Main

Internal conflicts—clashes between groups within a country across ethnic, religious and/or ideological lines—have claimed the lives of millions in the past three decades alone1. These conflicts are particularly long-lived due to their high rates of recidivism1. Understanding how to facilitate exits from internal conflicts, and in a way that avoids conflict relapse, is therefore a critical social challenge. How can we facilitate peace in the shadow of protracted conflict?

One potential resolution to internal conflict is through diplomatic efforts by political leaders. However, even if political leaders broker a peace deal, peace may be difficult to institute and maintain without the support of the populace. For example, in 2016, the Colombian government and the leadership of the Colombian Revolutionary Armed Forces (FARC) guerrilla movement signed a peace deal to end an internal conflict that has claimed nearly 250,000 lives since its inception in the 1950s. However, when the treaty was presented to Colombians in a national referendum, it was rejected (by a margin of less than 0.5%—that is, 50,000 votes among more than 13 million votes cast). A revised deal was subsequently ratified by the Colombian Congress (without a second referendum), but so long as a substantial proportion of the populace remains resistant, the risk of unsuccessful reintegration and conflict relapse will linger. That is, diplomatic efforts among political elites to achieve peace may not suffice—for a sustainable peace to take hold, it is critical to address psychological barriers among the public hindering conflict resolution2,3,4,5,6,7,8. Here we set out to develop a psychological intervention with the potential to precipitate a shift in non-FARC Colombians’ support for peace with and reintegration of FARC ex-combatants that, if successful, could be readily scaled up to help increase societal desire to pursue peace and facilitate the reintegration of ex-combatants.

While many debiasing and conflict resolution efforts focus on direct person-to-person contact9,10,11,12, these interactions can be logistically difficult to achieve for groups involved in conflict and are often limited to those already possessing sufficient motivation to meet face-to-face with ‘the enemy’13,14 (but see ref. 15). Technology provides alternatives to person-to-person contact. For example, technology has been used to facilitate computer-mediated or virtual contact between enemy group members16,17,18 and to provide ‘parasocial’ or vicarious contact19,20,21—exposing individuals to information about the outgroup and its views through live or pre-recorded audio or video media (that is, a ‘media intervention’).

Media interventions have several characteristics that make them particularly well suited for conflict or post-conflict settings. They can extend beyond the subset of people with the motivation and means to meet ‘the enemy’20,21,22,23,24,25, and people can engage with media interventions in comfortable settings (for example, their own homes), thus avoiding the anxiety that characterizes (and potentially thwarts) face-to-face encounters13,16,26. Media interventions can also be tuned to target specific psychological processes. For example, reading or listening to narratives highlighting the difficulties faced by individuals who are homeless, elderly, disabled or otherwise marginalized has been shown to increase affective empathy towards the target and positive affect towards the group that the target individuals are from27,28.

Media interventions can also target psychological processes beyond affect. For example, in Rwanda, a radio soap opera depicting fictitious tribes engaging in peace-affirming behaviours shifted social norms about obedience to authority thought to have contributed to the Rwandan genocide20. Media interventions can also be used to target specific beliefs about the ingroup, the outgroup or the conflict. For example, past work has used videos to effectively challenge the view that Palestinians are inherently violent28, that Muslims are collectively responsible for the actions of individual extremists29 and that the Israeli ingroup is particularly righteous relative to its Palestinian enemy24. Radio shows in countries in the Great Lakes Region in Africa have also been shown to reduce psychological barriers to conflict resolution, including the proclivity to compete over claims to victimhood22,23. Thus, taken together, existing research suggests that media interventions can successfully target a range of psychological processes, from increasing positive affect towards an outgroup to changing specific conflict-related norms, beliefs and mindsets.

Here we developed and tested a series of media interventions to increase support for reintegration and peace among non-FARC Colombians. Opportunities for direct in-person or virtual contact between the non-FARC Colombian public and FARC members are extremely limited. The FARC have historically occupied areas in rural and remote parts of Colombia separated from the rest of the population. Even demobilized ex-combatants are largely sequestered in rural demobilization camps with limited infrastructure (for example, no internet access), where they interact with only a small number of non-FARC Colombians, with mixed results30. Moreover, because the conflict has been going on for several decades, most non-FARC Colombians have never interacted with a FARC member, creating fertile conditions for the proliferation of unchecked negative intergroup beliefs and misunderstandings. This lack of interaction is compounded by individuals’ baseline tendency to consistently overestimate outgroup negativity towards the ingroup, particularly in competitive intergroup contexts—a tendency that recent work has shown can be corrected by exposure to disconfirming information31. We therefore aimed to increase support for peace and reintegration among non-FARC Colombians by exposing them to FARC ex-combatants and their beliefs.

To determine which specific beliefs were most critical to intervene on in this context, we started from a grounded perspective, adapting a well-established framework used in the construction of health messaging campaigns32. The Hornik and Woolf approach32 has several components: (1) a qualitative assessment to identify a superset of beliefs associated with the target outcome(s), (2) a formal quantitative test of the candidate beliefs to determine which are held by enough people that a shift in the belief could cause meaningful change in policy, (3) a choice of belief from those identified in the first two steps that could plausibly be changed, (4) the construction of a media intervention that challenges the target belief and (5) an assessment of media intervention efficacy in a randomized study.

