Gamers against science: The case of the violent video games debate

Authors


Abstract

This article explores the notion that scientific research programs and empirical findings are fundamentally devalued when they threaten a perceiver's social identity. Findings from three studies show the following: (1) identification with the group of “gamers” (i.e., people who play video games on a regular basis) influences the extent to which perceivers devalue research suggesting that playing violent video games has negative consequences; (2) this effect is mediated by the feeling that the group of gamers is being stigmatized by such research (Studies 1 and 2) as well as by anger about this research (Study 2); (3) the effect of in-group identification on negative research evaluations cannot be explained by attitude or behavioral preference inconsistency (Studies 1 and 3); and (4) strongly identified gamers not only devalue a specific scientific study but also generalize their negative evaluations to the entire field of violent video games research (Study 3). The findings suggest that the influence of social identity processes on the evaluation of research is larger than it has previously been recognized. Implications of these findings for science communication are discussed. Copyright © 2013 John Wiley & Sons, Ltd.

  • This is total **** And I know Im backed up by the millions of gamers out there!
  • If u say video games influences teens behaviors u are f***ing ignorant. I shoot mother f***ers on call of duty all day and i would not do a school shooting.
  • As a gamer, I should know, I have never met anyone who had violent behavior from a V[ideo] G[ame]. …1

In most societies, scientific research is funded by public money. The public can therefore expect scientists to provide answers to what they think are important and societally relevant questions. Scientists are supposed to contribute to understanding, describing, explaining, predicting, and solving technical, economic, political, or social problems. Especially in the domain of social sciences (e.g., sociology, psychology, education, and criminology), ordinary people usually feel more confident to subjectively evaluate the quality of scientific research, as social scientific findings are often strongly connected to people's everyday rationality (Flyvbjerg, 2001; Haslam & Bryman, 1994).

Research on science communication has shown that public evaluations of scientific reports are influenced and affected by recipients' prior beliefs (Scheufele, Corley, Shih, Dalrymple, & Ho, 2008), prior attitudes (Munro, 2010), prior knowledge (Allum, Sturgis, Tabourazi, & Brunton-Smith, 2008), or interactions of these factors (Ho, Brossard, & Scheufele, 2008). For example, people have a more critical stance toward nanotechnology when it contradicts their religious beliefs (Brossard, Scheufele, Kim, & Lewenstein, 2008). Moreover, people holding hierarchical-individual values perceive global warming as not dangerous and, therefore, devalue research findings that demonstrate the negative consequences of global warming. On the contrary, people holding egalitarian-communitarian values perceive global warming as dangerous and therefore devalue research showing no adverse effects of global warming (Kahan, Braman, Slovic, Gastil, & Cohen, 2008; Kahan, Jenkins-Smith, & Braman, 2011). More recent research shows that accepting versus rejecting scientific findings, such as the facts that HIV causes AIDS and that smoking causes lung cancer, is associated with conspiracy beliefs (Lewandowsky, Oberauer, & Gignac, 2013). Being confronted with research findings that are inconsistent with one's prior beliefs even sparks fundamentally critical attitudes toward science in general, such that people argue that a particular topic cannot be studied scientifically (Munro, 2010). Taken together, these findings show that attitudes and values held individually or derived from a social or cultural affiliation (e.g., Kahan et al., 2011) influence how laypersons evaluate scientific findings and how they “engage” with science in their daily lives.

The statements quoted at the beginning of this article display a particular way of “engaging” with science. They illustrate how members of a particular group (in this case, people who play violent video games on a regular basis, henceforth referred to as “gamers” for reasons of simplicity) disregard and even disqualify scientific research suggesting that playing violent video games can have detrimental social or developmental effects. One might argue that devaluing research reflects a form of motivated reasoning on an individual level (cf. Ditto & Lopez, 1992; Kunda, 1990). In the context of the violent video games debate, such motivated reasoning biases might reflect two forms of inconsistencies: First, gamers like and play these games and therefore oppose any statement that sheds a bad light on these games. Thus, research that might ultimately result in sales bans or accessibility restrictions is devalued because it collides with gamers' behavioral preferences (i.e., their gaming habits). Second, video game players who hold a positive attitude toward violent video games and do not believe in any negative effects might disfavor research that contradicts these beliefs. In other words, such research might be devalued because it is inconsistent with gamers' prior attitudes.

In the present article, we argue that these individual-level explanations cover only a part of the picture. We hypothesize that identification with a group is sufficient to trigger a biased evaluation of group-relevant scientific findings and that devaluing scientific research on the detrimental effects of playing violent video games can thus be better explained by group-based processes: Being a gamer also implies belonging to a social category, and the stronger a person's identification with that category, the stronger one's motivation to maintain a positive social identity by being a member of this category (Tajfel & Turner, 1979, 1986). From a social identity perspective, empirical research reporting detrimental effects of playing violent video games threatens gamers' positive social identity: Such findings imply a stigmatization of the group of gamers. Fundamentally criticizing such research can therefore be conceptualized as a form of “collective action” to defend the group's image and one's social identity (e.g., Van Zomeren, Postmes, & Spears, 2008).

GAMERS AS A SOCIAL CATEGORY

Gaming has become increasingly popular over the last decades (Williams, Yee, & Caplan, 2008). With an increased dispersal of online games over the Internet, people are now able to connect and play with and against each other in real time. These developments have contributed to a growing community of people who consider themselves as “gamers.” Because each game has its own fans, game-specific subcommunities have emerged in which gamers frequently communicate with each other and meet in the virtual or even real world. Besides sharing their gaming habits, gamers also commit and contribute to common activities: They are active in lobby groups (e.g., the “Video Game Voters” or “Gamers against Rejection”2), they organize common events (e.g., trade fairs such as the E3 or Gamescom), and they share a common (sub)culture, which is mainly defined in terms of common social practices and a shared identity created in the gamespace (Taylor, 2006). Thus, there is reason to expect that gamers constitute a visible social category and that people differ in the extent to which they perceive themselves to belong to this category.

