Introduction

In February of 2023, a significant shift occurred in Israel’s political landscape as an ultra-conservative government assumed power and set in motion a series of legislative changes aimed at curbing the authority and independence of the country’s judicial system. In response to this move, a wave of civil protests broke out across Israel involving hundreds of thousands of liberal demonstrators, who protested against what they have characterized as a threat to democracy. However, a curious phenomenon emerged in the wake of this mobilization. While the call to protect democracy resonated powerfully with liberals and galvanized them to take a stand, it was either ignored or elicited backlash among conservative audiences1.

While liberals predominantly viewed democracy as a safeguard for human rights and social equality, conservatives emphasized the meaning of democracy as majority rule. For conservatives, the argument that the protestors were acting to safeguard democracy was hypocritical, as they perceived the protests as an undemocratic attempt to undermine the policies of a government elected by popular mandate2.

This fundamental difference in the way democracy is understood underscores the critical importance of tailoring distinct messages to individuals with differing political ideologies. It highlights how political affiliation is intricately tied to underlying socio-psychological characteristics and motivations that shape the way individuals process information and perceive their social reality3.

Understanding how competing political affiliations may lead to different interpretations of social relations is especially significant in resolving intergroup conflicts. The current work wishes to elicit the importance of addressing the specific needs and motivations of competing political ideologies to increase Jewish Israelis’ support of social equality towards Arab citizens of Israel.

In the following sections, we present a methodological framework rooted in empirical research designed to match and target individuals from different ideological groups with precisely tailored psychological intergroup interventions. We commence with a brief overview, drawing comparisons between the conventional intergroup intervention approach and the principles of targeted communication. We then delve into the potential moderating effects of political identification on intergroup interventions focused on promoting social equality.

Subsequently, we introduce our proposed theoretical rationale and hypotheses that guide the selection of psychological intergroup interventions tailored to members of distinct political groups. Finally, we present the findings from two experimental studies that rigorously test these hypotheses and shed light on the path toward more effective intergroup interventions in the pursuit of social equality.

In recent years, social psychology has shifted towards an interventionist approach, mirroring trends in other psychological sciences4. Social psychologists have increasingly focused on evidence-based interventions to foster constructive relations among conflicting social groups5. These interventions aim to reduce intergroup violence, promote reconciliation, and support social equality6,7,8,9.

However, two major limitations exist. First, most research and public campaigns apply uniform interventions across diverse populations, ignoring individual variations in psychological needs and motivations, which can affect intervention effectiveness10,11,12. Second, translating empirical findings from controlled experiments into real-world campaigns is challenging, often leading to mixed results due to underestimating real-life dynamics13,14,15,16,17.

In contrast, modern persuasive communication effectively matches individuals with tailored messages, as seen in fields like medicine18 and marketing19,20. In 2023, 78% of social media campaigns were consumer-targeted, highlighting the effectiveness of targeted marketing21,22. Companies like Meta use A/B testing to assess and optimize campaign appeals based on audience response23.

Despite the prevalence of the self-congruity principle in targeting studies, these often lack robust theoretical frameworks and rely on intuition24,25, limiting their generalizability26. More complex psychological mechanisms are involved for interventions aimed at changing long-standing attitudes and political behaviors, necessitating a systematic theoretical framework for effective implementation in real-world contexts.

Political identification goes beyond ideology, encompassing personality traits, psychological motivations, and needs that shape one’s ideological leanings27. This identification significantly influences how information is processed in conflict scenarios, due to both the content and motivating properties of the ideology28,29,30. Thus, the effectiveness of psychological intergroup interventions on politically charged issues depends on one’s political identification. These interventions derive their power from specific psychological motivations, which can vary among different political groups. Unlike narrow personality traits or social orientations, political identification includes a wide range of socio-psychological characteristics crucial for shaping intergroup attitudes31. Targeting individuals across the political spectrum is practical, especially on social media, unlike mass targeting based on complex psychological traits.

Moral reframing studies show that aligning messages with an individual’s moral values can be an effective political communication tool32. However, what appeals to one group may offend another if they hold different moral foundations33. Moral reframing aims to persuade individuals to adopt an opposing viewpoint, which may not be effective in increasing support for already-held positions, like environmentalism among liberals. Its success depends on how well the underlying morality aligns with the target’s convictions, often relevant only in specific contexts34.

Empirical studies on political targeting have produced mixed results regarding its effectiveness26,35. Tappin et al.12 conducted an innovative study on political microtargeting by using machine learning to identify effective messages for different socio-demographic groups. They found that microtargeting had a larger persuasive impact compared to alternative strategies for garnering support for the US Citizens Act.

While this pioneering work bears significance, it comes with limitations. The study adopted an exploratory approach, relying on message matching and targeting based on previous survey data and automated algorithms, devoid of a comprehensive socio-psychological framework or specific hypotheses regarding intervention-moderating variables. Moreover, the research was conducted outside the real-world context of social media campaigns, relying on traditional surveys with experimental designs, which may not accurately represent the intricacies of social media’s integral role in message dissemination.

