This is an open access article distributed under the terms of the Creative Commons
Attribution License (Attitude change is one of the most challenging endeavors for social and political psychologists, because of the practical implications of the interventions in promoting positive societal change. As part of this endeavor, recently we developed a new approach and method termed paradoxical thinking (Bar-Tal, Hameiri, & Halperin, 2019; Hameiri, Bar-Tal, & Halperin, 2019; see also Swann, Pelham, & Chidester, 1988). It is based on using a message that is consistent with the targeted audience's held beliefs and attitudes, but is provided in an amplified, exaggerated or even absurd manner. Instead of eliciting inconsistency using counter-attitudinal information, or views, the paradoxical thinking message extrapolates extreme and absurd conclusions from the views held by the message recipient, which are meant to induce a deliberative examination of these beliefs and attitudes. This process may lead to the realization that something within the individual’s beliefs is perhaps wrong, nonsensical, improbable, unacceptable or strange, which might stimulate unfreezing of these beliefs and eventually their moderation. Indeed, experimental studies in the laboratory and in the field confirmed unequivocally the hypotheses derived from the conceptual framework (Hameiri, Nabet, Bar-Tal, & Halperin, 2018; Hameiri, Porat, Bar-Tal, Bieler, & Halperin, 2014; Hameiri, Porat, Bar-Tal, & Halperin, 2016).
As a next step, the present study aims to provide additional illumination of the framework’s underlying psychological mechanism. Thus far, in research on paradoxical thinking, which examined interventions in the context of the Israeli-Palestinian conflict, we found indications that the paradoxical thinking interventions were effective because of the following psychological mechanisms: (1) a surprised reaction, i.e., being surprised, and even shocked or disturbed by the absurdity of the messages (see Frankl, 1975); (2) a sense of threat to the recipients’ identities; (3) lower levels of disagreement with the message content, arguably because it is consistent with the views of the message recipients, as it was tested vis-à-vis inconsistent messages; and that the identity threat, surprise, and lower levels of disagreement, in turn, (4) challenge the validity of the recipient’s held views leading to their unfreezing (Hameiri et al., 2018).
Social Judgment Theory and Paradoxical Thinking
We argue that social judgment theory, proposed by Sherif and Hovland (1961), illuminates our conception and supplies empirical observations. The authors describe how people relate their personal attitudes on issues to messages that they encounter. According to the theory, aside from having their personal attitude, individuals hold latitudes of what they think about attitudes expressed in messages: how acceptable or unacceptable these messages are. Sherif and Hovland (1961) saw an attitude as an amalgam of three latitudes. First, there is the latitude of acceptance, which is the range of attitudes that a person sees as reasonable or worthy of consideration. Second, there is the latitude of rejection, which is the range of attitudes that a person sees as unreasonable or objectionable. And, finally, there is the latitude of non-commitment, which is the range of attitudes that a person sees as neither acceptable nor questionable. Another factor in the theory is ego involvement, which denotes the importance or centrality of an issue to a person's life, or in other words, how extreme or morally convicted a person is with regard to an issue, often demonstrated by membership in a group with a known stand. Religion, politics, and family are examples of issues that typically result in highly involved attitudes as they contribute to one's self-identity.
People who are highly involved in an issue and hold frequent and extreme opinions, have a wide latitude of rejection because they already have a strong opinion and are usually unwilling to change it. They also have a more restricted latitude of acceptance and thus it is harder to persuade them with inconsistent information. In contrast, individuals who care less about an issue, or have a smaller ego involvement, are likely to have a wide latitude of acceptance and as a result are more likely to accept new opinions about an issue. Consequently, when a message does not diverge greatly from the latitude of acceptance, it will be accepted (i.e., assimilated), and the person will shift towards the position expressed in the message (e.g., Atkins, Deaux, & Bieri, 1967; Peterson & Koulack, 1969). Yet, there are some indications that this is true only for individuals with a wide latitude of acceptance, while for individuals with a narrower latitude, inconsistent messages are not persuasive in any case (see Eagly & Telaak, 1972).
More importantly, when the message advocates a position far removed from the latitude of acceptance that falls within the latitude of rejection, a contrast effect occurs and the communication is evaluated as "unfair", "biased", or "farfetched". In these cases, we suggest that a boomerang effect will take place. That is, the person’s attitude is likely to shift away from the attitude expressed in the message by moderating it. The person does not want to be perceived as holding such an absurd belief or to be associated with a group that holds it as its group belief (Bar-Tal, 1990). Indeed, a study about prohibition, conducted by Hovland, Harvey, and Sherif (1957) demonstrated this effect of moderation that occurred among the participating individuals with messages that fell within the latitude of rejection, due to the discrepancy from the held beliefs. In previous research on the paradoxical thinking effects, we found that when the message was inconsistent with the held beliefs and attitudes and fell within the latitude of rejection, the person with high involvement ignored the content, activating automatic thinking and defense mechanisms (Hameiri et al., 2018; see also Eagly & Telaak, 1972; Swann et al., 1988). In the present research we examine for the first time the hypothesis that when a message is consistent, but of very extreme and absurd nature, individuals automatically reject it, without any further processing of its content.