Perceiving oneself as a member of a group creates the basis for group identification processes, which influence the way individuals interpret group-relevant information (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987). The question of whether or not playing violent video games has detrimental effects is highly identity relevant for gamers. Folk wisdom has associated heavy gaming with acts of intensive violence, such as school shootings and cruel murders (e.g., Ferguson, 2008). The stigmatization of gamers in the public discourse constitutes an intergroup context in which the social category of being a gamer becomes salient whenever the question about potentially harmful effects of playing violent video games is discussed. Accordingly, we argue that social factors should be considered in order to understand the dynamics and emotionality of gamers' reactions toward research on the effects of violent video games.

When an intergroup context is salient, an individual's reaction toward persuasive information is heavily influenced by the in-group's position toward this information (Cohen, 2003; Wood, 2000). In the field of gender differences, Morton, Haslam, Postmes, and Ryan (2006) demonstrated that a scientific study was more positively evaluated when it affirmed participants' gender identity. When a scientific study portrayed their own gender group in a positive light, participants considered the study to be “scientific” and were more interested in this research. The authors concluded that scientific findings are more likely to be perceived as credible and plausible to the extent that they provide people with a positive sense of identity, irrespective of the actual scientific state of the art. Research on social identity processes suggests that such effects should be particularly pronounced among individuals who strongly identify with their respective group (e.g., Ellemers, Spears, & Doosje, 1997). For example, Jetten, Spears, and Manstead (1997) demonstrated that strongly identified group members are more likely to behave in accordance with salient group norms than weakly identified group members. Regarding gamers' evaluation of research on the effects of violent media, we therefore assume that strongly identified gamers are more likely to reject and devalue such research (to the extent that this research constitutes a social identity threat) than weakly identified gamers.

CONSEQUENCES OF SOCIAL IDENTITY THREAT

When people strongly identify with their in-group and when an intergroup context is salient, their behaviors are more strongly shaped by their social self than by their personal self (Turner et al., 1987). For example, being a strongly identified member of a problematized social group goes along with increased perceived stigmatization (Major & O'Brien, 2005; Major, Quinton, & Schmader, 2003). In our case, if a scientific study corroborates an aggression-enhancing effect of violent video games, people who strongly identify with the group of gamers should feel more strongly stigmatized by such research than weakly identified gamers. Furthermore, perceived stigmatization increases the likelihood for defensive-hostile reactions against the source of that stigmatization (Twenge, Baumeister, Tice, & Stucke, 2001) and is assumed to increase the likelihood for collective action (Biernat & Dovidio, 2000). In other words, perceived stigmatization can mediate the effect of threat on defensive and hostile reactions toward the source of that threat.

Besides perceived stigmatization, people who strongly identify with a group facing an identity threat also experience more negative emotions (Ellemers, Spears, & Doosje, 2002; McCoy & Major, 2003; Van Zomeren, Spears, & Leach, 2008) such as anger or moral outrage. Furthermore, anger has been found to mediate the effect of threat on (collective) action tendencies (Leach, Iyer, & Pedersen, 2006, 2007; van Zomeren et al., 2008). Because strongly identified individuals who experience a threat to their social identity are also more likely to show aggressive reactions toward the source of that threat (Branscombe & Wann, 1992), there is reason to assume that anger explains not only collective action tendencies but also aggressive reactions—such as devaluing the source of that threat.

ALTERNATIVE EXPLANATIONS

Besides these social factors derived from social identity theory and research on collective action, individual factors might also account for a biased evaluation of research findings. For example, strongly identified gamers might simply believe that violent video games are harmless. In this case, the negative evaluation of scientific findings by strongly identified gamers might merely reflect an attitude inconsistency effect (Lord, Ross, & Lepper, 1979; Munro, 2010) instead of an instance of “collective action.” Another alternative explanation could be that identification with the group of gamers is strongly associated with the habit of playing violent video games. Thus, people who like and regularly play these games might fear that research showing a media violence effect might ultimately lead to sales bans or accessibility restrictions. In other words, research demonstrating that playing violent video games has harmful effects collides with gamers' behavioral preferences. Notably, it is theoretically possible that people devalue scientific evidence showing negative effects of a certain behavior (such as gaming) even though they acknowledge that such behavior can have negative effects. In line with this argument, Weinstein (1998) reported that smokers do indeed acknowledge that smoking has negative effects. However, at the same time, smokers tend to evaluate scientific information demonstrating negative effects of smoking as less credible than nonsmokers do. In a similar vein, gamers might acknowledge that playing violent video games can have negative effects, but they might still discredit scientific studies demonstrating these effects. Taken together, one might argue that the effect of social identification on the biased evaluation of scientific results merely reflects a behavioral preference inconsistency effect and/or an attitude inconsistency effect. However, we argue that social identification with the group of gamers can lead to a biased evaluation of scientific findings over and above such individual-level effects.

HYPOTHESES AND RESEARCH GOALS

The present article describes three studies designed to test three hypotheses: The first hypothesis is that one's identification with the group of gamers predicts a biased evaluation of scientific research findings (i.e., a fundamentally negative evaluation of findings demonstrating detrimental effects of playing violent video games) and of the entire area of media effects research (Studies 1, 2, and 3). The second hypothesis is that this effect is mediated by perceived stigmatization and anger (Studies 1 and 2). The third hypothesis is that identification has a unique effect on the biased evaluation of scientific findings over and above one's personal effect beliefs about violent video games (i.e., attitude inconsistency) or one's gaming habits (i.e., behavioral preference inconsistency; Studies 1 and 3).

STUDY 1

Study 1 provides a first test of the hypothesis that the extent to which people identify with the group of gamers positively predicts harsh and fundamentally critical attitudes toward research that empirically corroborates the violent-games-effect hypothesis and that this relation is mediated by perceived stigmatization. Additionally, Study 1 was designed to test whether the two aforementioned alternative explanations, that is, effect beliefs about violent video games and gaming habits, can account for the hypothesized effect.

Method

Participants

In all, 561 undergraduate students from various disciplines (56.5% women) participated in mass testing sessions. Two hundred and fourteen participants (i.e., 38.1%) were excluded from further analyses because they indicated that they have not played any video games during the last 12 months (n = 205) or had more than 25% missing values on one or more scales (n = 9). Therefore, the final sample consisted of 347 participants (32.6% women). Ages ranged between 18 and 50 years (M = 23.15; SD = 3.81).

Materials and Measures

A correlational design was used to examine the relationship between identification with the group of gamers and a biased evaluation of scientific research findings.