Our current endeavor does not seek to transplant targeted marketing into the conflict resolution field or merely transpose conventional socio-psychological interventions into social media campaigns. Rather, it aspires to amalgamate the strengths of both domains into a cohesive framework that harnesses political targeting to optimize the efficacy of social equality interventions in practical settings.

Selecting and matching targeted interventions involves aligning individual factors with intervention types based on motivational congruency and baseline attitudes, and the operationalizing of the intervention should be as similar as possible to the actual medium implementation.

  1. 1.

    Congruency and content: Interventions must match the psychological needs and motivations of each political group Tappin et al.12. However, their effectiveness also depends on how much the conveyed information deviates from individuals’ existing perceptions and attitudes36. For example, interventions promoting outgroup heterogeneity will have little impact on those already perceiving their outgroup as diverse.

  2. 2.

    Intervention selection: Interventions are chosen from those known to positively affect the desired outcome or psychological mechanism of each group37. Competing interventions should be linked to distinct psychological mechanisms for precise alignment with group needs38. Consistency in presentation is important in order to isolate the message and mechanism from confounding variables.

  3. 3.

    Implementation orientation: Intervention selection and operationalization should match the medium and context of their intended use for real-world effectiveness. For example, outdoor campaign interventions should be designed as graphic advertisements. Social equality interventions relying on mass communication should include a field replication stage to ensure their efficacy in practical scenarios.

Our goal is to optimize the alignment between individual-level factors and chosen interventions based on motivational congruency and baseline evaluations. We analyzed the psychological needs and social cognitions of Israel’s three major political groups: Conservatives, Centrists, and Liberals. Each group was matched with the most relevant social equality intervention from established psychological intergroup interventions. These selections serve as a starting point to showcase the potential of our framework.

Conservative individuals exhibit distinct traits characterized by a strong desire for order, certainty, and structure39. Their heightened sensitivity to threats, as highlighted by Wright and Baril40 and Jost and Amodio30, stems from a motivation to resolve ambiguity and address perceived threats. This sensitivity often leads conservatives to attribute exaggerated hostility to social outgroups41,42. Given this proclivity, conservatives should be particularly responsive to interventions designed to mitigate uncertainty and threat sensitivity by challenging their misconceptions about outgroups. A classic example of threat-reduction intervention is the “corrective information” intervention, which exposes members of a warring group to information about its outgroup members’ perceptions and positions43,44. Given that intergroup meta-perceptions tend to be negatively biased45, correcting them has been found to be a fruitful avenue for intervention44. A recent study conducted in the context of Jews and Arab citizens of Israel46 has found that presenting Jewish-Israelis with information about the attitudes of Arab Israelis (showing that the vast majority of Arabs oppose violence against Jews), led to a significant decrease in Jews’ support of violence against Arab citizens. This effect was especially prominent amongst conservatives, who attributed greater baseline hostility to Arab citizens compared to non-conservatives.

Following this rationale, we developed a 2-min video that presented findings showing that the majority of Arab Israelis oppose aggression towards Jews and believe in coexistence between Jews and Arabs in Israel. The findings portrayed in the clip were based on actual public opinion polls amongst the Arab population conducted by our research team in the past three years.

- H1: (a) Conservative Israeli Jews exposed to a corrective meta-misperceptions intervention will exhibit a significant increase in support for social equality compared to their pre-exposure support of social equality (within-subject effect), (b) as well as compared to conservatives assigned to the control condition (between-subject effect). (c) The positive effect of the meta-perception treatment on Conservatives’ support of social equality will be higher than the effects of the malleability or social norms conditions.

Centrists, as proposed by Robert Samuels in “The Psychology of Political Ideology” (2021), possess unique psychological characteristics that distinguish them from conservatives and liberals. They tend to avoid confrontation, exhibit sensitivity to social cues, and experience discomfort when confronted with attitudes that deviate from the mainstream view47. Centrists place great importance on social consensus, as evidenced by Girvan and Snyder’s48 research on functional independence. Thus, we posit that an effective strategy for reducing prejudice among centrists involves appealing to their sensitivity to social cohesion and the need for harmony. In terms of content, the intervention should emphasize the normative acceptance of social equality within the Israeli Jewish community, thereby positioning it as a mainstream and widely accepted stance. Building on the well-established potency of social norms interventions in promoting positive intergroup outcomes49,50,51, we developed a “normative” intervention featuring a two-minute clip showcasing recent findings indicating a growing trend of support for social equality and Jewish-Arab coexistence among Jewish Israelis.

- H2: (a) Centrist Israeli Jews exposed to a social norms intervention will exhibit a significant increase in support for social equality compared to their pre-exposure support of social equality (within-subject effect), (b) as well as compared to Centrists assigned to the control condition (between-subject effect). (c) The positive effect of the treatment of social norms on centrists’ support of social equality will be greater than the effects of meta-perception or malleability conditions.

Liberals are open to change and malleability and view social groups as malleable52,53,54. Despite this, chronic conflict may undermine their belief in malleability55. In light of this, the intervention’s content should reaffirm the plausibility and inevitability of human malleability in a broader, non-context-specific manner. Given the previously established effectiveness of malleability interventions in promoting conciliatory intergroup attitudes during conflicts, we developed a two-minute historical review emphasizing human adaptability without specific reference to intergroup relations.