Tying the concepts of latitudes of acceptance and rejection to the paradoxical thinking conceptual framework, and extending the arguments in Hameiri et al. (2018, 2019), we suggest that one of the reasons paradoxical thinking interventions are effective is because even though the messages fall within the latitude of rejection, they are processed, as they are consistent with the held beliefs and attitudes of the message recipients. Individuals consider the messages and as they fall within their latitude of rejection, a contrast effect may take place. That is, participants moderate their attitudes to move away from the attitude expressed in the messages. We argue that this is in contrast to what happens when individuals are exposed to very extreme absurd messages that are senseless, such that these messages are rejected automatically. All of the above suggests, but was never empirically examined, that if a paradoxical thinking message is too extreme in its content and sounds too absurd, it will also be rejected or dismissed (see Hameiri et al., 2019). Thus, in line with the thinking of Sherif, Sherif, and Nebergall (1965), we hypothesize that each individual has a ‘sweet spot’ in which a paradoxical thinking message is the most effective.
The Present Research
The aim of the present research, therefore, is to examine whether, first, messages can be consistent, but too extreme and too absurd, so they would not lead to a contrast effect, as they would be automatically rejected. This would be indicated by strong general disagreement, and consequently would not be effective in leading to unfreezing. Second, whether there is indeed a ‘sweet spot’, such that for different individuals, based on their own initial views, paradoxical thinking messages with varying degrees of extremity would be more or less effective. To test these hypotheses, contrary to previous research that manipulated the messages based on their consistency with the recipients’ held views (e.g., Hameiri et al., 2018; Knab & Steffens, 2019; Swann et al., 1988), in the present research all of the messages were consistent with the recipients’ views, but varied in their extremity and absurdity of the content and language used.
Furthermore, to expand the paradoxical thinking conceptual framework’s external validity, which so far was almost exclusively examined in the context of the Israeli-Palestinian conflict (see Hameiri et al., 2019), we conducted this research examining negative beliefs and attitudes toward refugees and asylum seekers in Israel. This issue has become a major global concern in the past few years, and therefore has been receiving increased attention by social psychologists in developing interventions and policies to help refugees and asylum seekers resettle in new host countries—for example, by changing the views of members of these host countries (see Esses, Hamilton, & Gaucher, 2017).
Israel absorbed through the years two categories of international migrants: economic migrants who entered the state either legally or illegally with the goal of finding jobs, and asylum seekers who entered the state with the goal of finding a safe place from persecutions in their home countries. Although nobody knows their exact number, the rough estimation is that in Israel there are more than 80,000 economic migrants and more than 40,000 asylum seekers (Population and Immigration Authority, 2018). The Israeli government has carried a clear policy against asylum seekers trying to concentrate them in a camp and eventually to deport them to a third country (Tartakovsky & Walsh, 2016). This policy polarized the Israeli-Jewish society: While many express strong anti-refugee feelings, blaming the migrants for increased crime and deteriorating conditions in several neighborhoods around Israel; others, viewing the issue from a humanitarian and moral obligations viewpoint, object to the governmental policy and even help the refugees in different forms (Hochman, 2015; Nuttman-Shwartz & Shinar Levanon, 2019).
One of the main reasons the government’s policy regarding asylum seekers has polarized the Israeli-Jewish society might have to do with the Jewish people’s history as victims of the Holocaust (Klar, 2019; Klar, Schori-Eyal, & Klar, 2013). Previous research has shown that at least some members of groups that were victimized feel that it is a moral imperative, or mandate, for their group to help other victimized groups (see Klar et al., 2013; Vollhardt, 2015; Warner, Wohl, & Branscombe, 2014). Specifically, in Klar et al.'s (2013) analysis of the Israeli-Jewish society, they argue that one of the moral mandates that Israeli-Jews learned from the Holocaust is to never be a passive bystander, which historically was manifested in instances in which Israel allowed the entrance of refugees that were refused in other countries. This corresponds with the findings of Warner et al. (2014) that showed that asking Jewish participants to consider the lessons of the Holocaust for Jews (rather than for Germans), increased their perceived moral obligation to help other (non-adversarial) victimized groups, such as refugees from Darfur. However, Klar et al. (2013) argue, that this lesson clashes with another moral mandate that Israeli-Jews learned from the Holocaust, namely to never be a passive victim again.
While these moral mandates learned from the Holocaust mostly have been examined with regard to Israel’s conflict with the neighboring Arab countries, and particularly, the Palestinians (e.g., Rosler & Branscombe, 2020; Schori-Eyal, Klar, Roccas, & McNeill, 2017; Warner et al., 2014), they were also found to have important implications for Israeli Jews’ attitudes toward refugees and asylum seekers. Specifically, the particularistic moral mandate (i.e., never again be a passive victim) predicted more hostile treatment of the asylum seekers; while the universalistic moral mandate (i.e., never again be a passive bystander) predicted less hostile treatment (Ariely, 2019). Put simply, the issue of refugees and asylum seekers is highly contested in Israel, and each side holds views that are based on what are perceived by many as moral mandates (Klar et al., 2013; Warner et al., 2014).
