Do Black NPCs Matter? Effects of Virtual Contact, Construed as Intergroup Contact with Non-playable Characters in Video Games

(1)

Warsaw University

Faculty of Psychology

Do Black NPCs Matter? Effects of Virtual Contact, Construed as Intergroup Contact with Non-playable Characters in Video Games

Konsekwencje wirtualnego kontaktu, czyli kontaktu międzygrupowego z postaciami kontrolowanymi przez komputer w grach video

Agnieszka Mulak

PhD thesis written under the supervision of prof. dr hab. Janusz Grzelak and dr Mikołaj Winiewski

Warsaw, 2022

(2)

2

Acknowledgements

I’d like to thank my supervisor, prof. Janusz Grzelak, for inspiring me to do research and for his incredibly insightful remarks about experimental methods and conclusions drawn from them.

I’m also grateful to my co-supervisor dr Mikołaj Winiewski for his support, understanding and guidance throughout the thick and thin of the entire process of researching and writing up this thesis.

My thanks also go to my friend and neighbour, dr Wojciech Kotlarski for tediously reading through the entire dissertation, restraining the urge to make fun of psychologists and providing me with very helpful remarks as to the shape and readability of the thesis.

I’m also thankful to my friends Hubert Sobecki, Oktawiusz Chrzanowski and dr Kinga Padzik for their ongoing support and wise words.

This dissertation was written as a result of the Polish National Science Centre research grant

“Attitude Change as a Consequence of Outgroup Contact in Virtual Realities”

No.2016/23/N/HS601590.

(3)

3

Abstract

The research presented in this thesis investigates the effects of virtual intergroup contact (contact with NPCs in video games) on players’ behaviour and attitudes. The first study tests the relationship between players’ social distance towards minorities and intergroup contact present in the games they played the previous month. Four subsequent studies investigate the effects of contact with Black Non- Playable Characters (NPCs) in various in-game conditions on players’ behaviour and explicit attitudes.

In the first study, the quality of intergroup contact proved the strongest predictor of players’ attitudes, with more positive contact predicting more positive attitudes towards real-world minorities. A similar effect was observed for the amount of contact, with a more diverse in-game world predicting more positive attitudes. These effects were observed across all game genres. Thus, the study demonstrated that the effects of virtual intergroup contact generalise to real-world minorities including those, whose representatives were absent in the game (secondary transfer effect). The relationship between attitudes and contact quality was much stronger for contact with NPCs representing real-world groups than for contact with fictional races (such as elves or alien species). The diversity generated by characters representing fictional races was not related to players’ attitudes.

The results of the subsequent experiments demonstrate that the outcomes of in-game intergroup contact with Black NPCs generalise onto behaviour towards other Black NPCs (experiments 2-4) and onto attitudes towards Black people, that is the real-world out-group that was represented by the NPC (the primary transfer effect, experiments 1 and 4). Attitude generalisation to the real-world group occurred only for players who were engaged in the game.

Thus, the experimental studies confirm that even brief virtual contact in a very simple game can improve players’ attitudes toward out-groups. They also demonstrate that several situational factors known to contribute to the outcomes of intergroup contact conducted in the real world similarly determine the effects of virtual contact. These included: intergroup cooperation, shared goals, establishing a common identity with out-group NPCs, intergroup fighting and norms presented in the world of a game (racial segregation or lack thereof).

(4)

4

Interaction between the level of fun players experienced and the experimental conditions had an impact on players’ attitude change. Additionally, combinations of positive and negative virtual contact yielded either positive effects or a lack of observable change as compared to the control condition (no intergroup contact). These results demonstrate that positive affect evoked by playing a game contributed to more positive results of in-game intergroup interactions.

Keywords

Intergroup contact, virtual contact, contact with NPC, video games, prejudice, social distance, attitudes, attitudes toward minorities, cooperation, intergroup relations

(5)

5

Streszczenie

Celem niniejszej pracy jest zbadanie wpływu wirtualnego kontaktu międzygrupowego – czyli interakcji z postaciami w grach komputerowych – na postawy osób grających. W badaniu korelacyjnym testowano związek między kontaktem wirtualnym w popularnych grach obecnych na rynku a postawami grających wobec realnych mniejszości społecznych. W czterech kolejnych badaniach eksperymentalnych analizowano wpływ kontaktu z Czarnymi postaciami sterowanymi przez komputer (NPC) na zachowanie i postawy graczy wobec osób Czarnych (realnej grupy).

W pierwszym badaniu jakość kontaktu wirtualnego była najsilniejszym predyktorem postaw – bardziej pozytywny kontakt w grze przewidywał bardziej pozytywne postawy grających wobec rzeczywistych mniejszości. Podobny efekt zaobserwowano dla różnorodności świata społecznego gry, tj. większa różnorodność skutkowała bardziej pozytywnymi postawami. Efekty zaobserwowano badając wszystkie typy gier. Badanie potwierdziło więc, że efekty międzygrupowego kontaktu wirtualnego generalizują się nie tylko na postawy wobec realnych grup, które były reprezentowane przez NPC w grze, lecz także na inne mniejszości. Związek między postawami a jakością kontaktu był znacznie słabszy w przypadku ras fikcyjnych (np. elfów). Różnorodność generowana przez postacie ras fikcyjnych nie miała związku z postawami wobec realnych mniejszości.

Wyniki badań eksperymentalnych wskazują, że efekty kontaktu międzygrupowego z Czarnymi NPC generalizują się zarówno na zachowanie graczy wobec innych Czarnych NPC (eksperymenty 2-4), jak i na postawy wobec Czarnych w realnym świecie (eksperyment 1 i 4). Generalizacja na postawy wobec grupy w realnym świecie występowała tylko u graczy, którzy byli zaangażowani w grę.

Badania eksperymentalne potwierdziły, że nawet stosunkowo krótki wirtualny kontakt międzygrupowy w prostej grze może skutkować bardziej pozytywnymi postawami graczy wobec grup społecznych. Pokazały również, że czynniki sytuacyjne, które wpływają na efekty kontaktu międzygrupowego w realnym świecie mają znaczenie także w przypadku kontaktu wirtualnego.

Czynniki te to: kooperacja, walka i rywalizacja, wspólne i sprzeczne cele oraz normy społeczne przedstawione w świecie gry (segregacja, lub powszechność kontaktu międzygrupowego).

(6)

6

Zaobserwowano, że interakcja między poziomem przyjemności z gry i kontaktem międzygrupowym ma wpływ na postawy grających. Ponadto, mieszanka czynników pozytywnych i negatywnych w grze skutkowała efektami pozytywnymi lub brakiem efektów. Wyniki te pokazują, że pozytywny afekt wywoływany przez grę przyczynia się do bardziej pozytywnych skutków kontaktu wirtualnego.