Negative evaluations

First, participants read a short description of how researchers investigate the effects of violent video games (APPENDIX). The description explicitly mentioned that researchers aim to investigate the aggression-enhancing effects of violent video games; according to our hypotheses, strongly identified gamers should even regard such a subtle description as a threat to their social identity. Next, participants evaluated the entire area of research on the effects of violent video games and the respective researchers using a 10-item scale (“I think the results of this research can be meaningfully applied to real-life contexts” (recoded), “…these researchers are sometimes not very competent,” “…the results of this research are unambiguous” (recoded), “…these researchers just find what they want to find,” “…this research yields reliable results” (recoded), “…this research yields important results” (recoded), “…scientists who do research in this field are often biased,” “…one can draw useful conclusions for the real life from this kind of research” (recoded), “…this kind of research is not very meaningful,” and “…the methodology is fundamentally useless to investigate the effects of violent video games”; Cronbach's α = .87).

Effect beliefs were measured by asking participants to what extent they think that violent video games increase versus decrease the following: (1) aggressive thoughts; (2) aggressive dispositions; and (3) hostile tendencies (Cronbach's α = .77). The items were presented together with three items indicating positive effects of video games in order to reduce demand effects. Ratings were obtained on a 7-point scale (−3 = decrease, 0 = no effect, 3 = increase). Thus, positive values indicate a belief in the detrimental effects of violent video games.

Gaming habits

Participants were asked whether they had played any video games during the last 12 months, and, if so, how many hours per week they usually spend playing each of eight different video game genres (i.e., “shooter” and battle games, role-playing games, strategy games, racing or sport games, massively multiplayer online role-playing games, skill games and jump‘n'runs, simulation, or browser games). These genres had been pretested by experts on violent video games (15 students of computer science with extensive experience in playing video games) on a 6-point scale (0 = no violent content, 5 = strong violent content). The intraclass correlation was .72 across the 15 raters. A violence frequency index was computed by multiplying the time spent playing each category with the violence rating for that category (for a similar approach, see Anderson et al., 2008; Möller & Krahé, 2009).

Perceived stigmatization of the group of gamers was measured with a two-item scale (“People who regularly play video games are often devalued in the public” and “The public takes a one-sided stance toward people who regularly play video games;” Cronbach's α = .67). Ratings were obtained on a 6-point scale (1 = strongly disagree, 6 = strongly agree).

Identification with the group of gamers was measured with four items (“Being a ‘gamer’ is an important part of my identity,” “Being a ‘gamer’ gives me a good feeling,” “Being a ‘gamer’ is an important part of how I see myself,” and “I feel connected with other ‘gamers’”; Cronbach's α = .91). Ratings were obtained on a 6-point scale (1 = strongly disagree, 6 = strongly agree).

Finally, demographic information was assessed. Completing the questionnaire took about 10 minutes. One MP3 player was raffled among all participants who completed the survey. Descriptive statistics and correlations between all variables are displayed in Table 1.

Table 1. Descriptive statistics and correlations between variables in Study 1
VariableM (SD)Correlations
(1)(2)(3)(4)(5)
  • Note: N = 347.

  • p < .10;

  • *

    p < .05;

  • **

    p < .01;

  • ***

    p < .001.

Gaming habits (1)11.19 (20.72)1.00    
Effect beliefs (2)0.44 (0.81)−0.21***    
Identification with the group of gamers (3)2.28 (1.29)0.46***−0.31***   
Perceived stigmatization (4)4.24 (1.05)0.13*−0.13*0.29***  
Negative evaluations (5)4.02 (0.90)0.21***−0.40***0.33***0.24*** 

Results

Negative Evaluations

According to our reasoning, strongly identified gamers should be more prone to discredit research and the researchers pursuing negative effects of violent video games. In line with our first hypothesis, identification with the group of gamers was positively related to a fundamentally negative evaluations of such research, r = .33, p < .001 (Table 1). However, gaming habits and effect beliefs were also significantly correlated with negative evaluations as well as with identification. Importantly, the effect of identification was still significant after controlling for gaming habits and effect beliefs, B = 0.14, SE(B) = 0.04, p < .001 (Table 2). Additionally, only effect beliefs also significantly predicted negative evaluations, whereas gaming habits did not. Thus, Hypotheses 1 and 3 were supported.

Table 2. Results of multiple regression analyses (Study 1)
Predictor variablesModel 1 (DV: negative evaluations)Model 2 (DV: perceived stigmatization)Model 3 (DV: negative evaluations)
  • Note: N = 347. DV, dependent variable.

  • p < .10;

  • *

    p < .05;

  • **

    p < .01;

  • ***

    p < .001.

Constant term5.2684.0154.777
Gaming habits0.002−0.0060.002
Effect beliefs−0.359***−0.063−0.351***
Identification with the group of gamers0.14***0.225***0.113**
Perceived stigmatization0.122**
R20.2080.0840.227

Perceived Stigmatization

We expected that the effect of identification would be mediated by perceived stigmatization of the group of gamers (Hypothesis 2). This hypothesis was tested with a mediation model (Figure 1). The indirect effect of identification via perceived stigmatization on negative evaluations was tested by inspecting bias-corrected 95% confidence intervals (CIs) obtained by bootstrapping using Hayes' (2013) PROCESS macro (5000 resamples). Even though the identification effect on negative evaluations was still significant after controlling for perceived stigmatization, the indirect effect was significant, B = 0.03, SE(B) = 0.01, 95% CI [0.01, 0.06], irrespective of whether or not gaming habits and effect beliefs were controlled for (indirect effect including covariates: B = 0.03, SE(B) = 0.01, 95% CI [0.01, 0.06]). In other words, perceived stigmatization partially mediated the effect of identification on negative evaluations under social identity threat, even when controlling for gaming habits and effect beliefs (Figure 1).

Figure 1.

Mediation model (Study 1). N = 347. †p < .10; *p < .05; **p < .01; ***p < .001

Discussion

The results of Study 1 provide first evidence for our notion that the evaluation of research findings is influenced by the degree to which gamers identify with their in-group and that this bias is fueled by a perceived stigmatization of the in-group. Notably, the effect of in-group identification (and the indirect effect of identification on negative evaluations via stigmatization) existed over and above gaming habits and effect beliefs. These findings support our hypothesis that potentially threatening research findings trigger harsh and fundamentally negative attitudes toward the respective research and the researchers and that this is particularly the case for strongly identified gamers.