- H3: (a) Liberal Israeli Jews exposed to a malleability intervention will exhibit a significant increase in support for social equality compared to their pre-exposure support of social equality (within-subject effect), (b) as well as compared to liberals assigned to the control condition (between-subject effect). (c) The positive effect of the malleability treatment on liberals’ support of social equality will be higher than the effects of meta-perception or social norms conditions. (In our preregistration, the malleability intervention is sometimes labeled “plasticity” or “human agility” due to plans to compare two approaches before consolidating them. Details on hypotheses, analyses, and deviations are shown in Table 3.)

These preregistered hypotheses formed the basis for our empirical investigation into the effectiveness of these targeted intergroup interventions.

Methods

Research design

To put our hypotheses to the test, we executed two distinct experimental studies: A controlled intervention tournament and a corresponding social media field experiment. These investigations were conducted within the intricate context of relations between Jews and Arabs in Israel, a nation characterized by a majority of Jewish citizens (~80%) and a minority of Arab citizens (around 20%) in 2023. Tensions and disputes persist between these ethnic groups due to internal issues such as discrimination, inequality, and uneven resource allocation for Arab citizens. The ongoing historical conflict with neighboring Palestinians, with whom most Arab citizens of Israel identify, further complicates these already fragile relations. These tensions have frequently erupted into violent clashes, with a notable escalation in May 2021.

Our study primarily focuses on the advantaged group, Israeli Jews, for several key reasons. Firstly, the advantaged group typically wields more societal power and resources, making them crucial agents in de-escalating conflicts. Secondly, they bear a relatively greater responsibility for perpetuating these conflicts. Lastly, interventions targeting the disadvantaged often have distinct goals, oriented more toward promoting social change than reducing prejudice56.

Interventions were developed through a series of collaborative meetings by our research team, followed by two preliminary empirical examinations that corroborated the internal validity of each intervention (these examinations, as well as the intervention materials, are provided in the supplementary materials folder).

Study One: Longitudinal intervention tournament

The aim of this study was to examine the effectiveness of three intervention conditions (meta-perception, social norms, and malleability) in promoting support for social equality among Israeli Jewish participants with different ideological orientations (conservatives, centrists, and liberals). The study involved a longitudinal design, with pre-treatment measurements of support for social equality followed by randomized assignment to one of three intervention conditions or a control group, after which post-treatment support for social equality was measured.

This study received approval from the Hebrew University Ethics Committee (approval documentation is available in the supplementary materials: https://shorturl.at/NXQmV). All participants provided informed consent to participate in the study before participation. Participants were informed of their right to withdraw at any time without penalty. The study was preregistered on May 16, 2023, and the full preregistration details, including the hypotheses and analysis plan, are available at https://osf.io/yndfe.

Baseline Measurement: From March 24 to April 18, 2022, we conducted a baseline study among 3979 Jewish Israeli participants (1936 women and 2043 men, Mage = 43.4, SDage = 15.3) who were randomly recruited by a leading Israeli sampling company (iPanel), via an online self-report survey platform (Qualtrics). Participants were paid 8$ for their participation in each of the two-time measurements. After filing demographic measures, participants were asked to express their level of opposition/support of social equality between Jews and Arabs in Israel, as presented by ten statements, with a response scale ranging from 1 (strongly oppose/disagree) to 7 (strongly support/agree) given in Table 1. These statements addressed social equality, with five focusing on general attitudes toward social equality between Jews and Arabs (“Arab citizens should have the same salary and employment opportunities that Jews have”) and the remaining five assessing willingness to take action to promote social equality [“I am ready to act (for example, share a post, sign a petition, etc.] to reduce inequality in the distribution of budgets in government offices”). The 10-item social equality scale exhibited excellent internal reliability, with a Cronbach alpha of 0.93 (Social Equality Scale: M = 4.7, SD = 1.4. See Table 1 for complete reliability analyses). This initial study provided a baseline assessment of participants’ social equality support before any intervention exposure.

Table 1 Alpha Cronabah inner reliability—Social Equality Scale (baseline)
Full size table

Intervention Tournament Measurement: The post-intervention measurement took place from May 12 to 18, 2023. It included 2288 Jewish Israeli citizens (1052 women and 1236 men, Age: M = 46.3, SD = 13.8) who had previously completed the baseline measurement (via the same sampling company and online survey platform). After specifying their political affiliation (as either liberals, centrists, or conservatives), participants were randomly assigned to one of five conditions using a randomized block design—ensuring a sufficient representation of each political group in each condition. Three of these conditions involved the presentation of a two-minute video clip (available in the supplementary materials folder: https://shorturl.at/I0C5v) featuring the experimental interventions: correcting meta-perceptions (N = 504), social norms (N = 478), and malleability (N = 471). The remaining conditions served as the control group (N = 835), comprising either an empty control group (N = 280) or a neutral control group (N = 555). Since there were no significant differences in any of the outcome items between participants assigned to the empty control and those assigned to the neutral control (B = −0.041, SE = 0.1044, Wald χ²(1) = 0.154, p = 0.695), they were combined into one control condition.