Słowa kluczowe

Kontakt międzygrupowy, kontakt wirtualny, kontakt z NPC, gry wideo, uprzedzenia, dystans społeczny, postawy, postawy wobec mniejszości, kooperacja, relacje międzygrupowe

(7)

7

1. Theoretical Background

This thesis investigates the effects of virtual intergroup contact (that is intergroup contact with characters in video games) on players’ real-world behaviour and attitudes. As research on the topic is extremely scarce, the hypothesis draws on broader findings on video games and on intergroup contact in general. Thus, the first section of the theoretical chapter summarises research on the relationships between video games and players’ social attitudes. The section concludes with the implications these findings may have for the topic of the thesis, providing the rationale for the research hypotheses. The second section summarises the vast literature on intergroup contact with a focus on the most researched effects. The section includes findings on indirect contact, that is intergroup contact that is not conducted face to face. The last section draws on both previous ones to present the research hypotheses.

Video games research

It is estimated that in 2021 there were close to three billion players of video games across the globe and the numbers are still rising (Newzoo, 2021). The huge popularity of video games has inspired researchers to examine their impact on players. It was recognised that depictions of specific concepts in games can make it easier for players to access these concepts in real-world interactions (priming), leading to short-term changes in attitudes and behaviour. Repeated play reinforces these effects, leading to permanent change (Buckley & Anderson, 2006). The most prominent trend in games research considers negative and antisocial effects, specifically the relationship between in-game violence and players’ aggression (e.g. Sherry, 2001, Anderson et al., 2010, Prescott et al., 2018), and aggression priming (Anderson & Bushman, 2002).

The negative impact of games on attitudes towards others

Especially relevant for this thesis is research demonstrating that computer games can influence attitudes towards specific social groups. In this regard, the most criticised aspects are, again, violence and stereotypes due to the ubiquity of both in titles present on the market (Dickerman et al., 2008;

(9)

9

Sisler, 2008). Researchers (Beasley & Standley, 2002; Burgess et al, 2011) have documented the prevalence of stereotypes in games and demonstrated that the frequency of playing video games was related to more negative attitudes outside of the game context, e.g. to attitudes toward Blacks and Asians (Behm-Morawitz & Ta, 2014) or to stronger sexism (Breuer et al., 2015, Stermer & Burkley, 2015).

Experimental studies have shown stereotypical characters in violent video games prime stereotypical associations (Burgess et al., 2011; Cicchirillo, 2015). Playing a violent video game as a Black avatar results in more negative attitudes toward Black people as compared to playing the game as a White avatar (Yang et al., 2014). Noteworthy, almost all of the research on the negative effects of stereotypic characters was conducted on violent video games. However, Saleem and Anderson (2013) demonstrated that playing a violent game, which featured stereotypical terrorist Arabs in the Middle Eastern setting, increased stereotyping and negative attitudes toward Arabs and they did observe similar but weaker effects if the game contained no violence or if it featured no Arabs and non- stereotypical setting, but did focus on the trope of terrorism. Thus, even though experimental research on the subject is sparse, exposure to violent stereotypical content in video games has been consistently linked to negative outcomes.

The positive impact of games on attitudes

There is also a growing pool of evidence for a positive impact of video games. A meta-analysis has shown that whereas violent video games reduce prosocial behaviour, games with prosocial gameplay increase it (Greitemeier & Mügge, 2014). Some of the studies investigated games designed specifically to elicit positive change and research has shown such games can successfully improve attitudes towards a designated target (Mavridis et al., 2017). In the context of social attitudes, Roussos &

Dovidio (2016) have shown that a simple online game designed specifically to elicit an understanding of poverty increased positive attitudes, empathic concern, and support for government-funded anti- poverty policies in some experimental groups. Other researchers have found more general effects, especially for prosocial games (i.e. games designed so that players’ in-game behaviour is prosocial, e.g. they help others). Playing a prosocial video game was shown to increase access to prosocial

(10)

10

thoughts (Greitemeyer & Osswald, 2011) and behaviour (Greitemeyer et al., 2012), with a very simple and popular game “Lemmings” as the prosocial experimental manipulation. Similarly, Gentile and colleagues (2009) have shown that exposure to prosocial games was related to lower approval of aggression and smaller hostile attribution bias and that playing a prosocial, nonviolent video game (parts of Super Mario Sunshine or Chibi Robo) directly results in more prosocial behaviour.

Noteworthy, the experimental research demonstrating the prosocial effects of games was often conducted on games that did not include violence.

Similar effects have been found for cooperation in games. Playing a collaborative game improved out- group attitudes in the case of playing with (whom participants believed to be) a human partner belonging to a different social group than the player (Stiff & Bowen, 2016) and for playing with a computer-controlled character. Moreover, even violent recreational games, when played cooperatively with another person (human partner) have been demonstrated to increase the tendency for future cooperation (Ewoldsen et.al., 2012). Moreover, cooperation in a game has been shown to attenuate the effects of in-game violence (Velez et al., 2016).

Vang and Fox (2014) found that both cooperation and competition with a Black character (player- controlled) increased liking of other Black avatars. Although the researchers examined cooperation, their results suggest experiencing intergroup contact itself in a game can impact players’ attitudes.

However, the effect was observed for players’ attitudes towards Black avatars and did not generalise to attitudes towards Black people in general. Recently, Breves (2020) found evidence that even neutral contact with a Black computer-controlled character (completing a task for a Black NPC), resulted in more positive attitudes towards Black people, although the effect was observed only for a robust simulation of a virtual reality (VR) game and not for the version played on the screen.

Despite the scarcity of research on intergroup contact in games, the effects of such contact have been postulated. Games have been criticised for the lack of diversity (Passmore et al., 2017, Rutberg, 2019) and researchers have proposed that an increased presence of diverse characters would lead to more positive intergroup attitudes among players and prejudice reduction (Amichai-Hamburger &

McKenna, 2006). In other words, interactions with NPCs that belong to various groups were

(11)

11

postulated to have a positive effect on players. Therefore, these claims imply that intergroup contact with NPCs will result in more positive out-group attitudes among players.

The rationale for conceptualising in-game interactions with NPCs as intergroup contact

The prevalence of violence in video games is not likely to change, as the research of over 200 thousand gamers shows that about 14% of male players play primarily for the experience of competition, 12%

for the experience of causing destruction (Yee, 2016), and many more players of both genders enjoy those experiences. However, for most players, the motivations are mixed (Yee, 2020) and competitive gameplay should not be treated as the opposite of collaboration and helping (Yee, 2017) and many games cater to both needs. Thus, the research on the impact on interactions in games should also include a broader framework. One that could account for all observed effects and provide a theoretical model for examining both positive and negative factors contributing to in-game interactions.