Using the terminology of the extensive literature on moral conviction (e.g., Skitka, Bauman, & Sargis, 2005; Skitka, Washburn, & Carsel, 2015), many people on each side of this debate are morally convicted, i.e., they hold their views toward refugees and asylum seekers as “a strong and absolute belief that… is right or wrong, moral or immoral” (Skitka et al., 2005, p. 896). Research on moral conviction has shown that it predicts, above and beyond other measures of attitude strength or extremity, less willingness to compromise on an issue at stake, more resistance and intolerance to alternative views, and is in general a strong predictor for attitude extremity, accessibility, and increased attitude-behavioral intention consistency (Luttrell, Petty, Briñol, & Wagner, 2016; Ryan, 2017; Skitka et al., 2015; Van Bavel, Packer, Haas, & Cunningham, 2012). In the terminology of the social judgment theory, we argue that moral conviction would be a good indicator for the range and width of positions that would be tolerable by message recipients (Eagly & Telaak, 1972). Specifically, for the purposes of the present study, we argue that moral conviction is a good indicator for the extent to which recipients will be willing to thoroughly process different levels of extremity of paradoxical thinking messages. Thus, based on the predictive characteristics and consequences of moral conviction, and the fact that views regarding African refugees and asylum seekers in Israel are moralized to varying degrees by Israeli Jews, in the present study we use participants’ moral conviction regarding African asylum seekers in Israel as our moderator (e.g., Skitka et al., 2005; Verkuyten, Altabatabaei, & Nooitgedagt, 2018).
Focusing on individuals who hold anti-refugee views, we hypothesized, first, that for those holding these views with low moral conviction, a rather moderate paradoxical thinking message would lead to more unfreezing of their views; while for high morally convicted participants, a more extreme version of the message would lead to unfreezing. Second, we hypothesized that a very extreme message would not lead to unfreezing with any of the participants, regardless of their moral conviction. And third, we hypothesized that this would be the case as participants would show much more resistance to the very extreme and absurd message, compared to all the other, more moderate messages.
Method
Participants
210 Jewish Israeli participants completed the questionnaires through the online surveying firm, Midgam Project, which is an opt-in panel that includes over 50,000 panelists aged 17 years and older within Israel. In exchange for participation, participants received 8 ILS (equivalent to US$2.20). We removed nine participants that had extreme scores in any of our dependent variables, with residuals over 3 × interquartile range (IQR) outside the IRQ range (see Cook, 1979). The final sample included 201 participants (Mage = 37.16, SDage = 11.91; 125 men, 76 women; 40.3% were secular, 20.4% considered themselves as observant, and 39.3% were religious; 63.8% had at least a college degree, 17.4% had some college education, and 17.9% had 12 or fewer years of schooling). Importantly, the pattern of results remains identical when analyzing the results with the full sample. Power calculations indicated that our sample size was big enough to detect with .80 power a medium effect size (Cohen’s f2= .24) for the condition main effect; and a rather small effect size (Cohen’s f2 = .04) for our key interaction term (i.e., between the Texts 1 and 3 comparison and our moderator, moral conviction; see below).
These participants were a sub-sample of 787 (Mage = 37.38, SDage = 12.42; 388 men, 399 women) Israeli Jews that completed the first phase of the study. Specifically, only 280 (35.6%) participants were found eligible for the second phase based on their (dis)agreement with a pro-refugee opinion editorial they read in the first phase of the study (see Procedure). In terms of political orientation, the sample included 7.0% extreme rightists, 59.2% rightists, 19.4% moderate rightists, 12.4% centrists, and 2.0% moderate leftists. Thus, as expected from the procedure, in which we only sampled participants who disagreed with a pro-refugee opinion editorial, compared with the adult Jewish Israeli population (Central Bureau of Statistics, 2013; Hermann, Anabi, Heller, & Omar, 2018), our sample was more religious and heavily skewed to the right. However, this skewed sample is comparable to previous studies (Hameiri et al., 2016, 2018) that examined the paradoxical thinking conceptual framework that targeted individuals who were more extreme and adamant in their hawkish views with regards to the Israeli-Palestinian conflict (see Hameiri et al., 2019). Finally, it should be noted that the results remain identical when analyzing the data only with the 172 rightist participants.
Procedure
The current study was presented as if it was ostensibly intended to examine the effect of comments and letters to the editor on the political discourse in Israel, due to the increased tendency of news outlets to cancel these channels of communication on online news websites. In the first phase of the study, participants were asked to read a real opinion editorial piece published in the Hebrew edition of Haaretz, an Israeli newspaper that is considered liberal. They were informed that this was the first phase of the study, and that they would be contacted again within the next few days for the second part, in which they would be asked to read a second opinion editorial that commented on the first editorial. The first opinion editorial argued that the National Health Insurance Law in Israel should be applied to cover the refugees’ health needs, not only for humane, just, and moral reasons, but also because it was prescribed by law (Haaretz Editorial, 2016). Then, participants were asked to indicate the degree of their agreement with the editorial’s “main argument that asylum seekers should be provided with health care.” A week later, those who disagreed were invited to take part in the second part of the study. They were asked to read a second opinion editorial that was ostensibly written by a rightist Parliament member as a response to the original editorial and published in Israel Hayom, which is considered a news outlet in Hebrew with center-right political alignment. After reading the editorial, participants were asked to answer two attention verification questions. Those who answered one of the questions incorrectly (14 participants overall) were given a second chance and were directed to the beginning of the segment.