Contrary to most studies employing an intervention tournament design57, the video clips used in the current study were developed and tailored to express a specific psychological intergroup intervention. In addition, before the intervention measurement, we engaged a group of experimental social scientists to evaluate each intervention (October 2022). They were exposed to each of the clips without prior knowledge of its underlying intervention, with the task of identifying the psychological mechanisms embedded in each intervention and how distinct the interventions are from one another. This measure ensured the internal validity of the intervention clips (i.e., whether each clip indeed entails the intended intervention mechanism). It was followed by a manipulation check pilot study that reaffirmed the hypothesized effect of each intervention on its corresponding mechanism. After exposure to the video clip (or immediately for participants in the empty control condition), participants were asked to rate their level of support for or opposition to social equality between Jews and Arabs in Israel—using items identical to those used in the Baseline Study (M = 4.78, SD = 1.33, Cronbach alpha = 0.928).

Finally, participants were asked to indicate their political affiliation on a single item (M = 3.582, SD = 1.401) ranging from 1 (extreme right) to 7 (extreme left), with intermediate positions including Right (2), soft-right (3), center (4), soft-left (5), and left (6). This scale aligns with the common measure of political affiliation in Israel, reflecting voting patterns. For the purpose of analysis, we categorized participants into three groups: conservatives (N = 1079) for scores 1–3, centrists (N = 642) for a score of 4, and liberals (N = 567) for scores of 5–7 (see Table 2 for the distribution of participants across conditions based on their political affiliation).

Table 2 Number of participants assigned to each condition by political affiliation
Full size table

A generalized estimating equations (GEE) analysis was performed to examine potential condition differences in support for social equality prior to intervention exposure (Time 1). The results indicated no significant baseline differences between conditions in support of social equality compared to control (Malleability: Wald χ²(1) = 1.703, Social Norms: Wald χ²(1) = 1.491, p = 0.222, Meta-Perception: Wald χ²(1) = 0.004, p = 0.952). An additional GEE analysis revealed no significant differences in support of social equality in time 1 between participants who completed both time measurements compared to those who dropped out after the first measurement (Completed vs Dropouts: B = −0.114, SE = 0.173, Wald χ²(1) = 0.440, p = 0.507).

Study Two: Social Media campaign Interventions

To assess the effectiveness of our interventions in real-life settings, we designed a real-world field study using social media campaigns based on the three interventions included in Study One. We collaborated with the “Israelis Talk” Facebook page, a platform dedicated to fostering discussions among diverse segments of Israeli society and certified by Meta to run ads on social and political issues. We shared the intervention video clips with the page admins, which granted us direct access to their ads manager, allowing us to examine the raw results of the intervention campaigns (unlike Study One, the field study was not included in the preregistration)

To gauge the effectiveness of each intervention across different political groups, we utilized Facebook’s targeting capabilities to create three distinct target audiences. These audiences were formed based on three criteria:

  1. a.

    Geographic location: Most of Israel’s municipalities are politically homogenous in terms of the voting patterns of the residents in each municipality. We analyzed the results of the 2022 national election58 and identified 150 neighborhoods and municipalities where the vast majority (over 78%) of votes went to either conservative, centrist, or liberal parties (about 50 municipalities per political camp). This geographical criterion promised that the majority of members in each target audience were affiliated with one of these political camps.

  2. b.

    Lookalike audiences: The Meta ads platform can generate a target audience that is similar to the followers of specific Facebook pages in users’ demographics and interests. In this case, we added an additional criterion to the geographical audiences based on similarities to users who follow pages associated with conservative, centrist, or liberal politicians and parties.

  3. c.

    Exclusion criteria: We excluded from each target audience users that were included in either of the remaining target audiences, to ensure that users could belong to only one target audience, allowing us to differentiate between conservatives, centrists, and liberals. We excluded Arabic-speaking users to align with our decision to target the high-power group.

For example, the liberal target audience consisted of users who liked (or are similar to users who liked) Facebook pages associated with liberal parties and politicians and live in the 50 areas where liberals gained the most votes—while excluding users who were included in the two remaining target audiences.

Meta ads platform can estimate the budget size necessary to reach a certain number of users (i.e., the amount of money necessary to expose a certain number of users to the ad). We calculated the campaign budgets based on the minimum ad reach of 30,000 unique users, with budgets distributed unevenly due to variations in the sizes of political groups in Israel.

For each of the three target audiences, we executed creative A/B testing campaigns, comparing different Facebook ads targeting the same audience but with distinct ad content. These ads ran for the same duration and budget, and they were placed in identical media placements. All ads included an identical call to action (“An Equal Israel is a Better Israel >> Click and make a difference”), along with a message addressing the content of each intervention clip (for the Meta-Perception clip: “What do the Arab citizens really think about the Jewish citizens?”; for the Social Norms clip: “What do the Jewish citizens really think about the Arab citizens?”; For the malleability clip: “Groups and peoples change frequently, and this change occurs everywhere every day”). Figure 1 displays these ads in English. Upon the predefined completion date of September 8, 2023, Meta’s ads manager provided a report detailing performance metrics for each ad within each target audience.

