Heroism is the original topic of literature, as evidenced by some of the earliest known human writing from approximately 2100 BC in the Epic of Gilgamesh. Yet, an empirical understanding of heroism is only just emerging. This research has variously investigated types of heroism (Franco et al., 2011), functions of heroism (Kinsella et al., 2015a), traits of heroes (Goethals and Allison, 2012), characteristics of heroes (Kinsella et al., 2015b), and gender differences among heroes (Becker and Eagly, 2004), as well as amalgams of these approaches (Riches, 2018). Allison et al. (2016) summarize a number of dichotomies made in the literature which try to distinguish between two classes of heroes (e.g., emergent vs. sustained; Kraft-Todd and Rand, 2016). In light of this research, an early consensus definition of heroism seems to be taking extraordinary action in service of the greater good with personal risk of significant sacrifice (Allison et al., 2016).

Thus articulated, the burgeoning science of heroism appears to sit squarely between two social science literatures: game theory and social norms. Game theory formalizes strategic decision-making between individuals by quantifying the costs and benefits at stake in an interaction (Von Neumann, 1959). This conceptualization allows for a precise definition of cooperation—an individual paying a cost to give another a benefit—which in turn presents a conundrum: why do people cooperate (Rand and Nowak, 2013)? A particularly challenging problem in game theoretic terms is understanding why an individual would pay a cost to give many others a benefit, i.e., contribute to public goods (Hardin, 1968)—a pressing problem shared by policy-makers in the real world (Kraft-Todd et al., 2015). In the language of game theory, then, the risk of sacrifice in heroism implies a potentially large cost paid by an actor in order to cooperate or contribute to the public good. Heroism thus may be understood as a special case of cooperation in which the actor incurs (or at least risks) a large cost (akin to extreme altruism; Marsh et al., 2014; Rand and Epstein, 2014). Further, there is good reason to believe that assessment of the costs and benefits might be relevant to our perception of heroism. Adults and children use information about the costs and benefits of others’ behaviors to make inferences about their character (Jara-Ettinger et al., 2016). As early as 2 years of age, these evaluations affect our preferences for interacting with others (Jara-Ettinger et al., 2015). Thus, ascriptions of heroism may rely on beliefs about the costs and benefits of an actor’s behavior.

Social norms are “rules and standards that are understood by member of a group, and that guide and/or constrain social behavior without the force of laws” (Cialdini and Trost, 1998). Two types of social norms are frequently distinguished: descriptive norms, which are about what people think others do; and injunctive norms, which are about what people believe others think they should do. Colloquially, our conception of what is “normal” lies somewhere between our conception of what is descriptively and injunctively normative (Bear and Knobe, 2016). In the language of social norms, then, the extraordinary action that defines heroism is descriptively non-normative (i.e., rare).

Conceptually situated within this overlap of game theory and social norms, four quantifiable dimensions of social perception may help elucidate a clearer empirical understanding of heroic behavior: (1) the cost to the actor; (2) the benefit to the recipient(s); (3) the descriptive normativity of the behavior; and (4) the injunctive normativity of the behavior. In a series of exploratory studies (total N = 408), we aim to discover the extent to which these constructs influence people’s perceptions (i.e., lay intuitions) of heroism.

Intuitively, it seems likely that the more a behavior is thought to be heroic, the greater would be the perceived cost to the actor and benefit to the recipient, while the lower would be the descriptive and injunctive normativity of the behavior. We use a “ground-up” approach to the concept of heroism, avoiding a priori assumptions about what “counts” as heroism (similar to the method of Kinsella et al., 2015a). In Study 1, we therefore ask subjects to generate acts of heroism. In Study 2, we ask a separate group of subjects to rate the extent to which these candidate behaviors are heroic. Finally, in Study 3, we ask yet another group of subjects to rate the extent to which a subset of these candidate behaviors are costly to the actor, beneficial to the recipient, descriptively normative, and injunctively normative.

Situating an empirical approach to heroism within the game theory and social norms literatures, we conducted three exploratory studies (total N = 408) on lay intuitions of heroic acts. We find that acts which are widely agreed upon as being heroic (exemplary heroism) can be distinguished from acts whose heroicness is unclear (ambiguous heroism), with exemplary acts having lower descriptive normativity and higher costliness to the actor—but not differential injunctive normativity nor benefit to the recipient. These results extend prior work on heroism by providing empirical evidence supporting the conceptual link between the emerging science of heroism and more established fields in the social sciences, while also clarifying the lay definition of heroism.

Our approach avoids being bound by academic preconceptions of heroism by utilizing subject-generated acts of heroism as our stimuli (Study 1, in which we asked, “Please name at least 3 and up to 10 real-life acts of heroism”) in subsequent studies. Though a few empirical studies of heroism employ this method, these investigate other aspects of heroism: e.g., “In your view, what functions do heroes serve?” (Kinsella et al., 2015a) and “What are the features that you associate with heroes and their heroic actions?” (Kinsella et al., 2015b). We believe a comprehensive understanding of heroism will be achieved by exploring various means of eliciting lay perceptions of heroism and finding consensus among them.

One major limitation to our investigation is the extent to which our results depend on the specific ten behaviors from Study 2 we chose to serve as stimuli in Study 3. Future research should test the robustness of our conclusions to the consideration of a wider range of heroic acts, and perhaps conduct additional pretesting to ensure candidate acts of heroism do not differ on dimensions (e.g., familiarity) which might affect variables of interest. In particular, the inclusion of candidate acts of heroism with intermediate values of the proportion of people considering them heroic (compared to relative extremes we investigate here)—and the inclusion of acts considered distinctly non-heroic—may shed further light on the relation of perceived heroism and our variables of interest.

