praise.

H8 and H9: Effects of Perceiver Attributes on Moral Responsibility Attributions

With respect to H8, we found that women ascribed more blame (M = 4.64, SD = 1.41) to the agents than men did (M = 4.00, SD = 1.60), t(444) = 3.87, p < .001, d = 0.43, and women ascribed more praise (M = 4.87, SD = 1.57) than men (M = 4.37, SD = 1.72), t(459) = 2.90, p = .004, d = 0.31. Also, women (M = 5.95, SD = 1.10) showed a higher environmental protection motivation than men (M = 5.75, SD = 1.18), t(905) = 2.29, p = .022, d = 0.18. People who did not identify as either male or female were excluded from this analysis because they formed an insufficient subsample. Furthermore, with respect to H9, environmental protection motivation correlated with overall blame, rs = .450, p < .001, and with overall praise, rs = .224, p < .001.

Discussion

Replication of the Knobe Effect and the Effect of Social Roles

We successfully replicated the Knobe effect and the social role effect, showing that attributions of intentionality and moral responsibility depend on the valence of a decisions’ side effect (H1), and that moral and causal responsibility attributions depend on the agents’ social role (H2 and H3). These findings add to the work of Kaspar et al. (2016) and Willemsen et al. (2018), confirming the attribution patterns for company decisions already observed. The manager’s moral responsibility and intentionality ratings were higher for negative than for positive side effects, reflecting a basic finding from the original scenario of Knobe (2003). In the negative side effect condition, participants strongly agreed that the manager intentionally harmed the environment and thus deserved substantial blame. In the positive side effect condition, they perceived that the manager helped the environment with little intention, which made him unworthy of much praise. Moreover, higher status roles (managers), compared to lower status roles (employees), apparently deserve more blame for a decision’s negative side effects on the environment, whereas lower status roles deserve more praise for positive side effects. Both effects can be explained in terms of default points for moral behavior that determine moral right- and wrongfulness, and thus the direction of the responsibility attributions (Kaspar et al., 2016; Knobe, 2010; Pettit & Knobe, 2009). These default points appear to be stricter for roles ranking higher in the company’s hierarchy, such as the manager, and for potentially harmful side effects, such as harming the environment. This explanation also fits the culpable control model (Alicke, 2000; Alicke et al., 2011), which postulates that the manager has more control over the company processes and is therefore normatively expected to prevent harm and to promote help. Violating these role-related expectations now quickly surpasses the manager’s moral threshold. In contrast, meeting role-related expectations represents only normative behavior that does not merit special praise (Wright & Bengson, 2009). For the employee, on the other side, the default point is lower, which makes it easier for them to merit praise and to escape responsibility for morally bad actions. Hence, theoretical accounts explain the responsibility attribution patterns in terms of varying normative expectations set by differing moral standards for different social roles and different side effects.

The Impact of Age Group

Aside from social role, we introduced age groups as an additional social context variable that was assumed to influence responsibility attributions. First of all, we found a main effect of age group on the attribution of moral responsibility (H4): The young participants of our sample blamed the middle-aged agents more for harming the environment than they blamed their peers for the same behavior. This finding reflects the generational divide concerning environmental protection, where young people are assumed to display more pro-environmental attitudes and behavior than older people (Gray et al., 2019). Although this is not necessarily true (Gifford & Nilsson, 2014), this widespread expectation might have influenced participants’ evaluations. However, participants did not ascribe more praise to young people for helping the environment than to middle-aged people. From the perspective of the culpable control model (Alicke, 2000; Alicke et al., 2011), if participants expected young people to protect the environment, it is plausible that simply adhering to this norm did not lead to “bonus points”. For middle-aged people, however, there might have been a positive surprise effect which attenuated their overall blameworthiness, ultimately leading to similar praise attributions to ingroup and outgroup members. Thus, the effect of age group on blame attributions, and the lack of an age group effect on praise attributions, could both be traced back to differing normative expectations.

