Intergroup conflict, including the extreme case of war, is ubiquitous in human social life [1–3]. Given the importance of understanding the nature of intergroup conflict, researchers from diverse disciplines, such as sociology, political science, economics, and other social sciences, have conducted studies to unravel its causes or reasons [4–6]. In social psychology, intergroup conflict has been attributed to intergroup bias [7, 8], which is the human tendency to behave altruistically or cooperatively toward members of the group to which they belong (i.e., in-group) and/or competitively or aggressively toward members of the group to which they do not belong (i.e., out-group). Various theories have been proposed as a psychological explanation of it [8, 9–11].
Recent experimental studies in social and evolutionary psychology have revealed that intergroup bias mainly stems from in-group cooperation, and out-group aggression rarely occurs in economic games [12–14]. However, based on an evolutionary perspective , the absence of out-group aggression in previous studies may be due to the lack of a critical condition, such as a cue that attacking the out-group would increase the benefit to the in-group. Thus, the current study introduces such a cue to an experimental game paradigm and examines whether out-group aggression emerges.
1.1 Coevolutionary model of intergroup bias
Intergroup bias is present even between arbitrarily created experimental groups (i.e., minimal groups) . Intergroup bias in minimal groups—in which no conflict of interest, negative stereotypes, or face-to-face interactions within or between groups exist—has been consistently observed in student  and non-student samples  across various cultures [19, 20]. Evolutionary and social psychologists have formed several hypotheses on the evolutionary origin of intergroup bias [21, 22].
One of the evolutionary hypotheses explaining intergroup bias is the coevolution model (CO model) [15, 23, 24]. The CO model holds that humans simultaneously acquire in-group cooperation and out-group aggression. Choi and Bowles  used an agent-based simulation and showed that agents who displayed both in-group and out-group aggression thrived, suggesting that in-group and out-group aggression co-evolved. The validity of the CO model has been examined using various datasets, ranging from ethnohistorical and archaeological  to experimental [26, 27], although evidence for the model has been mixed [14, 28–30].
A feature of the CO model that distinguishes it from other evolutionary models of intergroup bias  is the assertion of the existence of out-group aggression. Previous studies have typically employed the intergroup prisoners’ dilemma–maximizing difference game (IPD–MD)  to test the CO model experimentally. In the IPD–MD, players are randomly divided into two groups. They receive a fixed amount of money from the experimenter. They decide how much to keep for themselves and to invest in the two pools: the within-group and the between-group pools. The total amount of money invested in the within-group pool by the members of the group is doubled and divided among in-group members equally. In contrast, the total amount of money invested in the between-group pool is doubled; all in-group members receive an equal split and out-group members lose half the amount of the money (equal to the total invested amount before doubling). In other words, any investment in the between-group pool includes harming out-group members, i.e., out-group aggression. If people are inclined to pursue self-interest, they would keep the entire amount of the initial endowment for themselves. If people are inclined only to cooperate with in-group members, they will invest their money in the within-group and/or between-group pools, without distinguishing between the within- and between-group pools. Meanwhile, if people intend to cooperate with in-group members and harm out-group members, they would invest in the between-group pool that reduces the resource of out-group members.
The CO model predicts that individuals are inclined to invest more in the between-group pool than in the within-group pool in the IPD–MD . Nevertheless, several studies have revealed that individuals, in general, invest more money in the within-group pool than in the between-group pool [12, 32–35]. The same pattern has also been observed in studies in which researchers modified the game to encourage investment in the between-group pools [36–41]. Their findings offer conflicting evidence for the CO model: the empirical evidence does not suggest that individuals simultaneously acquire the tendency for in-group cooperation and out-group aggression.
1.2 Out-group aggression in the IPD–MD
Behavioral evidence from studies using the IPD–MD does not support the CO model. Nonetheless, in-group and out-group aggression not evolving together would be a premature conclusion given the uncertainty of whether out-group aggression in the IPD–MD (i.e., investment in the between-group pool) corresponds to the operationalization of this concept in the CO model. Thus, the present study argues the vital importance of revisiting whether the IPD–MD offers contextual cues to elicit out-group aggression that is arguably acquired during evolution.