Additionally, because our results are drawn from a convenience sample of mTurk workers, they may be culturally specific (in accordance with the WEIRD hypothesis; Henrich et al., 2010). Given that heroism is a socially designated role, it is reasonable to think that lay perceptions of heroism would be influenced by cultural norms; for these reasons, future work might seek to assess this phenomenon cross-culturally. It is worth noting that our investigation is on lay perceptions of heroism, rather than the decision-making process of heroes, and so future work might also explore the extent to which the variables we consider are relevant this process, or whether they are the result of more generalized intuitions (Glanville and Paxton, 2007; Rand and Epstein, 2014). One other potentially interesting question for future research would be to compare first-order beliefs about heroism (as we do here) with second-order beliefs (i.e., what subjects think others think about heroism), as the latter have been demonstrated to have a greater impact on prosocial behavior than the former (Jachimowicz et al., 2018). Finally, our investigation was exploratory, so replication and confirmatory studies should be conducted to provide greater faith in our findings and their interpretation.

We found the observation that exemplary acts of heroism were not perceived as more beneficial to recipients than ambiguous acts of heroism to be surprising, given our intuition that helping others is part what makes an act heroic [e.g., “a person jumping on a grenade to save fellow soldiers” m = 69.08 95% CI (62.51, 75.65) vs. “volunteering at a soup kitchen” m = 65.84, 95% CI (60.53, 71.16), t(200) = 0.76, p = 0.45]. The within-subjects design of Study 3 grants additional credence to this observation: each subject rated all ten of the heroic acts, and so presumably they could have compared one situation to the next and made these judgments relative to each other. Yet, the finding interestingly coincides with other empirical findings. For example, subjects participating in prisoner’s dilemmas are highly influenced by the framing of the situation and potential behaviors, beyond the simple material outcomes of the interaction (e.g., Zhong et al., 2007), and so too here might verbal associations weigh more heavily than numerical assessments in subjects’ judgments. In charitable giving, the “effective altruism” movement aims to direct giving toward more socially efficient causes—i.e., get more bang for the donor’s buck—yet effectiveness information often does not motivate greater giving (Berman et al., 2018). This finding is an example of the broader phenomenon of scope insensitivity (Carson, 1997), in which people do not exhibit greater valuation for increased amounts of an economic good. Scope insensitivity has been repeatedly demonstrated in the domain of prosociality (Desvousges et al., 1993; Hsee and Rottenstreich, 2004; Small et al., 2007). Thus, our findings are in a sense the converse: while previous research has shown that people do not value (via monetary donations) causes which provide a greater benefit to others, we show that people do not perceive a greater benefit to others from behaviors that are more valued (via judgments of heroism). Further, and more relevant to the characterological judgment nature of heroism, this (non)relation of social benefit to valuation is consistent with findings that people do not prefer consequentialist agents who are willing to inflict harm to provide a greater social benefit (Everett et al., 2018). Taken together, these many potential reasons for our surprising finding that exemplary acts of heroism were not perceived as more beneficial to recipients than ambiguous acts of heroism suggest a promising avenue for future research.

Our finding that judgments of heroism are linked to the cost to the actor but not the benefit to the recipient suggests numerous questions regarding the proximate mechanism of heroism perception. First, when a decision-maker is attempting to distinguish whether another’s behavior is heroic or not, it could be that the costs to the actor are more salient than the benefits to the recipient if this judgment is accomplished via imagining what it is like to be in the actor’s shoes (rather than the recipient’s). For example, it could be that when you are trying to decide whether “entering a burning building to save someone” is heroic or not, you engage in perspective-taking not with the person who might be saved, but with the person entering the burning building. Second, it could be that the costs of heroism are simply more observable than the benefits because calculating the latter requires an extra step of contrapositive reasoning: i.e., it requires knowing what would have happened if the hero had not intervened. For example, when a child stands up for another child being bullied, we know that child steps in the way of the bully’s fists, but we don’t know whether the bully would have broken the victim’s nose or just taken their lunch money.

Our finding that judgments of heroism are linked to the descriptive normativity of the action but not the injunctive normativity was also surprising to us, as our intuition was that “going above and beyond” was an important part of being seen as heroic. Our data indicate, however, that this is not the case. Many of the proposed acts of heroism in Study 1 included professions where taking risks to help others is part of the job expectations (e.g., military, firefighter; see Table 1 and Supplementary Table S1). Thus, for these people acting heroically may not be unexpected (i.e., is injunctively normative), but it still may be rare (i.e., is descriptively non-normative). The fact that such actions were still judged to be heroic indicates that unexpectedness (or injunctive normativity) does not appear to be a crucial component of lay perceptions of heroism.

Heroism, understood as rare (i.e., non-normative) and costly cooperation is a particularly timely concept to understand as the need to promote innovative solutions to global social challenges becomes increasingly clear (Kraft-Todd et al., 2018). We hope our conceptualization of heroism can help connect the emerging science to such pressing real-world issues. Heroism needn’t be confined to our cultural mythologies (Campbell, 1949/2008); we may find that we can cultivate it more effectively if we celebrate it in our science as well as in our stories.

All data are publicly available at: https://osf.io/be8mn/.

GK-T and DR designed the online experiments. GK-T conducted the online experiments and analyzed the results. GK-T and DR wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.