On Attitude Extremity and Involvement: Applying the Catastrophe Model
Bachelor thesis Social Psychology Isabel Barnard, 10798234 Supervisor: Jonas Dalege University of Amsterdam Date of submission: 2-06-‘17 Word count: 3629
Abstract
Extreme opinions are widespread these days. The Catastrophe model implies that attitude extremity is more likely to occur in people highly involved in a topic. The
current study examined this relationship between involvement and attitude extremity. It was attempted to decrease involvement by having participants reading trivializing texts on two topics, climate change and refugee policy. Afterwards, attitude extremity was measured. Results showed that involvement only decreased in participants that read the text on climate change. Subsequently, attitude extremity of these participants showed a small decrease. Participants that read the text on refugee policy showed a smaller and opposing pattern. Suggestions for further research were made.
On Involvement, Information and Attitudes: Applying the Catastrophe Model Every year, Amnesty International publishes a report, documenting the state of human rights in 159 countries. In their report on 2016, Amnesty emphasizes the upcoming populist movements and their dividing messages ‘bursting into the
mainstream’ (Amnesty International, 2017). Mentioned as targets of populist politicians are, among others: political elites, the EU and immigration policies. Many, sometimes oversimplified, discussions on these topics have shown a huge division of people, with many people having opinions on both ends of the spectrum. This division is hardly compatible with the tolerance that is needed for constructive dialogue. With diverse social media platforms at hand, everyone is able to give their opinion and generally people have proved not to be shy in doing so. People seem to be prone to take clear sides in every discussion and expect everyone else to do so as well. This seems to have
resulted in an extremity of attitudes on several topics; sometimes nuance appears to be missing these days. In the current study the role of involvement within attitude
extremity will be examined.
In social psychology, attitude extremity is a much-discussed subject. In an influential study, Lord and Ross (1979) described two important findings in this research area. In this study, research findings on capital punishment, containing both evidence in favor and against it, were shown to both proponents and opponents of the death penalty. First, the researchers found that subjects evaluated the research findings anything but neutral; they rated research that agreed with their preexisting position more positively. Opponents of capital punishment preferred methods of which findings supported their efficacy, while they rated the exact same methods as inferior when findings did not support efficacy. This phenomenon they called biased assimilation.
Besides this, the researchers also found that subjects’ attitudes had become more extreme after evaluating the research findings; initial proponents becoming more in favor of the capital punishment and opponents becoming more against it. This
phenomenon can be called attitude polarization: attitudes becoming more extreme, in the direction of the initial attitude.
Other studies showed more features of attitude polarization. For example, one study showed that the more time participants spend on thinking about a political issue, the greater the tendency to attitude polarization (Tesser & Conlee, 1975). Also, it was found that the degree of attitude polarization is dependent on the extremity of the initial attitude (Miller, Bane & Dowd, 1993). Participants with more extreme initial attitudes showed a stronger attitude polarization than did participants with moderate initial attitudes.
Various theoretical models were created to describe changes in attitudes, such as the Catastrophe Model, which tries to explain sudden transitions in attitudes (Van der Maas, Kolstein en Van der Pligt, 2003). The Catastrophe Model is a three dimensional model representing a relationship between attitude, information and involvement. Specifically, the association between attitude and information, in this model, is
dependent on involvement. In a situation of low involvement, information and attitude show a simple, close to linear relation. In this case, receiving negative information will lead to a more negative attitude and receiving positive information will lead to a more positive attitude. In a situation of high involvement, however, there is a more complex relation between attitude and information. In this case, only two states exist: strong positive attitude and strong negative attitude. In between these two areas there is an inaccessible transition area. When presented with enough information, a sudden jump can occur from a strongly positive attitude to a strongly negative one (or the other way
around). However, according to this model, people high in involvement, will never show a moderate attitude. This means that when people are high involved in a topic, a gap will occur between people on both ends of the spectrum.
According to the Catastrophe Model, involvement plays a big part in the attitude polarization. As to this model, attitude polarization is most likely to occur when people are highly involved in a matter. Several studies indicate that this, indeed, might be the case. For instance, the magnitude of attitude polarization, in groups, is found to be dependent on the topic of the issue. One study found that the less popular and
interesting a topic was rated, the more it would depolarize (Vinokur & Burnstein, 1978). In a similar study, it was found that groups of people with specific interest in the
relevant topic were more certain and more extreme in their view that groups without any interest (Kronberger, Holtz & Wagner, 2012). The latter group remained uncertain and avoided taking sides. An interesting topic, of course, is not necessarily the same as a topic in which a person is highly involved. However, because of a close relatedness of these concepts, these findings seem to support the role of involvement in attitude (de)polarization. These two studies examined involvement in attitude polarization in groups, not necessarily in individuals. Other studies, directed at individual attitudes, however, showed mostly the same results. For example, one study revealed that people high in involvement have a more extreme positive opinion on their preferred
presidential candidate (Brent & Granberg, 1982). In another study involvement was manipulated by telling students that certain proposals were being considered for
implementation next year either at their school (high involvement) or at another school (low involvement) (Liberman & Chaiken, 1996). It was found that issues eliciting high involvement generated more extreme attitude responses than issues eliciting low
involvement. All of the studies mentioned above seem to display the role of involvement in attitude polarization, both in groups and in individuals.
