Abstract
Abstract
Previous research has demonstrated the tendency for humans to anthropomorphize computers—that is, to react to computers as social actors, despite knowing that the computers are mere machines. In the present research, we examined the attribution of both primary (non-uniquely human) and secondary (human-like) emotions to ingroup (teammate) and outgroup (opponent) computer-controlled characters in a video game. We found that participants perceived the teammate character as experiencing more secondary emotions than the opponent character, but that they perceived the teammate and opponent character as experiencing equal levels of primary emotions. Thus, participants anthropomorphized the ingroup character to a greater extent than the outgroup character. These results imply that computers' “emotions” are treated with a similar ingroup/outgroup social regard as the emotions of actual humans.
Introduction
The Computers Are Social Actors (CASA) theory postulates that humans' emotions, thoughts, and motivations toward computers are similar to their emotions, thoughts, and motivations toward other humans, despite their objective awareness that computers do not actually possess human-like characteristics. 3 For example, people react to the breakdown or loss of a computer with the same language used to describe the death of a living thing, 4 demonstrate antagonism toward computers that seem to be “cheating” them, 5 and perceive differences in computers' “personalities” based on minimal cues. 6
Recent research has extended the CASA theory beyond humans' perceptions of individual computers to humans' perceptions of entire virtual reality environments. This research demonstrates that in video gaming contexts, players often anthropomorphize virtual representations (i.e., avatars) of themselves and others even when the “other players” are controlled by a computer rather than by humans.7,8 Significantly, however, players are more likely to anthropomorphize characters who are portrayed as ingroup members (e.g., teammates, sympathetic characters) than characters who are portrayed as outgroup members (e.g., opponents, unsympathetic characters). For example, when playing video games involving sympathetic and unsympathetic characters, people exhibit greater willingness to help the sympathetic characters 9 and greater discomfort with harming them.10–12
One specific manifestation of anthropomorphism, which has yet to be studied in a virtual reality context, is the attribution of emotions to video game characters—that is, the extent to which people perceive video game characters as experiencing human-like versus non-uniquely human emotions. Primary emotions, such as excitement or fear, are thought to be experienced by non-humans (i.e., animals) and humans alike, as these emotions do not necessarily reflect higher-level thinking. Secondary emotions such as despair or compassion, however, are thought to be experienced only by humans.
In research on intergroup relations, people have been found to attribute fewer secondary emotions to outgroup members than to ingroup members, an indicator that people anthropomorphize outgroups to a lesser extent than ingroups.13,14 Given that computers and video game characters are often treated as social agents, 3 the tendency for people to attribute fewer secondary (human-like) emotions to outgroups than ingroups may also extend to video game characters. Such a finding would demonstrate for the first time that people ascribe particular types of emotions—namely, emotions used to describe humans—to characters that on the surface bear little resemblance to human beings.
Overview and Hypotheses
In the present study, participants played a video game with two computer-controlled characters: one teammate and one opponent. After playing, they were asked about the degree to which they attributed various emotions to each character. Previous research has suggested that people make distinctions between ingroup and outgroup virtual characters in general.10,11 Extending this phenomenon to emotion attributions specifically, we hypothesized that participants would attribute fewer secondary emotions to the outgroup character (opponent) than the ingroup (teammate) character (Hypothesis 1). We also hypothesized that there would be no difference in participants' attributions of primary emotions to the characters (Hypothesis 2).
Method
Participants and design
Forty-seven undergraduate students from the University of Chicago (25 men and 22 women) participated in exchange for partial course credit. All participants rated the emotions of both the ingroup (teammate) and outgroup (opponent) characters; thus, the design was within subjects.
Materials and procedure
The competition game was Super Smash Bros for the Nintendo 64. In this game, characters hit and shoot each other in a cartoonish, low-violence fashion to knock each other off a platform as many times as possible within a given time limit. After the characters are knocked off, they return to the platform and keep competing until time runs out. In the configuration used for this experiment, there were three characters: the player, one computer-controlled “teammate,” and one computer-controlled “opponent.” The opponent was programmed to attack both the player and the teammate, whereas the teammate was programmed only to attack the opponent. In addition, the player and the teammate were rendered unable to harm each other through either accidental or purposeful attacks.