Even though the results of Study 1 generally support our predictions, there are some limitations and shortcomings. Firstly, one should note the generally low levels of social identification with the group of gamers in our sample (M = 2.28 on a scale from 1 to 6). Secondly, we used a rather subtle threat induction. Possibly, our effects would have been even stronger if the mean and the variance of identification scores had been higher in our sample and the threat induction had been stronger. Thirdly, all of our participants were students, which might limit the generalizability of our results. Fourthly, the stigmatization items were rather unspecific. A more specific measure of stigmatization, pointing to the perception that gamers are stigmatized particularly by scientific research on the negative effects of playing violent video games, might have been better suited in order to address our hypothesis.

Finally and most importantly, we did not implement a control group (i.e., an experimental condition in which a nonthreatening research program is described). Thus, we do not know whether the effect of identification necessarily depends upon a social identity threat. Possibly, strongly identified gamers evaluate research on the effects of violent video games more negatively regardless of its implications for the in-group. Thus, Studies 2 and 3 used an experimental approach in order to strengthen the internal validity and to replicate the biased evaluation effect. Additionally, we used non-student samples in order to increase the generalizability of our results.

STUDY 2

Study 2 aims at replicating the effect of identification and its mediation via perceived stigmatization on biased evaluations of scientific research with a stronger design and a more specific stigmatization measure. Additionally, this study was designed to illuminate the underlying mechanisms of this bias in more detail. Besides perceived stigmatization, anger has been identified as an important mediator in explaining aggressive actions toward the source of a social identity threat (Branscombe & Wann, 1992) and collective action tendencies (e.g., Leach et al., 2007). Thus, it seems reasonable to assume that strongly identified gamers might not only feel stigmatized as a group but also feel angry about research demonstrating that playing violent video games can have detrimental effects. Hence, Study 2 provides a test of the hypothesis that anger might be a driving force in the negative evaluation of potentially threatening research besides a perceived stigmatization.

Method

Design and Procedure

Study 2 used data from an online-based experimental survey. Measures were assessed at two occasions with an interval of 2 weeks between them in order to reduce carryover effects. At Time 1, identification with the group of gamers was measured. At Time 2, participants were confronted with a short text summarizing the findings of a published study on the effects of playing violent video games. Depending on experimental conditions, this study was said to either corroborate or refute a violent games effect. Afterwards, participants' reactions toward the particular study and its authors were assessed.

Participants

Data were collected with the help of a professional sampling agency. This agency has access to a large participant pool available for marketing research and online surveys. Participants can be sampled from this pool on the basis of multiple—usually demographic—criteria. In our case, no particular sampling criteria were applied. On the first page of the survey, participants were asked to indicate whether they play video games on a regular basis (yes/no). Because we were specifically interested in gamers' reactions to science on violent video games, only those who indicated playing video games regularly were allowed to start the survey. In all, 548 participants responded to the invitation; 361 (48.6% women) of them indicated playing video games regularly and were directed to the survey. The other 187 were thanked and redirected to the agency site. Of those who started the survey, 350 (97%) completed it successfully. As suggested by Paolacci, Chandler, and Ipeirotis (2010) and by Oppenheimer, Meyvis, and Davidenko (2009), the survey also comprised an attention test (“If I complete this questionnaire attentively and focused, I check a ‘three’ here”). Participants who failed this test (n = 67) or had missing values on more than 25% of the items on the identification measure (n = 1) were omitted from all further analyses; this reduced the number of cases to 282 (47.9% women). Completing this first survey took about 9 minutes; participants were rewarded with a raffle ticket worth 1800 “credit points” (approximately US$0.58).

Two weeks later, the same 282 individuals were invited to take part in a second survey on “consumer attitudes.” Of those, 201 (71%) followed the invitation and started the survey, of which 183 (91%) finished the questionnaire. Data from the two measurement occasions were matched on the basis of a personalized code. Again, we included attention tests, and those who failed these tests (n = 85) and had missing values on 25% of the items within any scale (n = 1) were excluded from all subsequent data analyses.3 Thus, the total dataset consisted of 97 cases (49.5% women). Ages ranged between 18 and 75 years (M = 43.0; SD = 13.15). Completing this survey took about 11 minutes; participants were rewarded with a raffle ticket worth 1800 “credit points” (approximately US$0.58).

Materials and Measures

At Time 1, demographic information as well as participants' identification with the group of gamers were measured with five items (“I feel solidarity with other ‘gamers’,” “I feel committed to the group of ‘gamers’,” “I am glad to be a ‘gamer’,” “I think that ‘gamers’ have a lot to be proud of,” and “It is nice to be a ‘gamer’.”). Additionally, we included one item to measure identification with the group of gamers on a broader level adapted from Postmes, Haslam, and Jans (2013) (“I identify with the group of ‘gamers’”). All items loaded on one factor and constituted a reliable scale (Cronbach's α = .88). Response scales ranged from 1 (not at all true) to 6 (very much true).

At Time 2, participants were first confronted with a short text summarizing the results from a published study on the effects of playing violent video games. Participants were randomly assigned to either a “harmful” (i.e., the study shows that violent games do increase aggressive tendencies; the study described here was taken from Anderson & Dill, 2000, Study 2) or a “harmless” (i.e., the study shows that violent video games do not increase aggressive tendencies; the study described here was taken from Ferguson et al., 2008, Study 1) condition. Importantly, identification with the group of gamers did not reliably differ between the two conditions, t(95) = 0.65, p = .52. The summary read as follows (“harmless” condition in brackets):

In an experiment conducted by Craig Anderson and colleagues [Christopher Ferguson and colleagues] participants were randomly assigned to one of two conditions. In one condition participants played a violent video game, participants in the other condition played a non-violent video game. After playing the video game all participants were asked to participate in a reaction time task in which they competed (ostensibly) against an opponent seated in another room. Whenever participants won a round, they had the opportunity to punish the opponent. The punishment consisted of a very unpleasant noise; participants were asked to calibrate the sound's duration and its volume; these settings served as measures of participants' aggressive tendencies. Comparing the average duration and volume settings between the two conditions showed that those who had played the violent video game reacted [did not react] more aggressively than those who played the non-violent video game. The authors of the study concluded that consuming violent video games leads [does not lead] to an increase in aggression. The authors stated that “violent video games provide a forum for learning and practicing aggressive reactions” [“playing violent video games does not constitute a risk factor for behaving aggressively”].