Fig. 1: Screenshots of the Facebook desktop Intervention ads translated into English.
figure 1

These figures present the three main ads used in each campaign. The social norms campaign ad is presented on the left, the malleability ad is presented at the center and the meta perception ad is presented on the left. The original ads were presented in Hebrew and were later translated into English for the reader's convenience. Each participant was exposed to one of these ads.

Full size image

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Study One

Our analysis strategy consisted of two main stages. First, an overall model was conducted to assess the main and interaction effects of time, condition, and ideology. Second, separate analyses were performed within each ideological group (conservatives, centrists, and liberals) to evaluate condition-specific effects, directly addressing the study hypotheses.

Despite our preregistered plan to conduct two-way ANOVA tests to examine differences in support for social equality across conditions and political groups, the outcome variable violated key assumptions required for this analysis. The Shapiro–Wilk test indicated significant deviations from normality for the dependent variable across all combinations of intervention type and political ideology (all p < 0.001). Additionally, Levene’s test revealed a significant violation of the assumption of homogeneity of variances (F(11, 2408) = 6.935, p < 0.001). The preregistered hypotheses, planned analyses, and justifications for deviations are detailed in Table 3.

Table 3 Preregistration hypothesis and analyses and deviation justifications
Full size table

Given the structure of the data—repeated measures of support for social equality (pre- and post-intervention) from the same participants clustered within different intervention conditions—a generalized estimating equations (GEE) approach was selected to test the hypotheses. GEE is particularly well-suited to this type of data for several reasons: (a) The study’s hypotheses involved both within and between-subject effects (changes in support for social equality by condition, while controlling for baseline attitudes). GEE allows for testing both types of effects in a single model by including pre-treatment support as a covariate, effectively adjusting for baseline differences and examining how each intervention influenced post-treatment outcomes59,60. By modeling the working correlation structure, GEE provides robust estimates of intervention effects while accounting for the within-subject dependence of the repeated measurements. (b) The outcome variable only marginally met the assumptions of normality. GEE is advantageous because it provides robust estimates even when the outcome distribution deviates from normality61, as it is less sensitive to deviations from normality than traditional parametric methods like repeated-measures ANOVA. (c) GEE can handle missing data in repeated measures designs by using all available data in an analysis, without requiring imputation or data deletion62. This is particularly useful in our dataset, where not all participants completed both the pre- and post-intervention surveys, leading to unequal group sizes and some missing values.

Unified GEE analysis with ideological group comparisons

The following section presents the results of the GEE analyses in two steps. First, an overall model was conducted to assess the main and interaction effects of time, condition, and ideology. Second, separate GEE analyses were performed within each ideological group (conservatives, centrists, and liberals) to evaluate condition-specific effects, directly addressing the study hypotheses.

Step 1: Overall GEE analysis (entire sample)

Model specification

The analysis employed a generalized estimating equations (GEE) approach with an exchangeable working correlation structure, assuming within-subject correlation across repeated measures. The dependent variable was support for social equality, measured pre- and post-intervention. Predictors included time (pre/post-intervention: 1, 2), condition (control, meta-perception, social norms, malleability), and ideology (conservatives, centrists, liberals), along with their interactions.

Omnibus tests (Type III Wald Chi-Square tests)

  • Main effect of time: χ²(1) = 17.93, p < 0.001, indicating a significant overall change in social equality support from pre- to post-intervention.

  • Main effect of condition: χ²(3) = 10.99, p = 0.012, suggesting that intervention conditions influenced social equality support.

  • Main effect of ideology: χ²(2) = 1407.43, p < 0.001, confirming significant differences in social equality support by political ideology.

  • Two-way interaction (Time × Condition): χ²(3) = 13.34, p = 0.004, indicating that the impact of intervention conditions on support for social equality changed over time.

  • Two-way interaction (Time × Ideology): χ²(2) = 1.11, p = 0.573, suggesting no significant differences in social equality support between ideological groups before the intervention.

  • Two-way interaction (Condition × Ideology): χ²(6) = 7.50, p = 0.278, indicating that the intervention conditions did not differ significantly across ideological groups before considering time.

  • Three-way interaction (Time × Condition × Ideology): χ²(6) = 15.44, p = 0.017, suggesting that the effect of intervention conditions on support for social equality was moderated by political ideology over time.

Baseline comparability (Time 1 comparisons)

Pairwise comparisons conducted separately for each time measurement (and are presented in Table 4) found that significant condition differences in support for social equality emerged only after the intervention exposure, but not in the baseline measurement.