Participants were randomly assigned to one of four conditions that differed in the extremity of the message. All texts were approximately a page-long and were made to appear as if they were an opinion editorial from Israel Hayom. They all had the same underlying message, consistent with the participants’ own views, that Israel should not grant free health care to the refugees, but rather expel them from Israel. The difference between the conditions was in the way the text was worded, i.e., the intensity of the language (rhetoric) and the extremity of the content (semantic). The first two paragraphs were rather identical and only included subtle differences (e.g., the use of a more benign word to describe asylum seekers as those who seek asylum in Text 1; then in Texts 2 and 3 they are described as those who want asylum; finally, in Text 4, they are described as those who demand asylum). The third and final paragraph, while sharing the same structure, included language and extremity that were considerably different in each of the conditions, such that the first condition was moderate (Text 1, n = 51; e.g., “Additionally, [the asylum seekers’] absorption can harm Israel's economy and increase unemployment among Israel's citizens”), and the second (Text 2, n = 48) and third (Text 3, n = 53) conditions were increasingly more extreme and absurd (e.g., “Additionally, [the asylum seekers’] absorption can significantly harm Israel's economy and increase unemployment among Israel's citizens”), with the fourth condition being very extreme and absurd (Text 4, n = 49; e.g., “This cancer harms Jewish identity and destroys the Jewish essence of the country. Leaving them here even for a short while may harm Israel's economy and significantly increase unemployment among Israel's citizens and even lead to emigration”). Finally, in Text 3 and particularly in Text 4 we included rather blatant racist remarks, as well as references to Nazi Germany and ISIS. For the complete texts translated to English see Supplementary Materials.
To pilot test the materials, we sampled 15 participants and asked them to read all of the texts in a randomized order and indicate the level of extremity and absurdity of each text, on a scale ranging from 1 = not at all [extreme / absurd] to 10 = very [extreme / absurd]. We combined the extremity and absurdity scores for each text (all rs > .51, all ps < .052) to obtain a single rating for each text. A repeated-measures ANOVA showed a significant effect of the texts on the ratings of perceived extremity and absurdity, F(3, 27) = 24.30, p < .001, η2p = .63. Post-hoc analysis indicated that all the comparisons were significant (all ps < .028) (for means and SDs of the pilot study’s variables, see Table 1).
Table 1
Descriptive Statistics for the Combined Extremity and Absurdity Ratings in Pilot Study
| Condition | M | SD |
|---|---|---|
| Text 1 | 3.90a | 2.43 |
| Text 2 | 5.70b | 1.81 |
| Text 3 | 7.13c | 2.07 |
| Text 4 | 8.33d | 1.81 |
Note. Means not sharing the same subscript within extremity and absurdity are significantly different from each other at p < .05.
After reading the opinion editorial, participants completed the dependent variables questionnaire. Finally, participants were asked to comment on the second editorial that they had read. This request is similar to what is common in this type of editorials, in which readers have the option to comment on the texts. See Supplementary Materials for complete study materials in Hebrew.
Measures
Moral conviction was measured with one item assessing participants’ moral conviction regarding their beliefs and attitudes concerning African refugees and asylum seekers in Israel (i.e., “How much are your feelings about the issue of asylum seekers in Israel connected to your core moral beliefs or convictions?”; adapted from Skitka et al., 2005). The response scale ranged from 1 = not at all to 6 = to a great extent. Unless indicated otherwise, all items were measured using the same scale.
Identity threat was measured with five items indicating the extent to which participants felt their identities are threatened by the opinion editorial (α = .78). Three items were adapted from Hameiri et al. (2018), and the fourth and fifth items assessed participants’ emotional reaction after reading the opinion editorial (e.g., “Israelis that support the views that are conveyed in the opinion editorial threaten how I perceive Israeli society”).
Surprise was measured with four items assessing the extent to which participants were surprised, and even shocked by the opinion editorial (e.g., “The opinion editorial surprised me”; α = .89; see Hameiri et al., 2018).
General disagreement was measured with two items indicating the extent to which participants generally disagreed with the message expressed in the opinion editorial (e.g., “The opinion editorial represents reality in a biased manner”; r = .46, p < .001; see Hameiri et al., 2018).