Researchers have recognised the value of intergroup contact theory for analysing video games’

influence (e.g. Hasler & Amchai-Hamberger, 2013, Amchai & McKenna, 2006, Kordyaka et al., 2019, Breves, 2020) but research on the topic is still extremely sparse. However, considerable knowledge was accumulated by social psychologists about the mechanisms, mediators, and moderators of intergroup contact conducted in other settings. It provides a framework for analysing negative effects of violent encounters in video games and positive effects of in-game collaboration. It also describes mechanisms and situational conditions important for attitude generalisation from an interaction partner (whether real or imaginary) to the real-world social group as a whole. It informs about factors that are crucial for the effects of intergroup contact and explains the positive effects of neutral intergroup contact.

Intergroup Contact

One of the first, and arguably the most famous formulation of the Intergroup Contact Theory (ICT) was made by Gordon Allport (1954). Allport acknowledged various possible outcomes of contact and listed conditions necessary for positive effects, inspiring a multitude of research – a trend that lasts to date. In the broadest of terms, ICT states that interactions between members of two distinct groups

(12)

12

can change the attitude of each of the participants towards the “other” group (the out-group). Allport believed that for intergroup contact to result in positive outcomes (e.g. reduced prejudice) interactions must be conducted in a situation that fulfils specific requirements. These requirements, termed

“optimal conditions” included: 1) cooperation 2) common, superordinate goals. Furthermore, intergroup contact should be 3) supported by institutional authorities, laws, norms or customs and 4) both groups engaging in contact should have equal status in the situation. Intergroup contact conducted under these optimal conditions was to produce positive outcomes and if the conditions were not met, contact was to result in conflict and negative effects. With these assumptions, intergroup contact theory is congruent with games research – the positive results observed for games that included cooperation while lacking violence and negative impact of gameplay that was violent and thus – in the light of intergroup contact theory – was highly competitive, included conflicting goals with out-groups and featured a setting with norms and laws supporting negative out-group relations.

A vast array of research followed Allport’s formulation of the ICT, confirming the positive impact of intergroup contact (e.g. Pettigrew, 1998; Hewstone & Swart, 2011) and demonstrating the positive effects on multiple dependent variables. Intergroup contact has been found to reduce both subtle and blatant prejudice (Hamberger & Hewstone, 1997). Many studies focused on cognitive outcomes of intergroup contact such as stereotype reduction or a decrease in semantic differences (Tropp &

Pettigrew, 2004). Further effects included more positive emotions towards the out-group in general (Paolini et al., 2006) and other emotional outcomes, such as reduced anxiety, decreased individual and collective threat, enhanced empathy and forgiveness, and many more (Pettigrew et al., 2011).

Notably, intergroup contact has also been shown to affect behaviour, promoting more positive actions towards the out-group members (Dovidio et al., 1997; Nier et al., 2001) and more positive behavioural intentions (Christ et al., 2010; Vezzali et al., 2012).

The most prominent evidence supporting the ICT to date are meta-analyses conducted by Pettigrew and his colleagues (Pettigrew & Tropp, 2006; Pettigrew et.al., 2011), including over five hundred studies from multiple countries each. The meta-analyses confirmed that direct intergroup contact is

(13)

13

strongly linked to prejudice reduction, regardless of actors’ demographic characteristics (such as age or gender), geographical settings, and the type of out-group. Moreover, they demonstrated that intergroup contact itself is enough to produce positive effects, even without the optimal conditions.

This finding explains the results observed by Vang and Fox (2014) with players’ increased liking of Black avatars after cooperative as well as competitive gameplay that included intergroup interactions.

Furthermore, the meta-analyses found that when the situation was structured so that it met the requirements of the optimal conditions, the positive effects were stronger. Additionally, the conditions were highly interrelated, which thus brought the authors to the conclusion they are best to be treated not separately, but as one joint indicator of whether the circumstances in which the contact occurs are favourable for positive change (the first study conducted as a part of this thesis adopted this concept into the design – see chapter 2).

Generalisation of Effects

For intergroup contact to reduce prejudice, the positive effects need to generalise from the intergroup interaction to other situations and from the out-group member who took part in the interaction to the entire out-group. This effect was assumed by Allport (1954) and indeed, most research tested for the generalised effects on attitudes towards the entire out-group, as opposed to the encountered individual. The meta-analytic test also confirmed, that the positive effects of intergroup contact typically generalised to the entire out-group (Pettigrew & Tropp, 2005). Several pathways have been suggested for this generalisation, each highlighting different aspects of the situation in which contact is conducted and of the interaction itself.

One of the simplest explanations used for the effects of intergroup contact involves the exposure effect. Research has shown that repeated exposure to a target, in and of itself, increases liking of the target and other related targets (Bornstein, 1989, Montoya et al., 2017). Such effects have been observed in case of social targets i.e. people and the increased liking of others that share similarities with them (Moreland & Beech, 1992, Burger et al., 2001). The exposure effect justifies why contact itself was found to reduce prejudice and explains the wide applicability of the ICT (Pettigrew et.al., 2011). In the context of video games, the exposure effect requires only the presence of out-group

(14)

14

characters, i.e. including a simple graphic depicting them or a mention, without the necessity for robust interactions with such characters or a story behind them. This mechanism could explain positive results observed for prosocial games with very simple design and graphics such as Lemmings (Greitemeyer et al., 2012) or Chibi Roboto (Gentile et al., 2009).

Aside from mere exposure, multiple mediating mechanisms have been found to promote the generalisation of the outcomes of contact. One of the first perspectives (Sherif et al., 1961) emphasises the role of functional relations, stating that positive interdependence (cooperation) produces more positive attitudes towards out-groups and negative interdependence (rivalry) produces more negative attitudes. Thus, the concept further stresses the role of cooperation identified by Allport as one of the key factors for positive outcomes. This approach has found support in subsequent research, linking cooperation and rivalry to changes in intergroup attitudes (Brewer & Miller, 1984) and intergroup attraction (Gaertener et al., 1999). Other researchers focused on behavioural factors (e.g.

positive behaviour towards out-group members resulting in cognitive dissonance and the process of its reduction, Miller & Brewer, 1986; Stephan, Finlay, 1999). In line with these findings, game mechanics that result in positive exchanges with out-group characters, especially cooperation, would lead to more positive out-group attitudes. This was demonstrated in games research for attitudes towards others in general (e.g. Stiff & Bowen, 2016). Similarly, according to intergroup contact theory game design that induces fighting with out-group characters would increase stereotyping and prejudice. This prediction is congruent with the effects observed for playing violent games that featured stereotypical in-game characters (Burgess et al., 2011; Cicchirillo, 2015).

Further, the literature has called attention to cognitive factors, e.g. learning new information about the out-group, building new, non-stereotypic associations or reducing uncertainty in future intergroup interactions (Dhont et al., 2011). Subsequent research has also pointed to a critical role of emotions, that is positive emotions in general (Paolini et al., 2006) and various specific emotions – such as collective guilt (Branscombe & Doosje, 2004), feelings of intimacy and closeness or empathy – in producing positive outcomes by reducing negative affect and increasing positive affective ties (Tropp & Pettigrew, 2004, for a more detailed review of mediating factors see Dovidio et al., 2003).