Table 4 Condition comparison per each time measurement
Full size table

Step 2: Ideological group-specific analyses with within- and between-subject effects

Conservatives (H1: Meta-perception intervention effectiveness)

H1a: Conservatives exposed to the meta-perception intervention exhibited significantly stronger support for social equality compared to their baseline support (Within-subject effect: B = 0.261, SE = 0.104, Wald χ²(1) = 6.23, p = 0.013, 95% CI [0.057, 0.465]), H1b: As well as compared to conservatives assigned to the control condition (between-subject effect: B = 0.231, SE = 0.052, Wald χ²(1) = 19.74, p < 0.001, 95% CI [0.130, 0.333]).

H1c: There was no statistically significant evidence that condition allocations affected social equality support in support of social equality between conservatives assigned to the control condition and those exposed to either the social norm intervention (B = −0.053, SE = 0.066, p = 0.623, 95% CI [−0.183, 0.077]), or the malleability intervention (B = −0.001, SE = 0.062, p = 0.974, 95% CI [−0.122, 0.120]).

Centrists (H2: Social norms intervention effectiveness)

H2a: Centrists exposed to the social norms intervention exhibited significantly stronger support for social equality compared to their baseline support (within-subject effect: B = 0.334, SE = 0.108, Wald χ²(1) = 9.60, p = 0.002, 95% CI [0.122, 0.546]). H2b: Additionally, centrists assigned to the social norms condition reported significantly greater post-intervention support for social equality than those assigned to the control condition (between-subject effect: B = 0.312, SE = 0.097, Wald χ²(1) = 10.32, p = 0.001, 95% CI [0.122, 0.502]).

H2c: There was no statistically significant evidence that condition allocations affected social equality support. in support of social equality between centrists assigned to the control condition and those exposed to either the meta-perception intervention (B = 0.164, SE = 0.057, p = 0.081, 95% CI [−0.013, 0.404]), or the malleability intervention (B = 0.043, SE = 0.034, p = 0.426, 95% CI [−0.132, 0.281]).

Liberals (H3: Malleability Intervention Effectiveness)

H3a: Liberals exposed to the malleability intervention exhibited significantly higher support for social equality compared to their baseline support (within-subject effect: B = 0.267, SE = 0.086, Wald χ²(1) = 9.54, p = 0.002, 95% CI [0.098, 0.437]), H3b: As well as compared to liberals assigned to the control condition (between-subject effect: B = 0.278, SE = 0.079, Wald χ²(1) = 12.21, p < 0.001, 95% CI [0.122, 0.434]).

H3c: There was no statistically significant evidence that condition allocations affected social equality support between liberals assigned to the control condition and those exposed to either the meta-perception intervention (B = 0.037, SE = 0.098, p = 0.709, 95% CI [−0.185, 0.194]) or the social norms intervention (B = 0.150, SE = 0.099, p = 0.132, 95% CI [−0.075, 0.329]).

Table 5 presents the intervention study hypotheses and their respective empirical corroboration, and Fig. 2 presents the estimated marginal means of support for social equality for three political groups (Conservatives, Centrists, and Liberals) across four experimental conditions (Control, Meta-Perception, Social Norms, and Malleability).

Table 5 Intervention Study Preregistered Hypotheses and their respective empirical corroboration
Full size table
Fig. 2: Adjusted support for social equality by political group and experimental condition.
figure 2

This figure displays the estimated marginal means of support for social equality for three political groups (Conservatives, Centrists, Liberals) across four experimental conditions (Control, Meta-Perception, Social Norms, and Malleability). Error bars represent the standard errors of the means. Annotations: Significance: Significant effects (p < 0.05) are highlighted in blue (Conservatives: Meta-Perception), red (Centrists: Social Norms), and green (Liberals: Malleability). Effect sizes and confidence intervals: Each condition with a significant effect is annotated with B values, SE, 95% confidence intervals, Wald Chi-Square, and p-values. Wald Chi-Square and p-values: The annotations indicate the strength of the effects as tested in the GEE analyses. Interpretation: For conservatives, the meta-perception intervention (n = 241) shows a significant increase in support compared to the control (χ² = 6.23, p = 013, B = 0.231, SE = 0.052, 95% CI [0.130, 0.333]), while other interventions do not yield significant changes. For Centrists, the Social Norms intervention (n = 137) results in a significant increase in support compared to the control (χ² = 9.60, p = 002, B = 0.334, SE = 0.108, 95% CI [0.122, 0.546]), with other conditions showing non-significant effects. For Liberals, the Malleability intervention (n = 117) was significantly higher compared to the control (χ² = 9.54, p = 002, B = 0.267, SE = 0.086, 95% CI [0.098, 0.437]), while Meta-Perception and Social Norms interventions were not.

Full size image

These results validate our main hypotheses, demonstrating that social equality interventions yield different effects among various political groups. Figure 2 visually presents the descriptive of our outcome variables, showing that 38% of conservatives assigned to the meta-perception exhibited average support of social equality (ratio of participants scoring higher than “4-Unsure”) compared to 25% in the control group, centrists assigned to the social norms condition exhibited the highest frequency of support for social equality (76%) compared to 58% in the control group, and the highest proportion of liberal participants supporting social equality occurred after exposure to the malleability intervention (95%, compared to 87% in the control group).