Unfreezing was measured with five items adapted from Hameiri et al. (2018; see also Hameiri et al., 2014) indicating the extent (from 0 = not at all to 100 = very much so) to which the opinion editorial made participants reevaluate their beliefs in general, and with regards to specific immigration-related themes (e.g., “To what extent did the opinion editorial make you reevaluate your attitudes regarding the issue of asylum seekers in Israel”; α = .86).i
Results
To examine the effects of our manipulation and the moderating effect of moral conviction (centered at the mean) on our dependent variables, we used Hayes’s (2018) PROCESS (Model 1) bootstrapping command with 5,000 iterations for a multicategorical independent variable by using indicator coding (Hayes & Montoya, 2017). PROCESS created three dummy variables, in which we specified the Text 3 condition as a reference group. Thus, throughout the analyses, D1 reflected the Text 3 versus Text 1 comparison, D2 reflected the Text 3 versus Text 2 comparison, and D3 reflected the Text 3 versus Text 4 comparison (we included additional comparisons in the Supplementary Materials, and only refer to them briefly in the text where appropriate). Finally, in order to visualize and probe interactions using simple slope analysis (Aiken & West, 1991), moral conviction was fixed at 1 SD below the mean, corresponding to low morally convicted participants, and 1 SD above the mean, corresponding to highly morally convicted participants (see Table 2 for means, SDs, and bivariate correlations for all variables).
Table 2
Descriptive Statistics and Bivariate Correlations Between Variables
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|---|---|
| 1. Unfreezing | 8.83 | 11.63 | – | |||||||
| 2. Disagreement | 2.36 | 1.18 | .23** | – | ||||||
| 3. Surprise | 2.24 | 1.24 | .21** | .49** | – | |||||
| 4. Identity threat | 1.80 | 0.86 | .41** | .64** | .56** | – | ||||
| 5. Moral conviction | 3.98 | 1.53 | .02 | -.02 | -.04 | -.07 | – | |||
| 6. Political orientation | 2.43 | 0.87 | .20** | .25** | .14* | .24** | -.05 | – | ||
| 7. Age | 37.16 | 11.91 | -.14† | -.12 | -.16* | -.11 | .08 | .20 | – | |
| 8. Gender | – | – | .09 | .02 | .03 | .01 | -.00 | -.02 | -.03 | – |
†p < .10. *p < .05. **p < .01.
Identity threat was marginally significantly predicted by D1, b = -0.29, 95% Confidence Interval (CI) [-0.62, 0.03], SE = 0.17, t = -1.78, p = .076, such that participants who read Text 3 (M = 1.87) tended to sense more threat to their identities compared with those who had read Text 1 (M = 1.57). All other main effects were not significant (all ps > .184). Furthermore, we found a marginally significant D2 × moral conviction interaction, b = -0.21, 95% CI [-0.44, 0.01], SE = 0.11, t = -1.88, p = .062; see Figure 1. Conditional effects revealed that while for the low morally convicted participants there were no significant differences between conditions (all ps > .118), highly morally convicted participants reported significantly more identity threat after reading Text 3 compared to Text 1, b = -0.46, 95% CI [-0.91, -0.01], SE = 0.23, t = -2.00, p = .047, and Text 2 (b = -0.49, 95% CI [-0.97, -0.01], SE = 0.24, t = -2.02, p = .045, and similar levels compared to Text 4, b = 0.10, 95% CI [-0.38, 0.57], SE = 0.24, t = 0.40, p = .692. All other interaction terms were not significant (all ps > .326). Participants sensed more threat to their identity after reading Text 4 compared with Texts 1 and 2, regardless of their moral conviction (see Supplementary Materials).
Figure 1
The interactive effect of condition and moral conviction on identity threat.
Note. Error bars represent SEs.
Surprise by the editorial was predicted by D1, b = -0.77, 95% CI [-1.22, -0.32], SE = 0.23, t = -3.38, p < .001, D2, b = -0.47, 95% CI [-0.93, -0.02], SE = 0.23, t = -2.05, p = .042, and D3, b = 0.46, 95% CI [0.01, 0.92], SE = 0.23, t = 2.01, p = .046; see Figure 2, such that participants reading Text 3 were more surprised (M = 2.44) compared to both Texts 1 (M = 1.67) and 2 (M = 1.97), and less surprised compared to Text 4 (M = 2.90) regardless of their moral conviction. All other effects were not significant (all ps > .172).
Figure 2
The effect of condition on surprise.
Note. Error bars represent SEs.
General disagreement with the text was predicted by D3, b = 0.47, 95% CI [0.01, 0.93], SE = 0.23, t = 2.03, p = .044; see Figure 3, such that reading Text 3 led to less disagreement (M = 2.28) compared to Text 4 (M = 2.76), and similar levels compared to Texts 1 (M = 2.19) and 2 (M = 2.23), regardless of moral conviction. All other effects were not significant (all ps > .138). Participants disagreed more with the text after reading Text 4, compared to all other texts (see Supplementary Materials).
Figure 3
The effect of condition on general disagreement.
Note. Error bars represent SEs.