We adapted this approach to conflict resolution by first conducting and filming a set of relatively informal semistructured street interviews with members of the Colombian public, of diverse ages, genders and backgrounds, to identify common concerns about FARC ex-combatants. These ten-minute interviews were conducted on the streets of Medellín, the capital of Antioquia (an area in Colombia that experienced some of the most intense fighting throughout the conflict33). Most notably, these interviews included questions about the most common everyday negative and positive sentiments and beliefs that respondents had heard expressed by others around them about the FARC (for a similar approach, see ref. 34; for a sample of the interviews, see https://player.vimeo.com/video/388593481). We then identified from the interviews what emerged descriptively as a consistent set of negative conflict-related beliefs about FARC. These included, most notably, beliefs about FARC members’ responsibility for their actions and beliefs about FARC members’ unwillingness and inability to reintegrate and give up violence. Of note, this set of beliefs is reminiscent of perceived group malleability, which has been highlighted in previous research as a key psychological barrier to peace and a belief that can be reduced to promote conflict resolution35,36.

Results

Preliminary Survey

To more formally investigate support for these beliefs and their correlates, we next conducted an initial cross-sectional online survey. We surveyed a sample of 435 non-FARC Colombians, who were broadly quota-matched to their respective demographic proportions on the dimensions of age, gender, socio-economic status and region in Colombia (see Supplementary Table 1 for more information about national demographic proportions). In addition to measuring the beliefs we highlighted above (identified from the round of street interviews), we asked the participants about four potential outcome measures: a multi-item measure of attitudes capturing support for peace, a multi-item measure of attitudes capturing support for the reintegration of FARC ex-combatants, a measure of blatant dehumanization of FARC members (strongly associated with support for aggression across contexts37,38,39,40) and the behavioural willingness to sign a petition expressing support for terminating the peace process. Whereas some research on conflict resolution emphasizes the importance of cognitive beliefs about the outgroup (for example, beliefs about unwillingness and inability to change35,36) or the conflict41,42, other research emphasizes more affective routes, highlighting variables such as affective prejudice and empathic concern that capture how individuals feel (versus think) about other groups43,44,45. We thus added here a measure of affective prejudice commonly used in intergroup research (“Feeling Thermometer”46; for complete information about the Preliminary Survey methods and results, see the Supplementary Information).

We found high median endorsement of the belief that FARC members are unwilling and unable to reintegrate and give up violence, with the average significantly above the midpoint (median = 61.53; mean = 59.71; s.d. = 26.01; scale midpoint, 50; t(433) = 7.78; P < 0.001; Cohen’s d = 0.37; 95% confidence interval (CI) = [7.26, 12.17]; see Supplementary Table 2 for variable means and intercorrelations). Moreover, this belief was significantly correlated with each of our attitudinal outcome measures (support for FARC reintegration policies: r = −0.42; 95% CI = [−0.50, −0.33]; support for peace: r = −0.48; 95% CI = [−0.56, −0.40]; dehumanization: r = 0.51; 95% CI = [0.43, 0.59]; all P < 0.001). It was also associated with the willingness to sign the petition supporting termination of the peace process (r = 0.39; 95% CI = [0.30, 0.47]; P < 0.001; see the Supplementary Information for a discussion of a second petition). We observed similar patterns for other beliefs we measured (for example, about the extent to which FARC members were responsible for their actions despite forcible recruitment, about their warmth towards non-FARC Colombians and about whether they had been sufficiently held accountable for their actions), although these patterns were descriptively weaker and the beliefs themselves generally less clearly endorsed (Supplementary Table 2). Of note, our measure of affective prejudice was only modestly correlated with the belief about the FARC’s unwillingness and inability to change (r = 0.29; 95% CI = [0.19, 0.39]; P < 0.001) and more weakly associated with the outcome measures: relative to the belief about the FARC’s unwillingness and inability to change, affective prejudice was significantly less correlated with support for reintegration (r = −0.13; 95% CI = [−0.23, −0.02]; P = 0.007; rdiff. = 0.29; 95% CI = [0.20, 0.41]; Steiger’s Z = 5.36; P < 0.001), support for peace (r = −0.16; 95% CI = [−0.24, −0.06]; P = 0.001; rdiff. = 0.32; 95% CI = [0.24, 0.45]; Z = 6.05; P < 0.001), anti-FARC dehumanization (r = 0.40; 95% CI = [0.31, 0.48]; P < 0.001; rdiff. = 0.11; 95% CI = [0.02, 0.25]; Z = 2.27; P = 0.023) and signing the petition (r = 0.16; 95% CI = [0.06, 0.25]; P = 0.001; rdiff. = 0.23; 95% CI = [0.13, 0.35]; Z = 4.24; P < 0.001).

