• Nie Znaleziono Wyników

54

be needed alongside negative contact to achieve prejudice reduction as compared to other forms of contact.

55

respondents declared they played games 5 to 10 hours a week (56.6% of the sample), and the rest declared they play more. The information on the average gameplay time did not include the kind of video games panellists played (e.g. whether these were console or PC games or less immersive mobile games).

Experimental manipulation

As in the previous study, experimental manipulation consisted of playing the research game (see the description of study 2 for more details). The game was played in the Polish language. Participants were randomly assigned to play in one of two experimental conditions, identical to the previous study.

These were:

1. The Diversity condition in which every NPC on the map was randomly generated as White or Black with an equal probability for both races so that the players’ team and rivals’ teams consisted of Black and White characters in roughly equal proportions. Thus, the condition featured both positive and negative intergroup contact in roughly the same amounts.

2. The All White condition in which all of the NPCs in the game were white and the game featured no intergroup contact. Players in the All White condition served as a control group.

Measures

Behavioural measure. A behavioural measure of the outcomes of intergroup contact was added to the game. For each participant the measure was applied twice, before and after the game, to accurately assess the change that occurred during gameplay. At the very beginning of each playthrough, players were informed that they need to hire delivery men and regional managers for their chain of restaurants in the game. They were presented with a set of 24 characters from which they chose a set of 10 delivery men and another set of 16 characters from which they chose 5 managers (see figure 6).

Each character was displayed as a face (Black or White) with a name and a set of four statistics (labelled: “fast”, “knows the city”, “likeable”, “helpful” for delivery men and “fast”, “accurate”, “fair”

and “liked” for regional managers). The names for White characters were chosen from the most common names in Poland and the names for Black characters were chosen from the most common

56

names given to children by Black parents in Europe. Every time the choice was activated, within each race faces were randomly assigned a name (from a predefined list) and statistics (from predefined sets). The characters were displayed in random order. Statistics were designed in such a way that during every single choice task almost all the characters had the same sum of points and for every Black candidate there was a White candidate with the same points distribution among traits).

At the end of the game, after players beat their opponents, they were informed additional staff was needed to cater to the restaurant chain they built and the task of hiring was repeated. The counts of Black NPCs hired before and after the game were significantly correlated (r = .356, p <.000). In the analysis, the number of Black characters hired before and after playing were compared.

Attention check. The behavioural measure in the experiment included attention checks. Two characters were added with considerably fewer skill points as compared to the rest of the candidates.

Figure 6. The choice of regional managers for the team before the game as part of a behavioural measure.

57

People who chose them could have done so judging characters based on their faces and names only.

However, statistic-based choices of companions and allies are fairly common in games so there was also a considerable chance that people who chose much weaker characters when they had many alternatives available made their choice at random, without paying attention to the characters.

Therefore, people who chose the very weak hires were excluded from the analysis.

Attitude toward Black people was measured with the Polish version of the social distance scale, the same as in the first experiment. The scale was coded so that higher scores meant higher acceptance of Black people (lower social distance). The scale proved reliable with Cronbach’s α = .919 (M = 14.49, SD = 3.54).

Results

Firstly, the results of the explicit attitude measure were analysed. A factorial analysis of variance was conducted to determine whether social distance toward Black people (dependent variable) differed in the two experimental conditions (independent variable). Demographic information (age, gender, size of the city of residence) were entered as covariates and the average amount of time spent playing games per week was included as a random factor. The experimental manipulation was not a significant predictor of social distance (F(1, 122) = 50.38, p = .201). Therefore, the second experiment did not replicate the influence of contact with out-group NPCs on social distance toward Black people.

Secondly, the scores on the behavioural measure were compared. To determine whether the experimental manipulation influenced behaviour, a repeated measures analysis of variance was conducted in a 2 (pre- and post-game Black hires) x 2 (experimental condition) design, with counts of Black characters respondents hired before and after playing the game as a within subject factor and the experimental condition as a between subjects factor. Again, the demographic data was entered as covariates.