Study Two: Facebook advertising campaign effectiveness by ideology

The second study leveraged Facebook’s Meta Ads Manager to conduct a real-world field experiment, delivering three distinct advertising campaigns (Meta-Perception, Social Norms, and Malleability) to Israeli Conservatives, Centrists, and Liberals. We evaluated the effectiveness of these campaigns using two primary metrics: (a) click rate (CTR): The percentage of users who clicked on the petition signature link. (b) Cost per click (CPC): The amount spent per unique link click (measured in NIS).

Across all campaigns, the advertisements reached over 69,000 Israelis, resulting in 26,000 total engagements and 1577 unique clicks on the petition signature link. Table 6 presents the descriptive statistics for each campaign by the ideological group, and Figs. 3 and 4 visualize these differences.

Table 6 Performance metrics of the ad sets used in each campaign
Full size table
Fig. 3: Cost per Click (CPC) by Ideology and Campaign.
figure 3

This figure presents the average Cost per Click (CPC) in NIS for each ideological group (Conservatives, Centrists, Liberals) across three advertising campaigns (Meta-Perception, Social Norms, Malleability). Error bars represent standard errors of the means. Lower CPC values indicate greater cost efficiency in engaging users. Sample sizes: Conservatives: n = 19,665−Centrists: n = 38,919−Liberals: n = 24,182. Significance annotations: *p < 05; **p < 01; ***p < 001; ns not significant.

Full size image
Fig. 4: Click rate by ideology and campaign.
figure 4

This figure presents the Click Rate (%), representing the percentage of users who clicked on the petition signature link after seeing each campaign. Higher click rates indicate greater engagement**. Error bars represent standard errors of the means. Each ideological group (Conservatives, Centrists, Liberals) shows distinct responses to the campaigns. Sample sizes: Conservatives: n = 19,665—Centrists: n = 38,919—Liberals: n = 24,182. Significance annotations: *p < 05; **p < 01; ***p < 001; ns not significant.

Full size image

Conservatives: Consistent with our hypothesis, Conservative audiences responded most favorably to the Meta-Perception ad, which yielded the highest click rate (2.56%) and the lowest CPC (2.91 NIS). The Meta-Perception ad’s CPC was significantly lower than the Malleability ad (MD = −1.83 NIS, t(11,492) = −195.90, p < 0.001, d = 0.85, 95% CI [0.78, 0.92]), as well as the Social Norms ad (MD = −0.39 NIS, t(11,492) = −44.55, p < 0.001, d = 0.31, 95% CI [0.24, 0.38]). The Meta-Perception ad significantly outperformed the Malleability ad in terms of click rate (z = 2.34, p = 0.019, h = 0.12, 95% CI [0.04, 0.20]). However, click rate differences between the Meta-Perception and Social Norms ads were not statistically significant (z = 1.44, p = 0.149, h = 0.07, 95% CI [−0.01, 0.15]).

Centrists: The Social Norms campaign was most effective for the centrist audience, producing the highest click rate (1.94%) and the lowest CPC (3.27 NIS). The Social Norms ad significantly outperformed the Malleability ad in terms of CPC (MD = −2.22 NIS, t(29,323) = −379.60, p < 0.001, d = 0.89, 95% CI [0.83, 0.95]) and click rate (z = 4.22, p < 0.001, h = 0.18, 95% CI [0.10, 0.26]). While the Social Norms ad’s CPC was significantly lower than the Meta-Perception ad (MD = −1.47 NIS, t(25,056) = −236.99, p < 0.001, d = 0.74, 95% CI [0.68, 0.80]), the click rate difference between the Social Norms and Meta-Perception ads was not significant (z = 0.48, p = 0.633, h = 0.02, 95% CI [−0.05, 0.09]).

Liberals: The Malleability ad was most effective for Liberals, yielding the highest click rate (3.03%) and the lowest CPC (3.23 NIS). The Malleability ad significantly outperformed the Meta-Perception ad in click rate (z = 3.56, p < 0.001, h = 0.14, 95% CI [0.07, 0.21]) and CPC (MD = −1.13 NIS, t(12,844) = −128.06, p < 0.001, d = 0.63, 95% CI [0.57, 0.69]).

Additionally, the Malleability ad significantly outperformed the Social Norms ad in both click rate (z = 7.60, p < 0.001, h = 0.25, 95% CI [0.17, 0.33]) and CPC (MD = −1.49 NIS, t(17,864) = −68.23, p < 0.001, d = 0.72, 95% CI [0.65, 0.79]).

Descriptive metrics: Conservatives engaged most with the Meta-Perception ad (25 likes), compared to Social Norms (24 likes) and Malleability (14 likes). Centrists reacted most positively to the Social Norms ad (85 likes), compared to Malleability (57 likes) and Meta-Perception (38 likes). Liberals engaged most with the Malleability ad (48 likes), compared to Social Norms (39 likes) and Meta-Perception (29 likes). These engagement metrics are presented for descriptive purposes only, and no inferential statistics were conducted.