After reading the summary, participants were asked to what extent they felt that this research stigmatizes gamers, with two items (“I think that this study was designed to devalue video game players” and “I think that this study denounces video game players;” Cronbach's α = .89). Response scales ranged from 1 (not at all true) to 6 (very much true).

Next, participants were asked to evaluate the study and the competence of the authors on seven items (six adapted from Study 1 plus the item “I think that this study was a waste of public money;” Cronbach's α = .91). Response scales ranged from 1 (not at all true) to 6 (very much true) with higher values indicating a more negative evaluation of the study and the researchers responsible for these studies.

Anger about the study was measured with two items (“I was irritated about the study” and “I was outraged when I read the summary of the study;” Cronbach's α = .91). Response scales ranged from 1 (not at all true) to 6 (very much true).

Descriptive statistics and correlations between all dependent variables and identification at Time 1 are displayed in Table 3.

Table 3. Descriptive statistics and correlations between variables in Study 2
VariableM (SD)Correlations
(1)(2)(3)(4)
  • Note: N = 97.

  • p < .10;

  • *

    p < .05;

  • **

    p < .01;

  • ***

    p < .001.

Identification with the group of gamers at Time 1 (1)2.81 (1.14)1.00   
Negative evaluations (2)3.29 (1.23)0.111.00  
Anger about the study (3)2.02 (1.40)0.20*0.64***1.00 
Perceived stigmatization (4)2.39 (1.51)0.39***0.49***0.47***1.00

Results

Negative Evaluations

According to Hypothesis 1, we expected that identification with the group of gamers (measured at Time 1) moderates the effect of condition on evaluation such that strongly identified gamers react particularly negatively and critically toward a study that corroborates the violent-games-effect hypothesis, but not toward a study that refutes it. This was tested via moderated regression analysis (cf. Cohen, Cohen, West, & Aiken, 2003). Condition was effect coded (−1 = harmless, +1 = harmful), and identification was centered prior to computing product terms (cf. Aiken & West, 1991). As expected, negative evaluations were predicted by the Condition × Identification interaction, B = 0.27, SE(B) = 0.11, p = .01, ΔR2 = .06. Simple slopes analyses revealed that the effect of condition was significant for strongly identified gamers (i.e., 1 SD above the sample mean), B = 0.38, SE(B) = 0.18, p = .03, but not for weakly identified gamers (i.e., 1 SD below the sample mean), B = −0.24, SE(B) = 0.17, p = .17. No other effects were significant on a 5% level. Predicted means are displayed in Figure 2.

Figure 2.

Negative evaluations by experimental condition and identification with the group of gamers (Study 2)

Perceived Stigmatization

According to Hypothesis 2, we expected that the Identification × Condition interaction effect is mediated by perceived stigmatization. This hypothesis was tested with a moderated mediation model (Figure 3). The Identification × Condition interaction effect on perceived stigmatization was only marginally significant, B = 0.21, SE(B) = 0.12, p = .08, ΔR2 = .02. However, simple slopes analyses confirmed that the effect of condition was significant for strongly identified gamers, B = 0.75, SE(B) = 0.18, p < .001, but not for weakly identified gamers, B = 0.27, SE(B) = 0.18, p = .15 (Table 4 and Figure 4). The conditional indirect effects of Condition × Identification on negative evaluations via perceived stigmatization were tested by inspecting bias-corrected 95% CIs obtained by bootstrapping using Hayes' (2013) PROCESS macro (5000 resamples). In line with our hypothesis, the indirect effect was significant for strongly identified gamers, B = 0.33, SE(B) = 0.11, 95% CI [0.15, 0.59], but not for weakly identified gamers, B = 0.12, SE(B) = 0.07, 95% CI [−0.001, 0.29].

Figure 3.

Moderated mediation model (Study 2)

Table 4. Results of multiple regression analyses (Study 2)
Predictor variablesModel 1 (DV: negative evaluations)Model 2 (DV: perceived stigmatization)Model 3 (DV: negative evaluations)Model 4 (DV: anger about the study)Model 5 (DV: negative evaluations)
  • Note: N = 97. DV, dependent variable.

  • p < .10;

  • *

    p < .05;

  • **

    p < .01;

  • ***

    p < .001.

Constant term3.272.372.222.002.18
Condition: study type (−1 harmless; +1 harmful)0.070.51***−0.160.16−0.04
Identification with the group of gamers at Time 10.050.43***−0.140.18−0.02
Condition × Identification (Time 1)0.27*0.210.180.27*0.13
Perceived stigmatization0.44***
Anger about the study0.55***
R20.080.280.290.100.43
Figure 4.

Perceived stigmatization by experimental condition and identification with the group of gamers (Study 2)

Anger

Finally, we hypothesized that the interaction effect of Condition × Identification is mediated by anger about the study. This hypothesis was again tested with a moderated mediation model (Figure 3). In line with our reasoning, identification at Time 1 moderated the effect of condition on anger about the study, B = 0.27, SE(B) = 0.12, p = .03, ΔR2 = .05. Simple slopes analyses revealed that the effect of condition was significant for strongly identified gamers, B = 0.47, SE(B) = 0.20, p = .02, but not for weakly identified gamers, B = −0.14, SE(B) = 0.20 p = .46 (Table 4 and Figure 5). The conditional indirect effects of condition on negative evaluations via anger were tested by inspecting bias-corrected 95% CIs obtained by bootstrapping using Hayes' (2013) PROCESS macro (5000 resamples). In line with our hypothesis, the indirect effect was significant for strongly, B = 0.26, SE(B) = 0.12, 95% CI [0.04, 0.52], but not for weakly identified gamers, B = −0.08, SE(B) = 0.10, 95% CI [−0.29, 0.11].

Figure 5.