(15)

15

The bulk of these findings poses considerable demands on video games. Whereas generally positive emotions could be evoked by the situation of playing any game, all other factors described above require out-group interactions in the game to be more robust. For example, for players to learn new information about the out-group the game would have to include such information in the characters’

storyline and convey it in dialogues and plotlines designed for those characters. Similarly, to produce feelings of closeness or empathy, the game would have to include a narrative that allows players to form a bond with those characters. Thus, only larger games in which character interactions were more robust would fit the criteria posed by these factors.

Some of the most prominent recent theories are based on the Social Identity Theory (Tajfel & Turner, 1979) which, in broadest terms, addresses the role of individual and collective identities with the latter based on an individual being a part of social groups (categories). Brewer and Miller (1984) proposed decategorization is necessary for intergroup contact to produce positive outcomes. That is, to result in a positive change, interactions need to be personal as opposed to category-based, and out-group members have to be differentiated from the out-group as a whole. Such contact would promote the exchange of individuating and non-stereotypical information and achieve more differentiation in the way a person perceives out-group members. Thus, the decategorization model assumes intergroup contact should remove the boundaries of social categories. Again, this pathway requires video games to be more robust, allowing players to learn more about the out-group character and to see them as an individual and not just as a part of a group or as a background for the game mechanics. A different approach was proposed in the in-group identity model (Gaertner & Dovidio, 2000). According to the authors, intergroup contact should be structured to allow for one, common identity with the out- group. Thus, the model implies the importance of common goals. In effect, social boundaries should be changed (recategorization) through contact so that the out-group is perceived as part of the same category, an ingroup, promoting more positive attitudes. For games, this pathway necessitates the presence of teams or sides and the inclusion of out-group characters on the players’ side (which is the case in some games on the market and the design of some experimental conditions in the studies presented in this thesis).

(16)

16

Hewstone and Brown (2005) noted that whereas both these models have gathered supportive experimental evidence, they include serious inherent problems. Firstly, they require participants of intergroup interactions to relinquish their own group identities up to some degree for the sake of individuation (both in the decategorisation model and, even more so, in the recategorization model which requires creating new categories). Whereas this is possible in experimental settings and with artificial categories, in real life people may be unable or unwilling to forsake their social identities.

Moreover, for the recategorisation model, if people feel the new categories are forced on them against their will, it could cause negative outcomes. Games’ research indirectly supports this critique by demonstrating negative effects for players who are forced to adopt avatars that differ from them markedly (Passmore, Birk, & Mandryk, 2018). Furthermore, Hewstone and Brown (1986) proposed that if the effects of intergroup contact are to generalise onto all other out-group members, then the initial group identities have to remain present throughout the contact. In other words, for the contact to be “intergroup contact”, group categories have to be salient throughout the interaction. The authors proposed this will lead to interaction participants recognising both groups’ respective superiorities and inferiorities. The model resulted in a focus on group salience, with a multitude of studies supporting its crucial role in intergroup contact (Brown & Hewstone, 2005). Therefore, according to this model, games including out-group characters should depict them unequivocally as part of the out-group.

Most recently, group salience has been reconceptualised with a focus on semantic distance, that is the degree of similarity or relatedness between the target (e.g. the interaction partner) and a frame (e.g.

group prototype). The model stems from prototype models of category structure (Rosh, 1978), stating that the smaller semantic difference will result in more generalisation whereas large semantic distance will result in seeing the target as non-representative for the group and to a creation of a

‘subtype’, leaving the attitude towards the entire out-group unchanged (Meleady et al., 2019).

Therefore, for greater effects, out-group characters in video games would have to be constructed so that they at least partially resemble the prototypical out-group. Additionally, fantasy races would have a lesser impact on out-group attitudes than real-world races depicted in the game (as demonstrated in the following experiments in this thesis). This could also explain why some games’ experiments

(17)

17

observed a change in attitudes towards other avatars but not towards real-life social groups depicted by those avatars (Vang & Fox, 2014).

Transfer effects

The positive effects of intergroup contact on intergroup relations require that interaction outcomes generalise from the interaction partner to the entire out-group. This effect is called the primary transfer effect (Pettigrew & Tropp, 2011). Research has also shown, that the processes of generalisation go further, with positive outcomes generalising to other, secondary out-groups, whose members were not part of the intergroup interactions (the secondary transfer effects, STE).

Furthermore, contact generalises beyond intergroup relations and extends to cognitive processes, resulting in cognitive flexibility (the tertiary transfer effects). Researchers documented the tertiary transfer effect, showing intergroup contact results in better academic achievement, skills and engagement as well as more flexible thinking (Meleady et al., 2019). Due to the goal of studies presented in this thesis, I will focus on the secondary transfer effects.

Numerous studies have demonstrated the secondary transfer effects (e.g. Lemmer & Wagner, 2015;

Pettigrew, 2009). For example, contact with immigrants was shown not only to relate to more positive attitudes toward immigrants in general, but also toward lesbians, gay men and Jews (Schmid et al., 2012). Living with a Latino roommate resulted in more favourable attitudes not only towards Latinos but African Americans as well (Van Laar et al., 2005). In a series of experiments, Tauch et al. (2010) confirmed the secondary transfer and ruled out alternative explanations for the effects, including:

contact with the secondary group, prior attitudes and socially desirable responding – thus demonstrating the impact of the attitude generalisation. Researchers observed secondary transfer effects not only for face-to-face contact but for indirect contact as well, showing STE as a result of extended (Vezzali et al., 2019) vicarious (Joyce & Harwood, 2014) and imagined (de Carvalho-Freitas

& Stathi, 2017) contact (see the “Indirect contact” section). This suggests that depictions of out-groups in video games might influence players’ attitudes towards numerous groups not present in the game.

Literature shows that whereas some of the mechanisms are specific to secondary transfer effects, STE share many of the mediators and moderators with primary transfer effects (Boin et al., 2021). Vezzali

(18)

18

and her colleagues (2020) argue that STE and prejudice are inherently intertwined as prejudice includes both, a component specific to a single out-group and a general component.

Most of the evidence for the secondary transfer effects stems from research on positive contact. The scarce findings on negative contact suggest that STE does occur and is a result of the same mediating processes as those found for positive contact, although not all the studies found significant effects of secondary attitude generalisation (Vezzali et al., 2021).