One study on attitude extremity is particularly relevant to the current study. In this study, it was shown that initial attitude extremity is related to the importance of the topic (Harton and Latane, 1997). More importantly, the distribution of attitudes became more bimodal as importance increased. In other words, the more important an topic was for participants, the more they reported having attitudes on both ends of the spectrum. These findings are perfectly in line with the Catastrophe Model mentioned above; apparently, participants high in involvement indeed did not enter the
inaccessible transition state described by the model. In the current study it is attempted to generate the exact opposite effect. Where it was found that attitude distribution became more bimodal as involvement increased, now it is aimed to decrease bimodality – and attitude extremity – by decreasing involvement. Since attitude polarization can have undesirable consequences this might be of way more value. As mentioned before, societal division can lead to intolerance towards people that are different or unfamiliar. Finding ways to reduce attitude polarization, and thereby decreasing societal division, might provide a useful tool for steering society towards a more tolerant orientation.
The current study was designed to test the hypothesis, based on the Catastrophe model, that a decrease of involvement will lead to a decrease in attitude polarization. To manipulate involvement decrease, participants will read trivializing texts on climate change or refugee policy. Attitude extremity will be assessed by means of a
questionnaire. Expected is that participants will have less extreme attitudes on topics on which they have read a trivializing text compared to topics on which they have read a neutral text. Moreover, comparing attitudes on one specific topic, participants having
read a trivializing text on that topic are expected to have less extreme attitudes than participants not having read the trivializing text.
Methods Participants
A total of 100 University of Amsterdam students (69 male) took part in the study in return for so called ‘participation credits’. Their mean age was 20.45 (SD = 2.15). Six participants did not finish the study. Their data was removed before further analysis.
Pilot study
With a small pilot study (N = 16), we measured student’s attitudes on a number of topics. Based on these findings, two topics, showing the most extreme student’s attitudes, were chosen for the current study. These topics were ‘Dutch government investing in counteracting climate change’ and ‘closing border for refugees’.
Materials
A questionnaire was constructed to measure participant’s attitudes on the two topics mentioned above and eight additional topics, for example: ‘anonymous job application’ and ‘increasing arts and culture subsidy’. On each of the topics, three items measured attitude and three items measured involvement. For instance, an item that measured attitude was ‘I find government investments on climate change …’ and an item
measuring involvement was ‘I care about closing boarders for refugees’. Participants gave their answer by moving a slide on different scales. The attitude scales ranged either from ‘negative’ to ‘positive’, ‘bad’ to ‘good’ or ‘against’ to ‘in favor of’. The involvement scales ranged from either ‘not important’ to ‘important’, ‘interesting’ to ‘not interesting’
or ‘disagree’ tot ‘agree’. In total, the attitude questionnaire contained sixty items. The climate involvement subscale of the questionnaire had a high reliability, Cronbach’s α = .90. The refugee involvement subscale of the questionnaire had a Cronbach’s α of .77, which represents a sufficient reliability. The climate extremity subscale of the
questionnaire had a high reliability, Cronbach’s α = .97. The refugee extremity subscale of the questionnaire also had a high reliability, Cronbach’s α = .95.
Four small texts were created, concerning each of the following topics: ‘climate change’ and ‘refugee policy’, ‘re-establishing student funding’ and ‘increasing arts and culture subsidy’. The two latter were purely informative; they held some neutral information relating the topic. The texts on ‘climate change’ and ‘refugee policy’, however, were created to be trivializing, by putting Dutch contribution into the bigger picture. For example, the climate change text showed how much several countries contributed to climate change by their CO2 emissions. With this, we attempted to
decrease participants’ involvement in these topics, without affecting their attitudes directly.