Anger about the study by experimental condition and identification with the group of gamers (Study 2)

Discussion

The results of Study 2 replicate and extend our findings from Study 1. They provide additional support for our notion that the evaluation of research findings is influenced by the degree to which gamers identify with their in-group and that this effect is mediated by group-based perceptions (stigmatization) and negative emotions (anger). When confronted with a research finding that corroborated the violent-games-effect hypothesis, strongly identified gamers reacted with more anger, a higher degree of perceived stigmatization, and more negative evaluations than did weakly identified gamers. Anger and perceived stigmatization mediated the effect of condition on negative evaluations only for strongly identified gamers; in other words, these participants were more likely to criticize a study reporting detrimental effects of violent games because they experienced anger about the study and perceived it as more stigmatizing. Notably, negative evaluations clearly extended beyond the particular study participants were confronted with: Some of the items referred to the researchers' competence and their (un)biasedness. This corroborates our notion that strongly identified gamers expressed a general bias toward entire research areas if they produce potentially identity-threatening findings.

Even though the results of Study 2 generally support our predictions, there are two limitations. Firstly, it is important to note that although anger about the study participants were confronted with was—technically speaking—a mediator variable in our model (Figure 3), it was measured after the evaluation of the study (that is, the dependent variable). This was performed in order to rule out the possibility that asking participants explicitly about their emotions might have created an artificial demand or might have unwanted priming effects, which, in turn, could have influenced evaluations artificially. Secondly, the attrition rate in the study was quite substantial. Although using attention tests in order to increase the statistical power of a test leads to comparable attrition rates even in laboratory settings (Oppenheimer et al., 2009), in our case, this might have been especially due to the payment politics of the hired sampling agency: Participants are not immediately paid for their participation. Instead, they receive virtual “credit points” in order to take part in raffles or are disbursed when they reach a certain amount. This might have motivated participants to skip reading the scientific summary in order to finish the study as quickly as possible. We tried to avoid this problem in Study 3 by recruiting participants in gaming forums in which the motivation to read about research on the effects of violent video games should be higher. Whereas Study 2 was designed to investigate potential mediators for the biased evaluation of scientific findings elicited by a social identity threat, Study 3 investigated whether the effect of identification is truly independent from an attitude inconsistency and/or behavioral preference inconsistency effect and generalizes to the entire field of violent video games research.

STUDY 3

Study 3 aimed at replicating the moderating effect of identification on negative and critical evaluations of social scientific research, and to once more rule out two previously mentioned alternative explanations, that is, beliefs about the effects of violent video games and gaming habits. We tested whether the effect of identification on negative evaluations persists even after controlling for effect beliefs and gaming habits. In contrast to Study 1, we used an experimental design to test this hypothesis. Additionally, we tested whether gamers' negative evaluations extend beyond the particular study and its authors and generalize to the entire field of violent video games research.

Method

Participants

The study was advertised in several German gaming Web forums and mailing lists. During the time this study was online, the forums were monitored for comments about the study, which would have influenced future participants' responses. No such comments were posted. In all, 533 people started the questionnaire; 228 of them (43%) finished it successfully. Twenty-nine participants were excluded from further analyses because these people indicated that they had not played any video games during the last 12 months (n = 4), failed the attention test (discussed later; n = 22) or had more than 25% missing values on one or more scales (n = 3).4 Therefore, the final sample consisted of 199 participants (7.5% women). Ages ranged between 16 and 45 years (M = 23.13; SD = 5.96). We included minors (16 years and older) in order to increase the generalizability of our results. More than two thirds of people in Germany between 14 and 17 years play video games, and especially young people are more likely to identify more strongly with the group of gamers (Quandt, Festl, & Scharkow, 2011). Full informed consent was obtained before participants started the study. Furthermore, the study involved neither deception nor did it have any negative consequences. Thus, all participants were treated in accordance with ethical guidelines.

Materials and Measures

The design of Study 3 was similar to Study 2 with the only exception that all measures were assessed at a single measurement occasion. First, participants were asked whether or not they had played any video games during the last 12 months. Those who answered “no” to this question were thanked and redirected to a different website.

Gaming habits

Those who indicated playing video games during the last year were first asked how many hours per week they usually spend playing each of eight different video game genres of which we computed a violence frequency index (Study 1).

Next, effect beliefs (Study 1, plus the item “aggressive behavioral tendencies;” Cronbach's α = .81) and identification with the group of gamers (five-item scale used in Study 1, plus one item adapted from Mael and Tetrick, 1992, which seems particularly important in the context of the violent video games debate: “When somebody criticizes gamers, it feels like a personal insult”; Cronbach's α = .85) were measured. Ratings were obtained on a 6-point scale (1 = strongly disagree, 6 = strongly agree).

Participants were informed that the study they would have to evaluate was randomly chosen from a pool of different empirical studies. This was performed in order to prevent participants from assuming that the presented study was prototypical for the entire field of research. The texts they read were the same as in Study 2. Again, participants were randomly assigned to either a harmful condition, in which the results of the study corroborated the violent-games-effect hypothesis, or a harmless condition, in which the results of the study refuted this hypothesis. Importantly, gaming habits, beliefs about the effects of violent video games, and identification with the group of gamers did not reliably differ between these two conditions (all t's(197) ≤ 0.47, p's ≥ .64). Next, participants evaluated the study using the same seven-item scale as in Study 2. Additionally, four items explicitly referring to the entire research area on the effects of violent video games (Study 1) were added to the scale (Cronbach's α = .85).

Finally, demographic information was assessed. Completing the survey took about 16 minutes. Among all participants who completed the survey, 10 online shopping vouchers (worth €20 each) were raffled. Descriptive statistics and correlations between all variables are displayed in Table 5.

Table 5. Descriptive statistics and correlations between variables in Study 3
VariableM (SD)Correlations
(1)(2)(3)(4)
  • Note: N = 199.

  • p < .10;

  • *

    p < .05;

  • **

    p < .01;

  • ***

    p < .001.

Gaming habits (1)36.41 (30.53)    
Effect beliefs (2)−0.10 (0.79)−0.21**   
Identification with the group of gamers (3)3.70 (1.23)0.45***−0.10  
Negative evaluations (4)3.81 (0.94)0.12−0.060.09 

Results

Our central hypothesis (i.e., strongly identified gamers express more fundamentally negative evaluations in the harmful condition) was tested via moderated regression analysis. Again, condition was effect coded (−1 = harmless, +1 = harmful), and identification was centered prior to computing product terms. As expected, negative evaluations were predicted by the Condition × Identification interaction, B = 0.15, SE(B) = 0.05, p = .001, ΔR2 = .04. Simple slopes analyses revealed that the effect of condition was larger for strongly identified gamers (i.e., 1 SD above the sample mean), B = 0.66, SE(B) = 0.08, p < .001, than for weakly identified gamers (i.e., 1 SD below the sample mean), B = 0.29, SE(B) = 0.08, p < .001, although the latter was significant as well. Additionally, we found a main effect of experimental condition, B = 0.48, SE(B) = 0.06, p < .001: Participants expressed more positive attitudes toward media violence research in the “harmless” condition than in the “harmful” condition. Predicted means are displayed in Figure 6.