58

The results showed that the experimental manipulation significantly influenced the behavioural measure (F(1,123) = 4.11, p = .045, η2 = .032). Contrast tests revealed that respondents in the two experimental conditions did not differ in their choices before the game (F(1,123) < .00, p = .985), but after the gameplay session, participants in the Diversity condition hired significantly more Black characters (F(1,123) = 5.501, p = .021, η2 = .043) than participants playing in the All White condition (see chart 2). The number of Black characters hired before and after the game by the participants in the All White condition did not differ significantly (F(1,123) = 0.76, p = .385). Respondents who played in the Diversity condition hired significantly more Black characters after the game (F(1,123) = 4.11, p = .045, η2 = .032) than before the game. Therefore, the behavioural measure of attitudes was congruent with the findings of the first experiment, demonstrating the positive impact of playing in a diverse game world and experiencing intergroup contact with Black NPCs on behaviour towards Black characters.

Figure 7. Average number of hired Black characters before and after the game in the two experimental conditions. The error bars represent the confidence intervals.

59

Discussion

The results of the experiment confirmed the findings of the previous one only partially. In the second study participants who experienced intergroup contact with Black NPCs (the Diversity condition) had more positive attitudes towards Black people after the experiment as compared to participants who did not experience such contact in the game. The third study did not confirm this finding as the groups did not differ significantly in their attitude measure scores. However, the results of the behavioural measure were congruent with the findings from the previous experiment. Participants who experienced intergroup contact with Black NPCs hired more Black characters after the game than players who experienced no such contact.

The main thing to consider while analysing the discrepancy between the results of the social distance towards Black people in the first and second experiments is the power of these two studies. For both experiments the power was low. Power calculated for the second experiment (Rosner, 2011) – assuming the means of the social distance scale from the first experiment2 and the standard deviation from the entire sample3 – would only be 42%. The observed power in the results for the experimental condition impacting attitudes was .192 (19.2%). Therefore the chances of detecting a significant difference (if such a difference occurred) were extremely low. Accepting the null hypothesis that intergroup contact with NPCs does not influence social distance based on this single study may not be reasonable, especially as the results of the previous experiment suggest otherwise (with a p=.048 and therefore a much lower probability of a mistaken conclusion).

The second thing to consider is the difference in subject pools in these two experiments and its implications. The second experiment was conducted on volunteers that did not receive any compensation for their time and effort. The third was conducted on panellists, paid for participation.

Therefore, for anonymous volunteers there was less of an incentive to start playing if they did not anticipate to enjoy the game. Moreover, it was very easy for them to quit the experiment if they weren’t

2 5.18 for the Diversity condition and 4.87 for the All White condition

3 SD = 1.01

60

engaged in the game. The panellists had a financial incentive to play and continue to the end. The cost of dropping out was higher for them as they would lose the reward and the system would note it in their personal statistics used in recruitment for future studies. Thus one might infer that the volunteers who completed the experiment enjoyed the game more and were more engaged in it. In this case, the game itself and the in-game contact with NPCs in particular could have had a stronger impact on them than on the less engaged panellists.

Additionally, although participants in both groups were considered gamers they might have differed in this respect as well. Subjects in the first experiment were people who joined Facebook groups for gamers (where the recruitment took place) and volunteered to play an unknown game. Information about the gaming habits of panellists available to the researcher was limited to the fact they play games several hours per week (including mobile games). It stands to reason that subjects in the first experiment could have been more involved with gaming and used to committing more time and attention to computer games. In consequence, it might have been easier for them to understand the research game and to engage in it than for the panellists. The effects of the attention check in the second experiment support this interpretation. The panel did not have any procedure associated with checking participants’ attention which means that it allowed for lower focus and engagement in studies conducted on the platform. The attention checks used in the experiment filtered out people who either did not understand the game or clicked at random for other reasons, one of which might have been low attention to the game. More than a third of the panellists recruited for the study failed the attention checks (34.5%, 69 out of 200 people). Nine panellists were excluded because they took an excessively long time to beat the game whereas there were only 4 such people among the volunteers. Even after the exclusion panellists needed significantly more time to beat the game4. This further indicates lower engagement or lower understanding of the game among the panellists. All of the above arguments suggest that intergroup contact with NPCs could have impacted panellists’

4 F(1,281)= 47.55, p<.000, panellists took 85.1 minutes on average with SD = 33.5 and the volunteers needed 61.9 minutes SD = 22.5

61

attitudes to a lesser extent due to their lesser engagement in the game, lesser enjoyment (less positive emotions) or less attention given to the game.