Negative intergroup contact

Intergroup contact theory initially aimed to specify conditions for positive change and since then, research has focused mainly on positive outcomes. Video games research stands in contrast to this approach, focusing primarily on negative consequences. Analyses of intergroup contact in games has focused on the correlation between gameplay and stereotype strength (Behm-Morawitz & Ta, 2014;

Stermer & Burkley, 2015), demonstrating the ubiquity of stereotypical NPCs (Dickerman et al., 2008) and its consequences: activating stereotypes (Burgess et al., 2011), inducing negative affect and stereotyping (Behm-Morawitz et al., 2016), and strengthening of negative attitudes (Saleem &

Anderson, 2013).

Intergroup contact research has also shown an association between negative contact and increased prejudice (Aberson & Gaffney, 2008; Barlow et al., 2012). Researchers define “negative intergroup contact” in various ways. Contact valence (its positivity or negativity) was often defined in general terms that is whether an interaction was pleasant or unpleasant (Paolini et al., 2010) or perceived as good/positive versus bad/negative (Barlow et al., 2012, Techakesari, 2015). Other studies of negative contact pointed to, among others, the prominent roles of inter-group anxiety (Pettigrew & Tropp, 2006), intergroup threat (Abrams & Eller, 2017) or conflict and rivalry (Brewer & Miller, 1984).

Games can evoke negative emotions (such as fear) and interactions in games often are negative (e.g.

players compete with characters in the game, are threatened by them and can be killed), which may suggest the ubiquity of negative contact in games. However, the overall situation of playing a game is a positive and pleasant one, which could imply that intergroup contact in games is interpreted more positively than similar contact in the real world. Notably for the stereotype prevalence in games,

(19)

19

contact with stereotype-confirming out-group members was also shown to strengthen stereotypical beliefs and cognitions, and result in more negative attitudes (Alvídrez et al., 2015).

Moreover, some researchers postulate negative contact may have a stronger impact on intergroup attitudes than positive contact (Barlow et al., 2012; Graf et al., 2014; Paolini et al., 2010), especially when it comes to the cognitive aspects of prejudice (Aberson, 2015). Acknowledging that negativity has stronger effects than positivity in other areas of psychological research, Graf and Paolini (2017) describe a valence asymmetry for intergroup contact – stronger negative effects of negative contact than positive effects of positive contact – demonstrating multiple evidence for its existence. Building on these differences between the two types of contact, Árnadóttir and her colleagues (2018) proposed that the effects of positive and negative contact are not additive. Instead, they are to interact with each other. Indeed, the team has demonstrated, that prior positive contact has a buffering effect, diminishing the effects of negative contact. Their research has also shown that prior negative contact has a facilitating effect, resulting in stronger positive effects of positive contact.

The valence asymmetry could have particular importance in the context of video games. Whereas in the real world, positive intergroup interactions were found to be much more frequent than negative ones (Graf et al., 2014, Pettigrew, 2008, Dhont and Van Hiel 2009), in virtual worlds violent interactions are common (Dill et al., 2005). Therefore, negative valence asymmetry could result in strong negative effects of intergroup contact in games. On the other hand, the ubiquity of negative interactions in games may result in a strong facilitating effect of positive contact in games. However, it is important to note that whereas the negative valence asymmetry was confirmed in indirect and imagined (Paolini et al., 2014) contact, the interaction effects (the buffering and facilitation effects) have not yet been observed for those types of contact (Árnadóttir, 2018).

Indirect contact

The initial intergroup contact hypothesis stated that the physical presence of two groups is necessary for an attitude towards a group to change. Subsequent research has expanded the theory to various other situations. The important role of one-on-one intergroup contact has been recognised (Dovidio et al., 2011), especially individual intergroup friendships (Hewstone & Swart, 2011). Pertinent to the

(20)

20

research presented in this thesis, even a one-time interaction with a single out-group member (e.g.

watching television together) was found to affect attitudes toward that group (Tal-Or & Tsfati, 2016).

This, in turn, implies that the time of a single video game playthrough might be enough for the effects of intergroup contact to occur.

The effects of intergroup contact have also been demonstrated for interactions other than face-to-face.

Wright et al. (1997) proposed people do not need to personally encounter out-group members for intergroup relationships to improve. The knowledge of an ingroup member’s friendship with a representative of an out-group (extended contact) should result in positive effects similar to those observed for direct contact. Since then, knowledge of cross-group friendships has been consistently linked to less prejudice (Paolini et al., 2004; Pettigrew et al., 2007; Wright et al., 2008). It has been demonstrated that such knowledge can reduce prejudice by reducing intergroup anxiety, by generating perceptions that norms are positive for contact and through the inclusion of out-group in the self (Turner et al., 2008) as well as through other cognitive and affective mediators (Vezzali &

Stathi, 2017). Positive effects of extended contact have been observed even while controlling for direct contact (Turner et al., 2007). The effects of extended contact are not limited to knowledge of one’s friend having an out-group friend, but also occur if it’s a previously unknown ingroup that has a cross- group friend (Tausch et al., 2011).

One of the recognised forms of extended contact, consistently shown to produce effects similar to those of direct interactions (Di Bernardo et al., 2017), is vicarious contact (Dovidio et al. 2011) i.e., observing an ingroup member interacting with an out-group representative. Research on vicarious contact further extends the theory to cross-group contacts that are observed in the media (Vezzali et al. 2014, Weisbuch, Pauker & Ambady 2009), read about or heard about on the radio (Vezzali & Stathi, 2020). This would imply that contact in video games could have similar or stronger effects than other media as games allow one not only to observe interactions but also to take part in them.

Furthermore, Turner and his colleagues (2007, Crisp & Turner, 2009, 2012) have demonstrated that simply imagining interactions with an out-group member results in effects observed for other forms of contact. Subsequent research provided evidence that imagined contact significantly improves

(21)

21

attitudes toward out-groups (Turner & Crisp, 2010), enhances intentions for future contact (Crisp &

Turner, 2009), reduces inter-group bias in attitudes, improves inter-group emotions, intentions and behaviour (Miles & Crisp, 2013), positively impacts attitudes at both explicit and implicit levels and is effective across a broad range of target groups (Meleady & Crisp, 2017).

Intergroup contact theory has been further expanded to computer-mediated interactions (online contact). Research shows that intergroup contact conducted on an online platform (Amichai- Hamburger, 2008), on Facebook (Schumann et al., 2012), and in chatrooms (White et al., 2015) weakens stereotypes and results in more positive attitudes toward the out-group. Lemmer and Wagner (2015) conducted a meta-analysis of prejudice-reducing interventions that were conducted either online or in the real world. Their results have shown that contact via the internet does decrease negative attitudes and that this change persists over time. More importantly for this thesis, their results have demonstrated that interventions conducted online did not differ from real-world, face- to-face interventions with regard to the strength of outcomes.