Figure 6.

Negative evaluations by experimental condition and identification with the group of gamers (Study 3)

Next, we tested whether the moderator effect of identification even held after controlling for gaming habits and effect beliefs. When controlling for these variables (including main effects and the respective condition interaction terms), the Condition × Identification interaction effect was still significant and in the expected direction, B = 0.13, SE(B) = 0.05, p = .01, ΔR2 = .02, whereas neither the Condition × Beliefs interaction nor Condition × Gaming Habits interaction effect significantly predicted negative evaluations (p's ≥ .22). Again, the effect of condition was larger for strongly identified gamers (i.e., 1 SD above the sample mean), B = 0.64, SE(B) = 0.08, p < .001, than for weakly identified gamers (i.e., 1 SD below the sample mean), B = 0.32, SE(B) = 0.09, p < .001, although the latter was significant as well. This demonstrates that social identification predicts a biased evaluation of research on violent games effects over and above effect beliefs and gaming habits.

Discussion

Results from Study 3 provide additional evidence that a biased evaluation of scientific research findings can be explained by gamers' identification with their in-group and that the effect of identification on such evaluations is independent from effect beliefs and gaming habits. Notably, strongly identified gamers expressed harsh criticism toward the entire field of media violence research and the respective researchers although they were told that the study they read was randomly chosen from a pool of different studies with potentially different results. We can therefore assume that after being confronted with such research, strongly identified gamers adopt a fundamentally critical style of receiving and engaging with the respective scientific area. We speculate that this is because this kind of research symbolizes a threat to the social identity of gamers.

However, two findings are worth discussing. Firstly, in contrast to Study 2, we found a science-discrediting effect also for weakly identified gamers. This might be due to the different population our samples were recruited from. In Study 2, “weakly” identified gamers had an average identification score of 1.58, whereas “weakly” identified gamers in Study 3 had a mean score of 2.47. Even though we did not use the same social identification items in both studies, the difference demonstrates that participants in Study 2 were less strongly identified with the group of gamers than participants in Study 3. This difference in participants' average levels of identification might explain why we found a biasing effect also for “weakly” identified gamers in Study 3.

Secondly, and in contrast to Study 1, identification with the group of gamers did not correlate significantly with personal beliefs regarding negative effects of playing violent video games (r = −.10, p = .16). This finding is surprising at first glance, but it might be due to the more diverse and more strongly identified sample compared with that of Study 1. Some strongly identified gamers might think that violent video games can principally have aggression-enhancing effects on some people, but not on themselves (“third-person effect”, Davison, 1983, also see Goldstein, 2005, for a similar argument).

GENERAL DISCUSSION

Research on the general public's understanding of and engagement with science has grown during recent years. This research has mainly focused on people's attitudes toward science as a function of their prior beliefs, prior attitudes, and prior knowledge about the respective scientific area (Ho et al., 2008) and on the cultural dependence of science evaluation (Kahan et al., 2011). Much less research has been devoted to the question of whether social identity processes also play a role for the formation of attitudes toward scientific research findings and programs. The present research aimed to fill this gap. We focused on the violent video games debate and showed that people who play video games on a regular basis are more likely to discredit scientific evidence demonstrating detrimental effects of violent video games and to express particularly harsh and critical attitudes toward video games effect research (and the respective researchers) when they identify strongly with the social category of gamers. This effect is at least partly mediated by perceived stigmatization and anger (Studies 1 and 2), and it cannot be explained by gaming habits or prior beliefs about the effects of playing violent video games (Studies 1 and 3).

These findings provide evidence for our assumption that research on the effects of violent video games symbolizes a threat to the social identity of gamers. We can therefore speculate that the harsh and extremely critical evaluations of this research expressed by gamers in Internet forums, social networks, or online discussions (see the quotes at the beginning of this article) represent some sort of “collective action” against science (cf. Leach et al., 2006, 2007; van Zomeren et al., 2008). We believe that this is a general principle that applies not only to research on the effects of violent video games, but to other research areas as well: Whenever scientific evidence has the potential to threaten a group's identity, we believe that the way in which group members react to such evidence can be explained by social identity and collective action processes. On a practical level, our research provides new insights into how scientific evidence is evaluated and how it can be instrumentalized in public debates. This might be especially interesting for debates in which certain groups are in the public focus, such as, for instance, gender groups, and also religious groups or vegetarians. For example, we would hypothesize that strongly identified vegetarians who are confronted with research showing that a vegetarian diet has negative consequences for the environment (e.g., through increased greenhouse gas emissions) would also react more negatively toward such research because they feel that their group's value is threatened.

Limitations and Directions for Future Research

The conclusions we drew from our data are partly based upon correlational evidence. Thus, we cannot fully rule out that a critical attitude toward video games effect research has already had a prior causal effect on people's social identification. Maybe those participants who agreed being identified with gamers are those who had critically evaluated social scientific research in the past. One way to test the assumed causal effect of identification on critical attitudes toward media violence research more directly would be to experimentally manipulate gamers' level of identification with their in-group (e.g., Jetten et al., 1997; Leonardelli & Brewer, 2001). Future research should use such experimental paradigms in order to establish the hypothesized causal link of identification more strictly.

Our findings provide first evidence for the influence of social identity processes, in our case, social identification with a group that has been targeted by social scientific research, on individual attitudes toward science. We will now discuss two possible psychological mechanisms for this effect, and we hope that this discussion will stimulate further research on this topic. These two mechanisms are as follows: (1) normative influences by one's in-group and (2) epistemic biases operating on the level of text comprehension.