The behavioural measure did confirm the positive impact of intergroup contact with NPCs found in the previous study. In study 3 the positive outcome was not seen in the explicit measure of prejudice towards Black people but was observed in behaviour. This discrepancy could stem from the nature of the measures themselves. Explicit measures require respondents’ willingness to report their attitude as well as their ability to judge the attitude accurately which behavioural measures do not necessitate.

(Olson, Zabel, 2016). This might indicate that the behavioural measure was more sensitive and thus allowed for the observation of change. Indeed, whereas behaviour was shown to correlate with implicit measures of prejudice, results for explicit measures are mixed (McConnell, Leibold, 2001).

Acknowledging that the correlation between attitudes and behaviour varies greatly in general and depends on multiple factors, Fazio (1990) proposed the MODE model in which he distinguishes spontaneous and deliberative processes as two modes linking attitudes to behaviour. The deliberative mode requires both motivation as well as opportunity to consider judgements or choices for the attitude to affect behaviour. Testing the implications of the model, Dovidio (1997) found the link between implicit prejudice and spontaneous behaviour and between explicit measures of attitude on deliberative judgements. In study 3, the judgements made by participants in the hiring task could have been deliberative, as participants had the task of considering the statistics of the candidates.

Their time for making the decision was unrestricted, providing an opportunity for a deliberative process. However, the scores on the hiring task (the behavioural measure) were not significantly related to the scores of the explicit measure. This might mean that participants in the second experiment did not have sufficient motivation for the deliberative process and based their decisions about whom to hire more on implicit attitudes and situational cues (experiences from the game) than on their explicit attitudes towards Black people (seen in the social distance scale). This would explain the discrepancy between the two measures. If the deliberative process did not take place in the third experiment, it could also contribute to the lack of observable differences in explicit attitudes between the two experimental conditions.

62

The differences between the two experiments could also be explained by the object towards which participants’ attitude was changed. According to the principle of compatibility (Ajzen & Fishbein, 1977) attitudes predict behaviour better the more the two are matched in terms of focus (e.g. attitudes towards “Black people as a political group” will better predict judgements about policy changes than it will impact reactions towards individual Black people met in a social situation). In the behavioural measurement, participants rated characters they were about to hire. Thus they judged the same group they were in contact with during the game. In the social distance scale subjects rated their attitude towards real-world Black people. Therefore, if the experimental manipulation changed participants’

attitudes towards Black NPCs it was more likely to show in the behavioural measure, which had the same object of focus (Black NPCs) than it was in the explicit measure scores (which had a focus on a slightly different group of Black people in the real world). According to the intergroup contact research, cross-group contact improves attitudes towards the interaction partner and this attitude change reliably generalises for the entire out-group (e.g. Pettigrew & Tropp, 2005). The results may indicate that in the second study the attitudes towards Black NPCs improved and the effect generalised onto Black people in the real world whereas in the third study the generalisation from individual NPCs to Black NPCs in general was observed with a behavioural measure, but further generalisation to real-world people was not. The explanation for such outcomes could stem from the same arguments made for the discrepancy between the two measures described before. There are reasons to believe that participants in the second experiment were less focused on the game (see previous paragraphs) which could impact the cognitive processes that are crucial for the generalisation of positive effects to the attitudes towards the entire out-group (Dhont et al., 2011). The affective mediators, which were shown to have an even stronger influence on outcomes (Paolini et al., 2006), could have been similarly worse in the third study if participants were less engaged in the game and enjoyed it less than volunteers in the previous experiment. Both cognitive and affective differences between gameplay experiences of participants in the two experiments may provide an explanation as to why in the third study the generalisation effects were weaker and were observable only for Black characters, but not for the entire real-world out-group as was the case in the previous study.

63

Powiązane dokumenty