The findings from the field study mostly corroborate the findings of the intervention tournament, indicating that different types of psychological intergroup interventions may be especially effective in promoting social equality among members of different political camps. This result aligns with our suggested intervention-matching framework. However, even though the meta-perception campaign yielded the most effective results amongst conservatives and the social norms yielded the most effective results amongst centrists, these conditions did not significantly differ among these two ideological groups in regard to their click rate.

Discussion

In contemporary persuasive communication, the practice of matching diverse individuals with various forms of persuasive messaging has become central18,63. However, when it comes to psychological intergroup interventions, the application of targeted messaging remains relatively limited10,57 and has mostly been examined using an exploratory approach rather than being derived from a well-defined, a priori theoretical framework. This discrepancy is noteworthy, as the core purpose of social psychology is to unravel the intricate interplay between individual psychological traits and social identities, attitudes, and behaviors64. Despite recent conceptual progress (e.g. ref. 65), intergroup interventions often employ uniform approaches across different target populations, overlooking crucial individual differences in psychological needs, motivations, and orientations. These variations can significantly influence the effectiveness of intervention messages, as they do not account for the influence of personal characteristics, context, and communication mediums13.

Building on recent efforts to enhance psychological intergroup interventions through political targeting based on moral congruency32 and inductive explorations12, our study introduces a framework for targeted intergroup interventions. This framework optimizes the potency of such interventions by aligning the most effective intergroup interventions with individuals holding different political ideologies, based on congruence with each political group’s unique psychological motivations. Specifically, this study aimed to identify the most effective psychological interventions for increasing support for social equality between Jews and Arab citizens of Israel among Israeli Jews from competing ideological groups (i.e., liberals, centrists, and conservatives).

In Study One, conducted as an Intervention Tournament56, we found an interaction effect of political affiliation on the effectiveness of psychological intergroup interventions. As anticipated, Jewish Israeli liberals showed the highest support for social equality after exposure to the general malleability condition, while centrists exhibited the highest support after the social norms intervention - and conservatives displayed the highest support for social equality between Jews and Arabs after exposure to the meta-perception intervention.

Study Two examined the replicability of the intervention tournament study by translating it into practical, real-world applications, focusing on social media campaigns. Reaffirming the main results of Study One, liberals continued to show the strongest support for social equality under the malleability intervention, centrists responded most effectively to the social norms intervention, and conservatives to the meta-perception intervention (and also to the social equality intervention). These findings underscore the necessity of tailoring interventions to align with the core values and principles of different political groups based on a framework that optimizes the match between the intervention and its intended audience.

While previous targeting studies often relied on the self-congruity principle without well-defined theoretical or methodological frameworks12 or focused on overcoming barriers within specific groups and outcomes34, this study is the first field examination of targeted intergroup interventions stemming from a well-defined theoretical rationale. This rationale is based on the motivational and attitudinal congruency between psychological needs and intervention mechanisms, which can be applied across various contexts and holds significant potential for extending utilization beyond the context of this study. Although the current work focused on social equality as the outcome of politically targeted interventions, we believe that political targeting can be effective for a wide range of intergroup attitudes that correlate with or are moderated by political affiliation.

Moreover, the political targeting employed in this study is based not solely on identification as conservative, centrist, or liberal but on the underlying values, social orientations, and other psychological characteristics associated with a particular political group. For example, we hypothesize that liberals should be more responsive to the malleability intervention due to their high openness to change, rather than their political ideology per se. Thus, this framework may be used to target a broad range of psychological moderators and desired outcomes.

Our research carries important practical implications for promoting social equality in polarized environments where individuals from competing political groups are often motivated by different psychological needs. It demonstrates that customizing interventions to align with political ideologies can substantially enhance their impact, focusing not only on which types of intergroup interventions may be most effective but also for whom. Although the study focused on relations between Jewish and Arab citizens of Israel, implementing political targeting of intergroup interventions is relevant in various contexts—from promoting support for humanitarian aid and conciliatory attitudes between Israelis and Palestinians to reducing political polarization and violence between ideological groups and any context where political ideology may moderate intervention effectiveness. Finally, the current work introduces a new methodological approach for empirically implementing targeted interventions in the field through social media metrics.

Limitations

However, this study has several limitations. The generalizability of our findings may depend on specific contexts, and ongoing changes in political affiliations warrant further investigation. Additionally, our study targeted a specific psychological moderator (political affiliation) and a specific outcome (support for social equality) within a particular intergroup context (Jewish and Arab citizens of Israel). Extending this framework to other contexts, outcomes, and targeting dimensions requires further research. Moreover, replication in real-life settings and platforms beyond Instagram and Facebook is crucial, as is examining how well the outcome measures used in this study (e.g., cost per click) capture our theoretical constructs.

Future research should explore several key areas, including the longevity of interventions and potential moderating factors. Strategies for sustaining behavioral changes over time should also be explored. Comparative studies across diverse contexts may reveal universal principles and cultural nuances that shape intervention success.

Conclusion

Our study represents a significant advancement in understanding how to promote social equality in divided societies. Tailored interventions offer the potential to bridge divides, foster meaningful dialogues, and promote social equality. As we navigate the complex landscape of social change, we hope these findings inspire the development of more targeted interventions, contributing to a more equitable world.