Moreover, the age of the company workers influenced the size of the social role effect, but only in the negative side effect condition: The differences in attributed blame between the manager and the employee (H5) as well as the difference in attributed causal responsibility between manager and employee (H6) were larger when young participants evaluated other young people (ingroup members) compared to when they evaluated people from an older age group (outgroup members). In contrast, in case of a positive side effect, the size of the social role effect was independent of the agents’ age. These findings indicate that the divergence between outgroup and ingroup perception extents to status differences. The results for the negative side effect are in line with prior findings on outgroup homogeneity (e.g., Bente et al., 2016; Linville et al., 1989; Rubin & Badea, 2007) and collective responsibility (e.g., Lickel et al., 2003; Lickel et al., 2006). Participants may have perceived outgroup members as more similar and were more likely to rate them regardless of their social role. A complementary explanation would be that the ingroup was perceived to be more heterogeneous than the outgroup (Mullen & Hu, 1989). Participants in the ingroup condition might have been more susceptible for differences between the agents’ social roles. In addition, as older generations are sometimes depicted as careless and negatively impacting the environment (Chazan & Baldwin, 2019), participants might have based their judgment on the (unconscious) impression that the outgroup was guilty as a whole. Hence, for the negative side effect, outgroup homogeneity might be reflected by less differentiated blame perception.

Some more reasons for this result pattern are conceivable: Ingroup identification is a powerful moderator of outgroup homogeneity effects (De Cremer, 2001; Lickel et al., 2006). However, the study did not assess participants’ identification with their age cohort. Assuming that adolescents identify with their age groups only to a small extent, it is possible that the intergroup effects would have been more pronounced for groups with higher identification potential. Furthermore, Pizarro et al. (2003) offered a possible explanation for the lack of social group influence on praise attributions by suggesting that blame and praise are processed differently, with different underlying motives. As harmful events prompt a search for blame-validating evidence (Alicke, 2000), the agents’ age was likely used as an additional cue to determine blame ascriptions because of the expected association between environmentally harmful behavior and age. For the positive side effect, however, in the absence of harmful events, participants might have been less motivated to search for cues to determine the agents’ responsibility. Consequently, if participants paid less attention to the agents’ age group in the positive side effect condition, they might have based their responsibility judgments solely on social role differences so that intergroup effects did not have a chance to play out fully.

The Role of Experienced Affect

Our results showed that higher negative affect experienced by the participants correlated positively with perceived blameworthiness of an agent’s decision (H7a), whereas positive affect correlated positively with attributed praise (H7b). However, negative affect and blame were stronger related than positive affect and praise. These findings are in line with the culpable control model (Alicke, 2000; Alicke et al., 2011) suggesting that affective processes play a driving role in responsibility attribution processes. While the relationship between negative affective states and blame attribution has already been empirically established (Bright & Goodman-Delahunty, 2006; Quigley & Tedeschi, 1996), research on the relationship between positive affect and praise attribution has been scarce so far. Nonetheless, the present data indicates that the ascription of moral responsibility varies with both negative and positive affective reactions, complementing recent studies on intentionality attributions and emotional processing (Zucchelli et al., 2019).

The contribution of affective mechanisms could also explain why individuals amplify ratings of intentionality and moral responsibility for negative outcomes to a larger extent than for positive outcomes (Alicke et al., 2008; Mazzocco et al., 2004; Wright & Bengson, 2009). Our data suggests that the relationship between positive affect and praise is weaker than the relationship between negative affect and blame. If moral responsibility attributions are partially driven by affect, and negative affect possesses more power over responsibility attributions than positive affect, it is plausible that the desire to blame some agents for negative outcomes is stronger than people’s desire to praise some agents for positive outcomes, as the former is additionally fueled by negative affect. This could be one more reason for the side effect asymmetry that should be scrutinized by future research.

Gender and Environmental Protection Effects

We exploratively analyzed the effect of gender on moral responsibility attributions for each side effect (H8). For both side effect conditions, women attributed higher moral responsibility to the agents than men did. Moreover, women reported a slightly higher environmental protection motivation than men, and environmental protection motivation was positively related to blame and praise attributions (H9). Prior studies confirmed that women show higher levels of environmental concern, pro-environmental attitudes, and pro-environmental behavior than men (Gifford & Nilsson, 2014). Women therefore might be more motivated to penalize harmful behavior, while also encouraging morally good behavior by ascribing praise.

Implications

This study’s findings entail several theoretical and practical implications. First, the study confirms the robustness and persistence of social role effects, which were long overlooked in responsibility attribution research. With medium- to large-sized effects, the social role effect appears to be very stable across studies and cultures (cf. Kaspar et al., 2016; Willemsen et al., 2018). This underscores the conclusion from Willemsen et al. (2018) that theories on responsibility attribution should systematically analyze the impact of social roles and the normative expectations associated with them. To achieve a more complete picture, it is essential to account for intergroup effects as well. Different social groups carry different expectations, which influence their moral and social evaluations (Hertel et al., 2002; Song et al., 2018). Ideally, intergroup research and research on responsibility attribution and social roles should go hand in hand to gain further insights into the variety of factors influencing causal and moral responsibility evaluations.