Hasler and Amchai-Hamberger (2013) argued that the internet might be effective for putting intergroup contact into practice due to the ease with which conditions for positive contact can be achieved and measures for its effects obtained. Online interactions are often more structured, which may reduce anxiety during intergroup contact. Moreover, the element of fun is often present in online activities, which was postulated to make computer-mediated interactions especially suitable to generate positive contact, leading to better intergroup relations (Amchai & Hayat, 2013, Amichai- Hamburger et al., 2015). Such conclusions could be extended to video games. Whereas in-game interactions are complex and difficult to design, they are almost always structured and games aim to be fun, which could contribute to more positive contact.

Video games have also been proposed as having the potential to influence intergroup attitudes, especially through players’ contact with in-game diversity (Amichai-Hamburger & McKenna, 2006), and more specifically, through intergroup contact with NPCs (Kordyaka, et al., 2019). Recently, Breves (2020) demonstrated that contact with an out-group NPCs in virtual reality influences

(22)

22

attitudes towards Black people using one of the bestselling titles on the market and demonstrating the effect for a robust, first-person VR version of the game.

Research hypotheses: Virtual Contact

This thesis proposes that virtual contact (contact with computer-controlled characters in games) can be construed as intergroup contact. As described previously, online interactions with people have an impact on attitudes similar to that of interactions conducted face-to-face. Research on vicarious and imagined contact demonstrates that actual human interaction is not necessary for intergroup contact effects to occur. Therefore it stands to reason that virtual contact proposed here, i.e. contact with NPCs in games, could follow the same pattern as other forms of intergroup contact.

This view is congruent with video games research and could provide a new perspective on some of the observed effects. As has been shown earlier, playing video games is known to change both implicit and explicit attitudes toward real-world groups (Yang et al., 2014), proving interactions conducted with characters presented in video games are in fact capable of changing attitudes toward a real-world out-group.

The negative impact of video games on attitudes was most commonly assigned to exposure to violence and stereotypes in games (Burgess et al., 2011). Conditions under which contact strengthens negative attitudes, as described by the intergroup contact research, include negative interactions (Techakesari et al., 2015) conducted under conditions of inter-group anxiety (Pettigrew & Tropp, 2006) and with stereotype-confirming out-group members (Alvídrez et al., 2015). Therefore, in the light of the findings of intergroup contact literature the most frequently researched elements of video games are specifically the ones that promote negative outcomes.

Games also have the propensity to result in a positive change in attitudes (e.g. Greitemeyer et al., 2012), although most of the research on the topic tested games in which violence was absent and the game elicited cooperation. Such conditions have also been described by intergroup contact researchers as highly conducive to positive change. However, due to the prevalence of violence in games, such conditions occur in very few titles. Yet intergroup contact literature could provide a new

(23)

23

perspective on the topic. Whereas games research focuses on the presence of violence or lack thereof, contact research demonstrates positive effects can occur even when conditions are not optimal.

Almost all games – whether violent or not – are designed to create feelings of fun and in-game interactions are always structured according to game mechanics. These aspects have been described as elements of positive contact and could contribute to more favourable intergroup attitudes.

Therefore further research is required to examine in-game conditions where both positive and negative interactions are present, as is common in video games, and intergroup contact literature provides guidelines for such inquiries.

The vast research on intergroup contact could serve as an excellent theoretical framework for assessing the conditions and factors that could contribute to positive and negative outcomes of players’ contact with NPCs. It could also provide ample mediators and moderators for generalisation effects from NPCs to real-world out-groups and inform guidelines on including diverse characters into games’ social worlds. Whereas some recognised pathways for generalising the effects of contact from the interaction partner to the entire out-group would require the in-game interaction to be more robust (e.g. result in learning facts about the out-group, elicit understanding and specific feelings such as guilt), others do not (e.g. exposure effect). This could imply that even contact in very simple games may elicit change in players’ real-world attitudes.

Researchers are beginning to recognise the potential of intergroup contact mechanisms for analysing games (Hasler & Amchai-Hamberger, 2013; Kordyaka, et al., 2019; Breves, 2020) although the studies on the topic are still very sparse. The effects of intergroup contact for real-world attitudes were previously shown only for a robust virtual reality simulation. Moreover, intergroup contact with NPCs was not previously tested in games in which both cooperation and fighting with NPCs were present as is often the case with games on the market. The studies presented here aim to close this research gap.

Research hypotheses in the correlational study

Firstly, research in this thesis examines whether the basic assumptions of intergroup contact theory apply to virtual contact in the most popular games on the market. Thus, study 1 tests the following hypothesis:

(24)

24

H.1. Intergroup contact in video games present on the market is linked to players’ attitudes towards real-world out-groups (secondary transfer effect)

H.1.a. More positive contact is linked to more positive attitudes towards out-groups

Research hypotheses in the experimental studies

The first study is followed by four experiments, which investigate the effects of virtual contact (contact with NPCs) conducted in a very simple game. The effects are examined in several experimental conditions, which differ as to the amount of intergroup fighting and cooperation, the presence of common or conflicting goals and varying norms for contact. Therefore, studies 2 - 5 test the following hypotheses:

H.2. Intergroup contact with NPCs in a very simple video game impacts players’ attitudes towards the real-world out-group the NPC represented.

H.2.a. Positive intergroup contact with NPC in a simple game results in more positive players’

attitudes towards the real-world out-group.

H.2.b. The positive attitude change occurs even if the game features fighting

H.2.c. Negative intergroup contact with NPC in a game results in more negative attitudes

Research aim

Thus, the studies reported here aim to expand intergroup contact theory to virtual contact (interactions with computer-controlled characters) and to demonstrate that virtual contact influences players’ attitudes even when it is conducted in very simple video games.

(25)

25

2. Empirical Evidence: Studies

Video games research provides evidence that gameplay can impact players’ attitudes (e.g. Behm- Morawitz & Ta, 2014; Sisler, 2008). The bulk of research focused on detrimental effects of exposure to stereotypical characters in violent games on intergroup attitudes (e.g. Beasley & Standley, 2002;

Burgess et al, 2011). However, studies have also demonstrated the positive impact of prosocial games on attitudes towards others (e.g. Greitemeier & Mügge, 2014; Greitemeyer & Osswald, 2011).

Simultaneously, researchers postulated more diverse virtual worlds are required for players attitude to improve (Amichai-Hamburger & McKenna, 2006; Passmore et al., 2017; Rutberg, 2019).

Intergroup contact research could provide the theoretical framework for reconciling these results and notions. It also allows for precise predictions about the outcomes of intergroup contact in games and for posing more detailed hypotheses about the effects of games where violence is present but is not the only factor affecting the contact situation, as is often the case in popular titles on the market.

Current literature has recognised the potential of intergroup contact theory for analysing intergroup interactions in video games (e.g. Amchai & McKenna, 2006; Hasler & Amchai-Hamberger, 2013;

Kordyaka et al., 2019). However, there are very few studies examining the impact of such interactions on players’ intergroup attitudes. Researchers examined such contact in robust video games with no violence and the results are mixed (Breves, 2020; Vang & Fox, 2014).