Participants were presented with the video game and a guide sheet explaining which controls to use for various character movements. They were then given 3 minutes to practice guiding their character around the screen and using the different moves in a non-competitive environment.
After completing the practice round, participants were shown a screen showing their own as well as two other humanoid, robot-like characters. The three characters were identical except for coloration. The game announced that this would be a “Team Battle,” in which both the participant and the teammate character were dressed in green, and the opponent character was dressed in blue to encourage ingroup/outgroup cognition. 15 Verbal instructions then introduced the player to their teammate (“Samus”) and their opponent (“Aran”). The participant played for 5 minutes, and the game was turned off before the participant could see the final score. This was done to avoid the possibility that the outcome of the game would influence participants' anthropomorphism of the characters.
Participants then completed measures of primary and secondary emotion attributions for both the teammate and opponent characters. Specifically, they rated how much they thought each character experienced each specific emotion (1=not at all, 7=very much so). The measure for each character contained six primary emotions (excited, joyful, fearful, pain, pleasure, and enraged) and six secondary emotions (guilty, friendly, compassionate, hopeful, remorseful, and despair), which were directly taken from previous research.13,14 Their ratings were averaged into separate composites for the teammate's primary emotions (α=0.74), the teammate's secondary emotions (α=0.78), the opponent's primary emotions (α=0.67), and the opponent's secondary emotions (α=0.65), with higher scores on each measure reflecting a greater attribution.
Finally, the participants completed a demographic survey. They were then probed for suspicion and debriefed.
Results
We hypothesized that participants would attribute more secondary (human-like) emotions to the ingroup (teammate) character than to the outgroup (opponent) character (Hypothesis 1), but that there would be no differences in the attribution of primary (non-uniquely human) emotions to the teammate and opponent characters (Hypothesis 2). Supporting Hypothesis 1, a repeated-measures analysis of variance revealed that participants rated the teammate character as experiencing significantly more secondary emotions (M=2.61, SD=0.93)than the opponent character (M=2.38, SD=0.83), F(1, 46)=6.47, p=0.01, ηρ2=0.12. However, and supporting Hypothesis 2, participants rated the teammate character (M=3.35, SD=1.07) and the opponent character (M=3.38, SD=1.03) as experiencing equal levels of primary emotions, F(1, 46)=0.10, p=0.75, ηρ2=0.002.
Discussion
According to the CASA theory, 3 people view and treat computers in a similar fashion as they do human beings. In this study, we examined a particular aspect of human–computer interactions: how people perceive the “emotions” of computers. In doing so, we applied a well-established finding from the intergroup relations literature—that people attribute more human-like qualities (including emotions) to ingroup members than outgroup members13,14—beyond human beings to virtual characters. The results of the present study confirmed our predictions: Overall, participants ascribed more secondary (human-like) emotions to the teammate character than to the opponent character, whereas they showed no differences in their ascriptions of primary emotions (which are attributed to humans and animals alike) to the two characters.
These findings emerged despite the fact that the teammate and opponent characters were similar in appearance and action, and that participants knew these characters were actually being controlled by a computer. Moreover, the characters did not display any personality traits during the game; rather, they were faceless, voiceless entities whose only actions were to combat each other. Thus, the game contained minimal social cues, which makes this paradigm an especially strong test of our hypothesis. Some previous research has shown that people behave more prosocially toward ingroup than outgroup virtual characters 11 ; however, no studies to date have examined attributions of specific emotions to characters. This is important, because it suggests that video game players can develop actual conceptions that characters are human-like, beyond any behaviors the players may exhibit on the surface (e.g., trying to defeat or help a character).
These results have implications for future designs of virtual reality and interactive media. Many video games are designed to educate children by teaching lessons in entertaining contexts, where capturing the child's attention is key to helping them engage with the material. Assuming the characters are regarded simply as machines would lead to less effective alterations of behavior in these contexts. Instead, designers would do well to encourage players (through social cues) to relate to or dislike the characters themselves. They could cue an ingroup mindset in order to make players want to follow a character's example, and cue an outgroup mindset in order to make players want to oppose or contradict a character. By capitalizing on the present findings, which demonstrate that ingroup/outgroup distinctions can be developed for computer-controlled characters, designers can make their games more immersive and convincing to otherwise apathetic players.
Footnotes
Author Disclosure Statement
No competing financial interests exist.