Moreover, moral responsibility research should include the contributions of affective reactions of the perceivers evaluating the agents. We therefore propose to include measures of affect into studies on moral responsibility. Moreover, the results on gender effects suggest that a systematic examination of participants’ demographic characteristics promises to provide valuable insights into the dynamics of moral evaluations.

Additionally, the findings reflect the current social debates on environmental protection and the climate crisis. The generational divide portrayed by media seems to persist and to influence responsibility attributions, especially for environmentally harmful events. Blame attributions do not only depend on the agents’ hierarchical status within their company, but also on their age. This means that if company decisions lead to negative environmental side effects, most blame will likely fall on senior CEOs or other middle-aged people ranking high in companies’ hierarchies. Apparently, older people must fight even harder to disprove the negative expectations of young people. The present study therefore acts as application of the culpable control model to environmental responsibility issues, which is an extended and new context for moral responsibility research.

In summary, the results complement the framework on moral responsibility attribution and provide further evidence for the importance of spontaneous evaluations and affect in the culpable control model. The study demonstrates that social context variables should no longer be neglected in moral responsibility research, and that these comprise not only social role effects, but also social identification processes.

Limitations and Future Directions

The findings are constrained by several methodological and thematic limitations. First, the operationalization of social groups as age groups entails the confounding variable of age stereotypes. Elderly people are often perceived as high in warmth, but low in competence (Fiske et al., 2002), and it is conceivable that young people might get more credit for having less working experience. This might have affected the moral standards associated with them, for example, because perceptions of incompetence might have increased blame attributions for the middle-aged agents or because young agents were evaluated more leniently. Since the study design did not allow disentangling effects of age from more general effects of intergroup perception, it remains uncertain whether they can be generalized to other forms of social groups. Moreover, the sample was limited to young people evaluating either young or middle-aged agents, so we cannot say anything about possible evaluation asymmetries among older people.

An additional limitation of the study design is that the gender of the company workers was not varied. Only (German) male terms were used to designate the agents’ roles, which might have influenced the responsibility ratings. The findings suggest that women tend to attribute higher moral responsibility to the agents. Given the demographic composition of the sample, with women constituting the large majority, this result might originate from more general gender perception effects. It is conceivable that women judge men stricter than they would have judged other women.

Similar shortcomings concern the nature of the side effects. Although the social role effect and the Knobe effect extend to positive and negative side effects for clients (Kaspar et al., 2016), it is not possible to infer whether the effects of age group, affect, and gender are also applicable to responsibility attribution contexts other than environmental issues. For instance, we do not know whether the responsibility attribution patterns would change if the decision’s side effects concerned particular subgroups like immigrants. So far, the findings are thus limited to the particular scenario and design.

Based on the listed limitations, we propose several starting points for future research. For instance, future studies should analyze social groups with higher identification potential than age groups, such as sports clubs or ethnic groups, to emphasize the intergroup aspect of the evaluations. This could reinforce the social group effects, and at the same time, it would be possible to test whether the findings generalize to different social groups. To further explore gender effects, future studies should vary the gender of the agents and compare how responsibility attributions change if male and female participants evaluate male versus female company workers. Another possible modification for future studies consists in turning the social role manipulation into a between-participants factor instead of a within-participants factor. This would allow to test whether the social role effect persists beyond the communicative aspects of a joint decision-making scenario.

Conclusion

Given the current intergenerational debate on the climate crisis, it is essential to understand how younger generations ascribe responsibility to companies and their workers for harming or helping the environment. The findings replicate asymmetries related to the valence of side effect and social roles, and add new insights to previous research on responsibility attribution by revealing effects of social group, gender, and affective state, which have been neglected until now. Theories on responsibility attribution and future research should consider the larger social context, the characteristics of the perceivers, and the associated normative expectations. Instead of focusing exclusively on the cognitive mechanisms of responsibility attribution, it is worthwhile to include the contributions of affective mechanisms as well. This approach offers promising new directions to disclose the driving factors of moral responsibility attribution.

Authors’ Contributions (CRediT Roles)

KR and KK conceptualized the study and research methodology; KK provided the resources and supervised the study; KR collected the data; KR and KK analyzed the data; KR and KK wrote the manuscript; KR and KK approved submission.