Science Criticism as Group Norm

People are more likely to adhere to group norms when the group's positive value has been threatened. For example, when group norms are salient, people who strongly identify with their group act more in accordance with salient norms (Jetten et al., 1997). In our case, strongly identified gamers may believe that a “gamer norm” would prescribe to discredit all kinds of potentially threatening research (“science-hostile group norm”), and their fundamentally critical style of evaluating the “harmful” study might reflect a kind of group norm-consistent behavior. Recently, Sjöström, Sowka, Gollwitzer, Klimmt, and Rothmund (2013) theorized about a “hostile science effect” as a counterpart to the “hostile media effect.” A science-hostile group norm would lend itself to explain such a “hostile science effect.” The assumption that devaluing scientific research reflects a group norm could be tested either by measuring inferred group norms among strongly identified gamers or by manipulating group norms experimentally.

Cognitive Mechanisms of Group-based Science Discrediting

A somewhat related question that future research should address is to what extent the discrediting effect that we found in our studies is based on an intentional, deliberative cognitive process (such as following a group norm) or rather an automatic process that operates on a perceptual level. Regarding the latter perspective, one could assume that strongly identified gamers read a threatening scientific study in a different fashion than weakly identified gamers do. One way to investigate whether the effect of social identification on evaluative biases already operates on a perceptual level would be to adopt existing paradigms from text comprehension research (e.g., Richter, Schroeder, & Wöhrmann, 2009). Richter and colleagues found that participants with strong prior background knowledge rejected false assertion as efficiently as they verified true assertions even when the assertions were learned under additional cognitive load. According to Richter et al. (2009), this demonstrates an automatic validation process that protects mental representations from being contaminated by false and inaccurate information. If our effects are based on an automatic process, strongly identified gamers should reject threatening information (e.g., “Violent video games can lead to increased aggressiveness”) as quickly and automatically as objectively “false” information (e.g., “Violent video games can lead to increased body height”), even under cognitive load.

Consequences for a Public Engagement with Science

Our research suggests that social identification with a targeted group affects people's attitudes toward research findings and research programs. We also suggested that expressing these attitudes can be conceptualized as a form of “collective action.” If this assumption is correct, one might wonder about the downstream consequences of such fundamentally critical attitudes toward social scientific research. For example, gamers who evaluate media effects research more negatively might also be more likely to publicly express their opinion about such research and about these researchers; they might even demonstrate against the stigmatization they feel to be exposed to. In June 2009, hundreds of gamers in three German cities protested against being stigmatized by the public debate on video game violence. Their slogan was “We are gamers: Demo for video game culture.” 5

Another form of collective action would be to make use of the opportunities provided by Web 2.0 (Brunsting & Postmes, 2002; Postmes, 2007). Web 2.0 allows an extremely rapid distribution of information (and misinformation) and, furthermore, gives many people unlimited access to immense quantities of information (Lewandowsky, Ecker, Seifert, Schwarz, & Cook, 2012). Thus, it creates a perfect space (especially for computer-affine people) to discuss group-relevant scientific findings. In our case, gamers might advocate their own opinion to a broader public because they fear that a study confirming a violent games effect might lead to a stigmatization of gamers. Besides a motivation to counteract an anticipated stigmatizing effect of a scientific study on the public, blogging, twittering, or posting in social networks about potentially threatening research might also serve a self-affirmative goal (Toma & Hancock, 2013). In social networks, for example, other gamers should be especially interested in negative posts about undesirable research findings because such posts might have an affirming effect on their social identity. Furthermore, strongly identified gamers should also be more likely to react toward these negative comments with even harsher science-discrediting comments themselves, affirming the positive identity of themselves and other gamers. This might result in a positive feedback loop, which is often observed in social networks or Web forums. It would therefore be interesting to test these predictions in a simulated online environment in which people have the possibility to post comments and to react directly toward comments made by other participants.

Conclusion

The present research sheds light on why gamers may react so critically toward the research on the effects of violent video games. These findings emphasize the role of group identification, perceived stigmatization, and emotions in people's engagement with science. Theoretically, our research shows that the social identity approach can be useful to describe and explain harsh negative evaluations of scientific findings and entire research programs, as illustrated in the quotes at the beginning of this article. On a more applied level, our results emphasize the role of identity concerns in polarized public debates in which agents use arguments backed up by scientific studies, and it corroborates the importance of social identity processes for motivated reasoning and science communication.

ACKNOWLEDGEMENTS

This research was supported by a German Research Foundation (Deutsche Forschungsgemeinschaft; no. GO 1674/2-1) grant to the second author. We thank Johanna Noemi Kues and Lara Ditrich for their help in conducting the studies.

  1. 1

    Gamers' comments on the question of whether violent video games contribute to youth violence. Retrieved 23 January 2013, from http://videogames.procon.org

  2. 2

    See https://secure.videogamevoters.org or http://www.digitale-generation.de (retrieved 23 January 2013).

  3. 3

    When individuals who failed the attention tests were included in the analysis, the Condition × Identification interaction effect neither predicted perceived stigmatization (p = .48) nor anger about the study (p = .79). Furthermore, anger about the study did not mediate the conditional effect of condition (i.e., study type), neither among weakly (95% CI [−0.08, 0.19]) nor among strongly identified gamers (95% CI [−0.13, 0.16]). However, perceived stigmatization still mediated the conditional effect of condition (i.e., study type) among strongly identified gamers (95% CI [0.03, 0.27]), but not among weakly identified gamers (95% CI [−0.01, 0.19]).

  4. 4

    When individuals who failed the attention tests were included in the analysis, the pattern of results remained the same.

  5. 5

    See http://www.wirsindgamer.de (retrieved 8 November 2013).

APPENDIX

ENGLISH TRANSLATION OF THE DESCRIPTION OF RESEARCH ON THE EFFECTS OF VIOLENT VIDEO GAMES IN STUDY 1

The following part asks for your evaluation of a field of research, namely the research on the effects of violent video games.

Research on the effects of violent video games aims to investigate short-term as well as long-term impacts using either experiments or surveys.

Experiments might look like this: First, people are randomly assigned (e.g., by the flip of a coin) to play a video game with either violent or non-violent contents for about 20 minutes. Afterwards, participants' tendency to behave aggressively is measured. If those who played a violent video game are, on average, more aggressive than those who played the non-violent game, the result indicates an aggression-enhancing effect of the game.

Surveys might look like this: People are asked about their regular consumption of violent video games and about their aggressiveness twice with, for example, an interval of one year in-between the two surveys. Statistical procedures are used to assess whether consuming violent video games influences aggressiveness over time.

Ancillary