As the predictions of intergroup contact literature have not yet been tested in games where both fighting and positive interactions are present, as is often the case with games on the market, the first study presented in this thesis was conducted on a vast array of popular games and their players. It tests the relationship between players’ intergroup attitudes and intergroup contact in the games they played.

The effects of intergroup contact have been demonstrated for online communication with people and for interactions that are but imagined. Literature provides some evidence for the effects of intergroup contact with computer-controlled characters in a robust VR game (Breves, 2020). However some of the pathways described by intergroup contact researchers could also work for simple games (e.g.

(26)

26

Brewer & Miller, 1984; Paolini et al., 2006; Pettigrew et.al., 2011). Thus, the thesis presents four experiments conducted with a very simple game. These experiments tested the effects of positive, negative and mixed virtual contact (intergroup contact with NPCs) on players’ attitudes towards the out-group the NPCs represented.

Study 1: Correlational evidence for intergroup contact in video games on the market (secondary transfer effect)

The first study¹ was designed to examine the general applicability of intergroup contact research to interactions in video games. More specifically, the study tested the hypothesis that the average valence (level of positivity-negativity) of intergroup contact in popular video games is related to players’

attitudes towards out-groups.

The study included an analysis of multiple video games popular on the market (with an evaluation of intergroup contact in these games) and an assessment of generalised out-group attitudes among people who play them. Thus, the study was conducted as an online survey of gamers, followed by an evaluation of games they played conducted by independent judges. The questionnaire presented to gamers comprised questions about games they played the most in the past month and of measures of prejudice against minorities. Independent judges assessed the games most popular among the participants on aspects pertaining to intergroup contact. They listed all minorities present in those games and evaluated the quality of in-game contact with each minority.

1 The results of this study, along with conclusions and following arguments have been published in Cyberpsychology: Journal of Psychosocial Research on Cyberspace (Mulak, Winiewski, 2021). This chapter draws heavily on the publication.

(27)

27

Participants

Participants were volunteers, recruited with posts on three largest Polish Facebook groups for gamers.

No rewards, financial or otherwise, were offered for participation. The initial sample included N = 2324 Polish gamers. Participants who quit during the study and did not provide answers to all the scales were excluded from the analysis, leaving N = 1857 who were taken into account. The procedure of games’ evaluation further restricted the sample. Participants who indicated 3 of 3 games that were rare and thus not evaluated by competent judges were not included in the analysis. The final sample consisted of N = 1627 gamers.

In the analysed sample, players’ age ranged from 11 to 52 years old, with an average age of 19 years (SD = 6.53). Respondents were mostly male (64.7%). Females comprised 13% of the sample. The rest of the respondents self-identified as “other” (0.4%) or did not state their gender (21.8%). Although large, this gender disparity may be a reflection of Polish gamers’ population. Whereas women constitute roughly half of the players in Poland, they are, on average more of casual gamers in contrast to PC and console players who are mostly male (Polish Gamers Observatory, 2018). Female players did not differ significantly from male participants in average gameplay time per week (t(1) = 0.031, p

= .86) but were slightly better educated (t(1) = 16.43, p < .001, η² = .007), older (t(1) = 20.94, p < .001, η² = .008) and lived in larger cities (t(1) = 13.88, p < .001, η² = .007). On average, they were also less prejudiced (t(1) = 103.43, p < .001, η² = .036 for the generalized social distance towards minorities).

Poland is a racially homogenous country. According to the last national census, 97.1% of the residents are of Polish ethnicity. The second and third largest ethnic groups in the country are also White and Polish (the Kashubians and the Silesians, 1.1% and .05% respectively; Statistics Poland, 2011).

Therefore, there was a very small likelihood that participants, who lived in Poland and spoke fluent Polish were of other ethnicities than the vast majority. Thus, following the assumptions made by the Polish Prejudice Survey (Winiewski, 2017), the current study was analysed under the premise that respondents were of Polish ethnicity.

(28)

28

Measures

Two kinds of measures were used in the study. The first encompassed individual-level measures that pertained to each participant’s experiences with games and their attitudes. These were obtained from the questionnaire completed by gamers. The second kind included game-level measures which described characteristics of games and were provided by independent judges during their evaluation.

Individual level variables

Games. Participants were asked to name at least one and up to three games they played the most in the previous month. The games were stated in order of the magnitude of time committed to the game (i.e. the most played game, the second most played game and the third most played). All the answers taken together provided for 1953 game titles played often by at least one participant. The titles were then grouped so that games that were part of a series were listed under one title of the series which resulted in 802 titles (including 362 titles mentioned by participants in the first spot). Most of the games were mentioned by several participants (M = 24.56, SD = 6.35). The most popular game was played by 587 people (including 307 people who placed it as the first game), whereas 458 titles were mentioned by a single person (including 195 games that appeared once as the first game). For practical reasons, the 44 most popular titles were subjected to the judging procedure (the popularity was assessed by the titles named in the first place). Each of the selected games was played by at least nine participants. This choice allowed for the inclusion of 70.6% (N = 1627) of respondents in the final analysis (whereas adding the next most popular title would increase the sample by only 0.3%). The complete list of chosen games along with basic information about each of them is presented in the appendix, in table A.1.

Gameplay Time. Respondents were asked how many hours per week they spent playing computer games in the last month. On average, respondents spent 25.92 hours per week gaming (SD = 19.73).

Attitudes Towards Minorities. Attitudes towards minorities were measured with the modified version of the Bogardus (1933) scale, which consists of three questions about acceptance of a minority representative as a family member (e.g., “Would you accept a relationship of a member of your family

(29)

29

with a Jew?”), a co-worker (e.g., “Would you accept it if a Jew was hired in your workplace?”), and a neighbour (e.g., “Would you accept a Jew as your neighbour?”). The answers were given on a 4-point Likert scale, coded so that a lower score signified higher social distance (i.e., lower out-group acceptance). The scale was validated in Poland in three nationwide, representative prejudice surveys (Bilewicz, 2009; Winiewski, 2017). The three questions of the scale were repeated for six minority groups strongly stereotyped in Poland: Jews, Roma, Eastern Europeans, Asians, Black people and Muslims. The reliabilities of the scales were high, ranging from α = .88 for the acceptance of Muslims to α = .91 for the acceptance of Roma (see table 1). The presence of these minorities in chosen games, as assessed by the independent judges, was scarce. Thus, instead of including each separate measure of social distance towards a particular out-group in the analysis, a mean for all scores was computed, creating a generalised out-group acceptance scale (reversed generalised social distance), reliable at Cronbach’s α = .96. This approach is based on the concept of secondary transfer effect, that is that attitudes towards an out-group generalise onto other out-groups (Pettigrew, 2009; Lemmer &

Wagner, 2015, see the “Transfer effects” subsection in the “Intergroup contact” part of chapter 1) thus making it possible to quantify generalised out-group attitudes (Genkova & Grimmelsmann, 2020;

Parrillo & Donoghue, 2013). The generalized social distance served as the main dependent variable in the study.