Unfreezing was predicted by D2, b = -5.24, 95% CI [-9.76, -0.73], SE = 2.29, t = -2.29, p = .023, such that participants showed more unfreezing after reading Text 3 (M = 11.21), compared to Text 2 (M = 5.96), and by participants’ moral conviction, b = 2.13, 95% CI [0.10, 4.16], SE = 1.03, t = 2.07, p = .040, such that the more morally convicted participants were, the more unfreezing they reported. All other main effects were not significant (all ps > .169). Furthermore, we found a significant D1 × moral conviction interaction, b = -3.81, 95% CI [-6.78, -0.85], SE = 1.50, t = -2.54, p = .012; see Figure 4. Conditional effects revealed that for the low morally convicted, there were no significant differences between the conditions (all ps > .137). Highly morally convicted participants reported significantly higher levels of unfreezing after reading Text 3, compared to both Text 1, b = -6.71, 95% CI [-12.90, -0.51], SE = 3.14, t = -2.14, p = .034, and Text 2, b = -9.08, 95% CI [-15.60, -2.56], SE = 3.31, t = -2.75, p = .007, and marginally significantly higher than Text 4, b = -6.31, 95% CI [-12.78, 0.15], SE = 3.28, t = -1.93, p = .056. All other interactions terms were not significant (all ps > .106). Unfreezing was marginally significantly higher in Text 1 compared to Text 2 for the participants with low moral conviction (see Supplementary Materials).
Figure 4
The interactive effect of condition and moral conviction on unfreezing.
Note. Error bars represent SEs.
Analysis of comments. Finally, we conducted an analysis of 170 comments, excluding 31 comments that we were unable to analyze, because either they were too short or incoherent (e.g., “No,” or “Short”); or participants did not follow the instruction to comment on the editorial they had read (e.g., “I do not write comments”). Two of the authors, blind to the condition and the participants' moral conviction, coded the comments. Following, the two coders discussed the shared emerging themes and integrated them with categories comprising the paradoxical thinking paradigm. When the two coders did not agree on a given theme, the issue was discussed until an agreement was reached and consequently, recoding of that given theme was conducted in adherence to what was agreed upon. Analysis was conducted on both the content/semantic and rhetorical levels. In terms of context, the broad societal context of negative attitudes toward refugees was addressed. Specifically, consistent with the paradoxical thinking conceptual framework (Bar-Tal et al., 2019), the coders assessed on a dichotomous scale (0 = no, 1 = yes) the following variables: (1) whether participants totally agreed, (2) agreed to an extent, (3) completely disagreed, or (4) tended to agree with the content, but not with the coarse rhetoric of the message source.
A binary logistic regression in which the DVs were regressed on the condition moderated by participants' moral conviction showed, first, that there were neither main effects, nor interactions when predicting total agreement, agreement to an extent, and disagreement (all ps > .188). However, the analysis with agreement, but not with the coarse rhetoric of the message showed a significant main effect of condition, χ2(7) = 24.74, p < .001; see Table 3 for descriptive statistics. Specifically, participants who read Text 3 reported marginally significantly more agreement, but not with the coarse rhetoric, compared to both Texts 1 and 2, b = -1.62, SE = 0.86, Wald χ2 = 3.56, p = .059, odds ratio (OR) = 0.20; and b = -1.23, SE = 0.72, Wald χ2 = 2.93, p = .087, OR = 0.29, respectively, and significantly lower levels compared to Text 4, b = 1.10, SE = 0.49, Wald χ2 = 5.09, p = .024, OR = 3.00. All other effects were not significant (all ps > .696).
Table 3
Descriptive Statistics of the Existence (0 = no, 1 = yes) of Agreement With the Text, but not With Its Coarse Rhetoric Style Divided by Condition
| Condition | No | Yes | Total |
|---|---|---|---|
| Text 1 | 38 (95.0%) | 2 (5.0%) | 40 |
| Text 2 | 40 (93.0%) | 3 (7.0%) | 43 |
| Text 3 | 36 (80.0%) | 9 (20.0%) | 45 |
| Text 4 | 24 (57.1%) | 18 (42.9%) | 42 |
| Total | 138 (81.2%) | 32 (18.8%) | 170 |
The comments provide an additional, interesting window into the understanding of the psychological mechanism at play, specifically in our attempt to pinpoint the ‘sweet spot’ in which the reader experiences a shift in attitude resulting from exposure to an extreme and absurd message. In other words, the reader initially agrees with the general attitude conveyed, but encounters an extremity (but not one which is too extreme and too absurd), which triggers a shift in thought. Table S1 in the Supplementary Materials represents the levels of agreement/disagreement identified in the comments, in line with the three latitudes and the identified fourth state of 'agreement with the message, but not with the coarse style'.
The following sample comment #1 expresses agreement, but not with coarse style:
"The expressive style of the writer of the editorial is coarse and reflects a basic lack of compassion towards human beings. Even if I agree in principle that the infiltrators should be removed from the State of Israel, I do not agree with the way proposed by the writer and his disregard for human life. The fact that we are dealing with infiltrators, which in my opinion for the most part are not being persecuted, but are rather in search of a better life, does not mean that the State should ignore the small percentage of those who are in fact persecuted and provide a solution." (Woman, 45, Text 3).