Human minds are thought to have domain-general and domain-specific features [42, 43], and evolutionarily acquired psychological mechanisms are often characterized as domain specific . For instance, the percentage of accuracy in performing logical reasoning tasks increases when the task is framed to detect norm violators [45, 46]. That is, “cheater detection” is thought to be an adaptive reasoning system specific to the domain of social exchange . Previous studies have identified various domain-specific psychological systems , including theory of mind , intuitive physics , and folk biology . Most importantly, intergroup bias is also considered a domain-specific feature .
Situational cues that imply a certain adaptive task reportedly activate a corresponding domain-specific mindset and behavior. An example is in the way disease priming, which induces pathogen infection threat assessment, leads to the preference for a symmetrical face, which indicates a high-functioning immune system , and negative attitudes toward immigrants, who are associated with the possibility of unknown pathogens [53, 54]. Cues of an important social exchange situation may also facilitate cooperation in a prisoner’s dilemma game [55, 56]. If intergroup bias stems from an evolutionarily acquired domain-specific mindset , then out-group aggression may be triggered by the cue of an adaptation task, such as intergroup conflict.
Choi and Bowles  described out-group aggressions predicted by the CO model as in-group-benefitting behavior. One of the main causes of intergroup conflicts is the competition for resources, such as food, territory, and mating opportunities [57–59]. The in-group-benefitting aspect of out-group aggression is an important tenet of out-group aggression in the CO model. Several evolutionary models and simulations have suggested that out-group aggression might evolve when individuals gain resources from an out-group [24, 60]. Moreover, people display hostile intergroup behavior when out-group resources can be acquired ; people also tend to be sensitive to threats arising from hostile intergroup relationships [62, 63]. These findings suggest that out-group aggression is favored when it benefits the in-group.
However, out-group aggression in IPD–MD seems to show a discrepancy with how it has been theoretically defined in the CO model. More specifically, previous studies using the IPD–MD have operationalized out-group aggression as simply reducing the payoff of out-group members: investment in the between-group pool is typically framed as merely about increasing and decreasing payoffs for in-group and out-group members, respectively . Thus, the in-group-benefitting aspect of out-group aggression seems to be missing in the IPD–MD, and this might be why previous studies using the game have yielded inconsistent results with the CO model. In other words, if out-group aggression in an IPD–MD (i.e., investment in the between-group pool) is framed such that it leads to financial gain for the in-group, then out-group aggression would emerge, which provides empirical support for the CO model.
1.3 The present research
The present research aimed to test the effect of framing in an IPD–MD and examine whether out-group aggression would emerge when the contribution to the between-group pool is framed as having benefits to the in-group. More specifically, this study formulated two conditions that varied in the framing of the between-group pool: a control condition and a steal condition. In the former, the instructions on the two pools were identical to that of a basic IPD–MD. In the latter condition, the investment in the between-group pool was framed such that it led to stealing money from an out-group for their own group.
The current research tested two hypotheses regarding the framing effect on out-group aggression. Böhm  proposed the “weak” and “strong” hypotheses of out-group aggression based on the CO model: the amount invested in the between-group pool should be interpreted not only in terms of its absolute amount but also as a relative amount compared with the amount invested in the within-group pool, and “a larger contribution to the between-group pool relative to the within-group pool may be interpreted as negative attitude toward the out-group, an equal contribution to both pools as indifference regarding the out-group’s welfare, and a smaller contribution to the between-group pool as positive attitude toward the out-group”  (p. 2). Based on the CO model, a negative attitude toward the out-group can be expected to lead to aggressive behavior in the steal-framing condition. Therefore, the present study set the “strong” hypothesis (hypothesis S) as follows:
Hypothesis S: In the steal condition, individuals would contribute more to the between-group pool than to the within-group pool.
Even if hypothesis S is not supported, as in previous studies [36, 38, 40, 64], steal framing may nonetheless lead to a less positive attitude. Therefore, the current study set the “weak” hypothesis (hypothesis W) as follows:
Hypothesis W: Individuals in the steal condition would invest more money in the between-group pool than those in the control condition.
The research aimed to further explore how the steal framing would influence out-group aggression. Framing effects in experimental economic games can be divided into effects of changes in preference and belief . Preference tends to be stable regardless of the actions and reactions of an interacting partner. Meanwhile, belief refers to an expectation about others’ behavior. If the framing affected the investment in an IPD–MD, then whether the effect would be attributable to changes in preferences and/or beliefs merits investigation.