Table 1. Reliability and Descriptive Statistics of the Social Distance Scales

Scale Cronbach’s α No. of items M SD

Social distance to Jews .89 3 2.05 0.82

Social distance to Roma .91 3 1.56 1.01

Social distance to Eastern Europeans .88 3 2.21 0.81

Social distance to Asian people .89 3 2.20 0.84

Social distance to Black People .88 3 2.15 0.84

Social distance to Muslims .88 3 2.28 0.79

Generalized social distance .96 18 2.01 0.72

(30)

30

In-Game Behaviour. Situational factors that determine the effects of intergroup contact (e.g., cooperation or competition) in video games are determined by the game’s design but could also be influenced by individual choices in a game. For that reason, a measure of in-game behaviours was included. The list of assessed behaviours was based on Bartle’s player typology (Hamari & Tuunanen, 2014) as it groups player behaviours into four categories clearly relating to the type of players’

interactions in the game: socializing, aggression, exploring and achieving. The typology assumes that two of these behavioural categories (the socializing and aggression) pertain to interactions with people or other characters and thus correspond to the preference for contact. The two remaining categories (the exploring and achieving) encompass interactions with the game world and its mechanics and thus a preference for interactions other than contact with players or characters in the game. Of the categories that correspond to in-game contact, one includes behaviours related to negative contact (the aggression scale) and one encompasses behaviours related to positive contact (the socialising scale).

Thus, participants were presented with a list of behaviours common in video games and were asked to assess how often they performed each of those actions in the games they played in the past month.

Assessments were made on a 7-point Likert scale with the answers ranging from 1 “never” to 7 “almost all the time” (e.g., “How often did you use weapons (sword, machine gun, stick etc.)?”). The socializing scale included an assessment of the frequency of three behaviours: 1) giving advice about the game to others, 2) showing others how to do something and 3) cooperating with other players or NPCs to complete a task. The scale proved reliable with Cronbach’s α = .80 (with scores M = 10.9, SD = 4.51).

The exploring scale (M = 10.16, SD = 4.07, α = .69) comprised an assessment of the frequency of three types of behaviour: 1) exploring the game environment, 2) building objects or structures and 3) designing or creating something in the game. The aggression scale (M = 13.96, SD = 3.68; with low reliability of α = .58) encompassed: 1) causing damage to objects 2) damaging, hurting or killing characters controlled by other players 3) damaging, hurting or killing NPCs and 4) using weapons (guns, knives, swords etc.). Lastly, the achieving scale (M = 14.76, SD = 3.39, α = .72) measured the frequency of: 1) working toward a high score, 2) striving to win (a race, match, battle, game etc.) and 3) working on increasing skills.

(31)

31

The items in the aggression scale were only weakly correlated and the reliability could not be improved by removing any specific item. This might be because in games each type of the aggressive behaviours in the scale requires a lot of effort to design and implement, and often in one game only one of the types is fully accessible (e.g., combating NPCs without the possibility to fight other players or the ability to damage many objects in the environment). Due to its low reliability, the scale was excluded from the analysis.

Game level Variables

The game level variables were coded by two independent judges. The judges were familiar with gaming and games assessment due to their professional careers. One of the judges was a 27-year-old male working as a games tester and the second was a 35-year-old female game designer. Both judges were avid players, already familiar with most of the games. The judges were compensated financially for their work. Each of the judges assessed each of the 44 games. For each game, the judges watched online playthrough videos and accessed an online wiki of the game, reading about the characters that appear in its virtual world. After familiarizing him- or herself with a game this way, the judge filled out an online questionnaire about it.

The questionnaire included a question about the genre the game belonged to, allowing judges to choose from a list of nine genres (action, shooter, adventure, role playing game, strategy/management, massively multiplayer online RPG, massively multiplayer online battle arena, arcade and casual). The answers to the question were reliable across the judges’ evaluations (with Pearson Chi-square = 164.22; p < .001). However, they were not unanimous (see Appendix A, table A.1). Considering that genres are categorical, the disparities made including the game genre in the analysis impossible.

The main part of the questionnaire consisted of questions aimed to assess intergroup contact in a game. The judges were asked to state whether the gameplay featured any interactions with NPCs or with characters controlled by other players. The judges were unanimous in their evaluations. For each game, in which characters were present, the questionnaire continued with questions about the presence of intergroup contact and the quality of such contact in the game.

(32)

32

Intergroup Contact. The judges were presented with a list of the same minorities for which we measured players’ social distance. For each group judges indicated whether it appeared in the game (answering with a “yes” or “no” for each group). Because attitude change due to intergroup contact generalises to different out-groups, the questionnaire for the judges inquired about any additional out-groups, other than those evaluated by gamers. Thus, the judges listed all additional real-world groups (that were out-groups to the players) and, separately, all sentient fictional races present in a game (from this point on referred to as out-groups or minorities). Next, judges evaluated what percentage of characters seen by the players belonged to each of the out-groups. The judges used a slider with the scale from “0 - not present in the game” to “100% of the characters in the game” to indicate the percentage, with a numerical representation of the choice displayed next to the slider and an option to enter the value using numbers.

Statistical assessment of the reliability of the judges’ evaluations proved difficult due to the nature of the games. The out-groups presented in each game were not always easily identifiable and in some cases, judges’ evaluations differed. The problem was most prominent for fictional minorities the judges listed as present in a game but it applied to real world minorities as well. Thus, the judges disagreed as to the classifications and divisions in the game world, i.e. they were less than unanimous as to what defines a minority in the game. For example, where one judge defined real world minorities presented in a game based on nationality (e.g. listing the Irish, Scots, Germans, Italians, Scandinavians etc.) the other saw religion as the key to identifying minorities (e.g. the heretics or pagans). Fictional minorities followed a similar pattern e.g. where one judge saw one minority of elves, the other listed three separate races of elves and when one judge used in-game names for races, the other used a pop-culture equivalent (e.g. demons or spirits). Instead of attempting to map the groups from one judge’s evaluations to the second judge’s evaluations, aggregate measures for each game were computed. The measures were the count of out-groups present in a game and the percentage of out-groups in the virtual society (for real-world and fictional minorities). Moreover, the minorities for which gamers’ attitudes were assessed were very scarce and appeared in very few games, which made it impossible to analyse the direct link between the presence of each minority in a game to real-world attitudes towards that particular minority. Therefore, these groups were included in the aggregate

Do Black NPCs Matter? Effects of Virtual Contact, Construed as Intergroup Contact with Non-playable Characters in Video Games