Procedure
The study was conducted online. Participants who chose to carry out our study were directed from a special University of Amsterdam website to our questionnaire on Qualtrics. First, they were instructed to read and sign the informed consent. After doing so, they were directed to the first part of our study: reading the texts. We randomly assigned the participants to one of two conditions. In the climate condition (N = 49), participants read the trivializing text on climate change. Participants in the refugee condition (N = 51) read the trivializing text on refugee policy. In both conditions participants read the neutral texts on ‘re-establishing student funding’ and ‘increasing
arts and culture subsidy’. By adding these neutral texts it was aimed to obscure the purpose of the study. When finished reading, participants continued to the second part of the study: the attitude questionnaire. Participants were asked to truthfully fill in our 60-item questionnaire on both the relevant topics (‘climate change’ and ‘refugee policy) and irrelevant topics that were added to extend the questionnaire and to conceal the purpose of the study. Finally, after a short debriefing, participants were thanked for their participation and got instructions on how to get their participation credits.
Data-analysis
First a manipulation check will be carried out. For each condition, a mean involvement score will be made based on scores for all involvement items on the topics ‘climate change’ and ‘refugee policy’. In doing so, four main scores will be calculated and a comparison can be made both between conditions and within subjects. By executing a two-way mixed ANOVA, a comparison will be made both between (condition) and
within subjects (items). For example, subjects in the refugee condition will have read the trivializing text about refugee policy. Their mean involvement score on refugee items will be compared to the mean score of refugee items of subjects in the climate condition, who have not read the trivializing texts on refugee policy. Also, their mean involvement score on refugee items will also be compared to their own mean score on the climate items.
For our main analysis, a mixed ANOVA will be executed. Identical to the
manipulation check, this analysis will provide a comparison both between and within subjects; the between variable will be condition and the within variable will be item type. First of all, an attitude extremity score will be calculated by subtracting the scale middle from scores on every individual item. After this, for both conditions a mean score
will be calculated for both type of items (refugee and climate). This way, four mean scores will be created and then compared.
Results Manipulation check
Kolmogorov-Smirnov test showed that involvement scores both on climate items, D(100) = 0.136, p < .001, and on refugee items, D(100) = 0,093, p = .034, were
significantly not-normal. Levene’s Test showed that for involvement scores on refugee items error variances were equal, F(1, 98) = .85, p = .358. Moreover, for involvement scores on climate items error variances were unequal, F(1, 98) = 10.27, p = .002. Due to these violations of assumptions, it was chosen to use the Greehouse-Geisser correction. This correction is applied to all test statistics reported below.
There was no significant main effect of item type, F(1, 98) = 0.29, p = .594, η2 =
.003. This means that overall involvement scores on climate and refugee items did not differ from each other. There was, however, a main effect of condition F(1, 98) = 2.41, p = .124, η2 = .024. This means that involvement scores differed between participants of
our two conditions. Figure 1 shows that participants in the climate condition had an overall lower involvement score than participants in the refugee condition. No
significant interaction effect between condition and item type was found, F(1,98) = 1.56, p = .215, η2 = .016 (see figure 1). This indicates that involvement scores on the two item
types were not different for participants in the climate condition and refugee condition. Contrasts showed that scores on climate items where slightly lower for participants in the climate condition than for participants in the refugee condition. This difference, however, was only a trend, F(1, 98) = 3.33, p = .071, η2 = .033. Furthermore, scores on
F(1, 98) = 0.38, p = .542, η2 = .004. These findings show that the climate text lead to a
small involvement decrease, while the refugee text did not seem to have any significant effect.
Figure 1. Mean involvement scores on refugee and climate items for participants in both conditions.
Main analysis
Kolmogorov-Smirnov test showed that attitude extremity scores both on climate items, D(100) = 0.142, p < .001, and on refugee items, D(100) = 0,132, p < .001,
significantly deviated from normal. As with the manipulation check, because of these violation of assumptions it was chosen to use the Greehouse-Geisser correction.
Levene’s Test showed that error variances were equal for attitude extremity scores on climate items, F(1, 98) = 2.21, p = .141, and for attitude extremity scores on refuge items, F(1, 98) = 2.26, p = .136.
There was no main effect of item type, F(1, 98) = 3.47, p = .066, η2 = .034. This
means that for all participants together, attitude extremity scores of climate items did not differ from refugee items. Also, there was no main effect of condition, F(1, 98) = 3.53, p = .063, η2 = .035. This indicates that participants of both conditions did not differ from
each other on their overall attitude extremity. There was no significant interaction effect between item type and condition, F(1,98) = 0.54, p = .466, η2 = .005 (see figure 2). This
indicates that attitude extremity scores on the two item types did were not different for participants in the climate condition and refugee condition. Contrasts revealed that attitude extremity scores on climate items where slightly lower for participants in the climate condition compared to participants in the refugee condition, F(1, 98) = 3.68, p = .058, η2 = .036. Furthermore, attitude extremity scores on refugee items were slightly
higher for participants in the refugee condition compared to participants in de climate condition, but this was a non-significant difference F(1,98) = 1.38, p = .243, η2 = .014.