In this comment, it is apparent that even though the commenter agrees with the writer of the editorial, the coarse style provides a negative perception of the writer as an individual lacking compassion and thus creates a desire for the commenter to disassociate herself, as an individual who does have compassion, from the writer. Perhaps the desire to disassociate herself led to her understanding of the complexity of the situation that is expressed in the commenter's acknowledgement of an existing problem and the difficulty of solving it. As revealed in the analysis of the questionnaires, a process of unfreezing or shift in attitude may take place when the message is extreme, but not too extreme and absurd. In other words, in order for the message to be effective, it needs to reach a point that moderates the view of the message recipient, consequently leading to a contrasting effect.
The following sample comment #2 expresses disagreement with the message: "The comparison to Hitler's "pure blood policy" is not acceptable. This is not our way!!!" (Man, 36, Text 4). In this comment, the commenter does not accept the comparison to Hitler's policy (neglecting the comparison to ISIS) as normative and reflecting the norms and ways of the Israeli society, presumably due to Jews’ history of persecution including by Hitler’s regime. Furthermore, the commenter expresses his intense disagreement with the trajectory that the Israeli society is taking, as reflected by the writer of the editorial. This comment reveals the recipient's disagreement and rejection of the message due to the message's overly extreme content. We argue that at this point the writer has lost the reader's interest in the message.
Discussion
The current research examined the notion that in order for the paradoxical thinking messages to be effective they need to hit a ‘sweet spot’, in that they need to fall within the recipients' latitude of rejection, and correspondently lead to attitude moderation, or to “displacement away from the subject’s position” (Sherif & Hovland, 1961, p. 149). But, at the same time, the messages should not be too extreme and too absurd to raise strong disagreement, and automatic rejection. Specifically, we found among Israeli Jews that the paradoxical (Text 3), but not very extreme and absurd (Text 4), opinion editorial was effective in leading to unfreezing of anti-refugee attitudes for the high morally convicted recipients. Results indicated that for these participants, both Texts 3 and 4 led to higher levels of surprise and identity threat compared to the other texts. Thus, in this respect, Text 4 had the potential to lead to the desired paradoxical thinking effect similar to Text 3. However, the very extreme and absurd manner in which the text was written rendered the text to go too far in the latitude of rejection (at least for some of the participants). This was manifested in high levels of measured disagreement, compared to all other texts, and nearly half of the commenters disagreed with the editorial writer's rhetoric coarse style.
The current study, thus, takes a step forward with regard to the qualification of the conceptual framework of paradoxical thinking. The study indicates that the extreme message has to fall within the ‘sweet spot’ of the latitude of rejection. This is a very meaningful condition that has to be considered by researchers and practitioners who plan such an intervention in the real world. In the present case, in trying to change the views of Israeli Jewish objectors to humane and moral treatment of the refugees by using the paradoxical thinking method, we need to take into account these individuals’ latitude of rejection. Indeed, very extreme messages may be rejected if they fall too far within this latitude. The method works well with extreme messages that moderated Israeli Jewish hawks’ views regarding the Israeli-Palestinian conflict, as was shown in a number of studies (Hameiri et al., 2014, 2016, 2018). This study extends the range of areas in which this approach was examined and provides evidence that the paradoxical thinking approach can be used also in the context of anti-refugee attitudes, which is very important and relevant to many other countries worldwide.
We also hypothesized that participants with low moral conviction would be mostly affected by the more moderate conditions. However, the results painted a more complicated picture. In reference to our hypothesis, in the case of unfreezing, Text 1 tended to lead to more unfreezing, at least compared to Text 2, while it was only descriptively higher than levels of unfreezing in Texts 3 and 4. As mentioned above, levels of surprise and disagreement were lower in these conditions regardless of participants’ moral conviction. Finally, for participants with low moral conviction, Text 4 still raised the highest levels of identity threat.
One possible explanation to account for this unexpected finding stems from the fact that all four texts were consistent with the views of participants with varying degrees of extremity and absurdity. Thus, we compared the more moderate texts with the more extreme and absurd ones. In previous studies, when the paradoxical thinking messages were compared with either neutral or inconsistent messages, clear differences were found (e.g., Hameiri et al., 2014, 2016, 2018; Knab & Steffens, 2019; Swann et al., 1988). The present study, however, had a different purpose, going beyond the previous studies. We designed the study such that Text 1 would serve as a moderate, almost neutral, condition. However, we found that this condition was somewhat extreme and absurd in the pilot study (see Table 1), rendering it to be a suboptimal control condition. Future research could replicate this study’s design while either including a control condition that is completely neutral, or that includes inconsistent information.
Another limitation that should be noted is that our study was relatively underpowered. Indeed, the very low percentage of participants (35.6%) that were eligible to take part in our study, based on their disagreement with the pro-refugee editorial in Haaretz, led to the fact that we ended up with a slightly smaller sample than we intended. However, as we detail above, our sample size allowed us to detect a medium effect size for the condition main effect, and a rather small effect size for our key interaction term (i.e., with the comparison between Texts 1 and 3, which eventually yielded a slightly smaller effect size of Cohen’s f2 = .031). Nevertheless, the fact that the present study replicated previous studies examining the paradoxical thinking conceptual framework (Hameiri et al., 2018), and that the qualitative analysis corresponded with the quantitative findings, lends confidence in the robustness of the quantitative findings of the present study.