These results show that reading the text lead to a small attitude extremity decrease for participants in the climate condition and to a marginal increase of attitude extremity for participants in the refugee condition.
Figure 2. Mean attitude extremity scores on refugee and climate items for participants in both conditions.
Discussion
The aim of this study was to examine whether a decrease of involvement would lead to a decrease in attitude extremity. It was attempted to decrease involvement by having participants read trivializing texts. Results on our manipulation check, first of all, showed a main effect of condition. Participants in the refugee condition were more involved in both climate change and refugee policy than participants in the climate condition. There was no main effect of item type; all participants combined showed no
difference between involvement in climate change or refugee policy, that is, participants seemed to be equally involved in both topics. More importantly, results showed that in the climate condition, participants displayed a tendency to lower involvement on climate change and subsequently a slightly lower attitude extremity on this topic than did participants in the refugee condition. Meanwhile, participants in the refugee
condition showed a different pattern. They showed a slight increase of involvement on refugee policy and subsequently also a slight increase on attitude extremity. These results seem to support our hypothesis; although non-significant, they are in the expected direction. As expected by the Catastrophe model, involvement decrease was followed by an attitude extremity decrease and involvement increase was followed by attitude extremity increase.
One big shortcoming of the current study can be pointed out. The manipulation check showed that the trivializing texts failed to have the expected effect on
participants’ involvement. Firstly, involvement in participants in the climate condition only decreased slightly. This indicates that the manipulation might not have been strong enough. People are confronted with information on climate change on a daily basis. Compared to this, our text might just be a minor influence and therefore not decisive in attitude formation. Further research might include presenting trivializing information in a different – high impact – fashion to counteract the omnipresent flood of information presented everyday.
Secondly, involvement in participants in the refugee condition showed a small increase instead of the expected decrease. This means that our ‘trivializing’ text on refugee policy was not trivializing at all, but in fact seems to have emphasized the importance of the problem. Now the question rises how almost exactly the same way of trivialization used on both topics, refugee policy and climate change, could have had
such opposing effects on people. The answer to this might be that the trivializing texts were not as similar as we thought they were. For both topics, it was tried to put Dutch contributions to the relevant problem in to a bigger (worldwide) picture. For climate change, we showed that Dutch pollution is only a small portion of global pollution and therefore the Netherlands is only responsible for a small part of the problem. For
refugee policy, we showed that Dutch accommodation of refugees is only a small portion of global refugee accommodation and therefore the Netherlands is only responsible for a small part of the solution. In other words, in the first case the text focuses on the
problem while in the latter case the text focuses on the solution. This distinction between responsibility for either the problem or the solution may have resulted in the difference in effect on the participants. For future research it is recommended to either chose more related topics that can be trivialized in the same way, or to create trivializing methods specific to each of the topics.
While this study did not provide indisputable confirmation that decreasing involvement reduces attitude extremity, it at least implies that this is the case. It is important that this relationship will be further investigated in the future. Attitude
extremity and polarization can lead to several negative consequences for society, such as intolerance to other people. Therefore, it is important to rapidly make progresses in indicating factors that can reduce these phenomena.
References
Amnesty International. (2017). Amnesty International Report 2016/17: The state of the world’s human rights. Retrieved from:
https://www.amnesty.org/en/latest/research/2017/02/amnesty-international-annual-report-201617/
Brent, E. E., & Granberg, D. (1982). Subjective agreement with the presidential
candidates of 1976 and 1980. Journal of Personality and Social Psychology, 42(3), 393-.
Harton, H. C., & Latane, B. (1997). Information-and thought-induced polarization: The mediating role of involvement in making attitudes extreme. Journal of Social Behavior and Personality, 12(2), 271-299.
Liberman, A., & Chaiken, S. (1996). The direct effect of personal relevance on attitudes. Personality and Social Psychology Bulletin, 22(3), 269-279.
Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of personality and social psychology, 37(11), 2098-2109.
Miller, A. G., McHoskey, J. W., Bane, C. M., & Dowd, T. G. (1993). The attitude polarization phenomenon: Role of response measure, attitude extremity, and behavioral consequences of reported attitude change. Journal of Personality and Social Psychology, 64(4), 561-574.
Tesser, A., & Conlee, M. C. (1975). Some effects of time and thought on attitude polarization. Journal of Personality and Social Psychology, 31(2), 262-270. Van der Maas, H. L., Kolstein, R., & Van der Pligt, J. (2003). Sudden transitions in