It should also be noted that all of the previous recent research, conducted as part of the paradoxical thinking conceptual framework (Hameiri et al., 2014, 2016, 2018; for a review, see Hameiri et al., 2019) was conducted in the context of the Israeli-Palestinian conflict, targeting the beliefs of Israeli Jews who hold conflict-supporting beliefs. Furthermore, previous research almost exclusively relied on either a paradoxical thinking media campaign (i.e., “The Conflict”) or a leading question paradigm (see Hameiri et al., 2019). Thus, two important contributions of the present research are, first, that it introduced a new research paradigm, and possible intervention, for studying the paradoxical thinking conceptual framework; and, second, that it examined these questions in the context of (negative) attitudes held by Israeli Jews toward refugees and asylum seekers. Moreover, based on the literature on which the paradoxical thinking conceptual framework was developed, and on some more recent evidence, we argue that these paradoxical thinking principles should be applicable in other contexts and for other populations. For example, Swann et al. (1988) managed to moderate conservative attitudes regarding women among students in the United States using paradoxical leading questions; and in a recent study, using an adapted version of Swann et al.’s (1988; see also Hameiri et al., 2018) leading questions technique, Knab and Steffens (2019) found that it moderated anti-refugee attitudes among German participants.
Nevertheless, more research is still needed to extend the external validity of the paradoxical thinking phenomenon. The current study and the research we reviewed do not indicate that by targeting liberal or more leftist beliefs, paradoxical thinking interventions can also moderate these beliefs among liberal or leftist individuals. Previous research on ideology has suggested that conservatives and liberals differ in various respects, such as the degree of their need for closure, and their fear of threat to their self-esteem (Jost, Glaser, Kruglanski, & Sulloway, 2003). This puts forward the question whether the paradoxical thinking interventions will be equally effective with liberal participants, as a core mechanism that underlies the paradoxical thinking observed effects is the threat to identity it evokes. A future study could examine, for example, whether exposing liberal participants who hold pro-refugee beliefs with an amplified version of their beliefs will lead to their moderation; and whether this will be moderated by their political orientation, such that it will be more effective the further to the left they are, or the more morally convicted they are.
Despite these limitations, we believe that the present research has several important implications. In terms of theory, one of the main limitations of previous research on the mechanisms underlying paradoxical thinking is that while it was evident that identity threat, surprise, and disagreement all played a role in the process, it was unclear what the order was in which these constructs were activated (Bar-Tal et al., 2019; Hameiri et al., 2018). The present study promotes our knowledge on the psychological mechanisms underlying paradoxical thinking by, first, manipulating the degree of extremity and absurdity of the messages, and thus showing that disagreement with them has a causal effect, and that using a very extreme and absurd message serves as a boundary condition. Second, by showing that regardless of disagreement levels, Texts 3 and 4 led to high levels of surprise and identity threat, it suggests that these mechanisms are activated before disagreement with the message leads to resistance (see also Hameiri et al., 2018). Also, as noted, the present research contributes to social judgment theory (e.g., Sherif et al., 1965; Sherif & Hovland, 1961) by showing empirically for the first time, to the best of our knowledge, that messages that are consistent with the recipients’ held views, but much more extreme and absurd, can fall outside their latitude of acceptance, and in their latitude of rejection. Nevertheless, rather than leading to a contrast effect, these very extreme and absurd messages raise strong resistance, rendering them ineffective in leading to attitude moderation.
Finally, in addition to the introduction of a new potential intervention, based on the paradoxical thinking principles, one additional important practical implication of the present study is that it underscores the importance of rigorous pilot testing before initiating a new study or intervention based on the paradoxical thinking principles. In previous studies in which paradoxical thinking interventions were examined (Hameiri et al., 2014, 2016, 2018), these were rigorously pilot tested in a process of trial and error until they evidenced the desired psychological process, rather than disagreement, and resistance (see thorough discussion in Hameiri et al., 2019). In these previous attempts, the pilot testing was part of the development of the paradoxical thinking materials. The present research shows empirically, for the first time, the possible consequences of using messages that do not hit the paradoxical thinking ‘sweet spot’, such that they can either not induce a surprised reaction or identity threat, or elicit strong disagreement with the message, leading to resistance. Indeed, it is crucial to thoroughly pilot test paradoxical thinking interventions, as they can potentially have deleterious effects.
Ultimately, this means that in translating the paradoxical thinking approach to interventions, such as when designing a new campaign, we face different challenges, including the serious possibility that it may backfire. However, the present research also suggests that researchers and practitioners can avoid this potential pitfall by identifying the ‘sweet spot’ for the particular intervention. It is an important, innovative, and practical tool for those who want to design effective interventions. We hope that these implications will be taken into account by researchers and practitioners in future endeavors involving the paradoxical thinking conceptual framework.