In combination, then, theorizing about the role of group malleability in conflict resolution35,36 and our empirical findings in the Preliminary Survey among non-FARC Colombians pointed to the potential importance of challenging beliefs about the FARC, and perhaps particularly beliefs about FARC members’ unwillingness and inability to integrate. On this basis, we ensured that the interviews we conducted with FARC ex-combatants in designing our media interventions (see below) directly addressed questions about their attitudes towards reintegration.

To develop our media interventions, we collaborated with Colombian filmmakers with context-specific knowledge and access to a FARC demobilization camp (formally, a Territorial Space for Training and Reincorporation, abbreviated ETCR in Spanish), one of 27 such camps in all of Colombia, and one of five in Antioquia; these camps were constructed after the accord was signed47. The demobilization camp directly abuts a village and is within walking distance of a small town. The camp itself has a cordon around it maintained by a police unit, while a military unit provides security. To create realistic and engaging media interventions, we interviewed demobilized FARC members on camera. We designed interview questions that would allow FARC members to express their thoughts about the set of beliefs we aimed to target (that is, about their willingness and ability to reintegrate and give up violence). We took several steps to ensure broad representation of FARC members and the authenticity of their responses (for the videos, see Supplementary Table 3).

Prior research suggests that individuals are apt to resist information that runs counter to their views about another group (for a recent review, see ref. 48), especially when the source of the information is the outgroup itself49,50,51. We therefore reasoned that our intervention might be especially persuasive if the ingroup (that is, non-FARC Colombians) validated perspectives expressed by the FARC outgroup. We thus also interviewed non-FARC Colombians from the neighbouring town and village, the police maintaining the cordon around the camp and members of non-governmental organizations (NGOs) who work in the camp. As with the FARC interviews, we took several measures to increase the representativeness of interviewees and the authenticity of their responses (Methods).

In designing our intervention, we drew from responses to interview questions that we asked about reintegration efforts, and we highlighted evidence from the demobilization camp of successful coexistence between the FARC ex-combatants and the local non-FARC residents. For example, FARC members reported (and non-FARC Colombians corroborated) that despite initial apprehensions on both sides, within months FARC ex-combatants had established a mixed soccer team including FARC members, people from the neighbouring village, military and police; that the FARC camp regularly hosts bingo nights and fundraisers for poor villagers; and that members of all communities had worked together to rebuild a central road after it was washed out. Of note, although we reasoned that the responses of FARC and non-FARC interviewees would convey the impression that FARC ex-combatants were willing and able to change, we did not directly ask them to comment on that.

In collaboration with our team, the filmmakers created a narrative structure that brought together these interview responses in a 5.5-minute video. To directly address potential resistance associated with message content and messenger identity, we first presented responses by FARC ex-combatants (3 minutes), then responses by non-FARC Colombians (that is, ingroup members; 2.5 minutes) that corroborated the sentiments expressed by the ex-combatants about FARC members’ willingness and demonstrated ability to reintegrate. We presented the FARC responses first to allow non-FARC responses to directly counter any scepticism that viewers might have experienced in reaction to statements by FARC ex-combatants. We focus in the main text on this version of the intervention (‘FARC-Integration’ v.1); however, to experimentally interrogate our speculation that providing both FARC and non-FARC responses (and in this order) would most effectively challenge beliefs about FARC members’ willingness and ability to integrate, and ultimately increase support for peace and reintegration, we created three alternative edits of the FARC-Integration media intervention (‘FARC-Integration’ v.2–4). We included these as part of our ‘intervention tournament’ testing the various interventions we designed in Study 1. Briefly, we found that v.1 (which first included FARC responses and then non-FARC corroboration) was descriptively the most effective relative to the control; that said, and given that the other edits of the FARC-Integration video were also generally effective, the difference between the v.1 edit and the other edits was often not itself statistically significant (see Supplementary Tables 6 and 11 for the detailed results).

Beyond challenging perceptions of the FARC’s willingness and ability to reintegrate in the FARC-Integration video, we also considered a variety of other plausible intervention targets, broadly inspired by our findings from the Preliminary Survey and Hornik and Woolf’s32 methodological approach. In particular, we considered beliefs about FARC honesty, and specifically the sincerity of their commitment to the peace process. We targeted these beliefs with another video, designed from interviews in which FARC members commented on their demonstrated commitment to—and the reasons underlying their eagerness for—peace. The video drew on segments of the interviews highlighting that FARC ex-combatants had followed through on their commitments to peace (for example, fully disarming months earlier, as stipulated by the peace treaty) and expressing their desire to stop the cycle of violence to meet the goal of creating a better future for all Colombian children. Paralleling the structure of the FARC-Integration videos, these sentiments were corroborated by footage from non-FARC community member interviews. We also featured the same interviewees and again created a set of four alternative edits (‘FARC-Peace’ v.1–4), varying the order in which they presented views from the FARC and/or community members (see Supplementary Table 3 for more information).

We reasoned that the videos described above would operate at least in part by changing specific beliefs about FARC ex-combatants, beyond any effects on affective warmth towards the group. That is, the videos described above were designed primarily to change beliefs about FARC ex-combatants’ intentions and capacities that might serve as barriers to peace rather than specifically to evoke feelings of empathy, sympathy or warmth towards the group (notwithstanding the potential relationship between negative beliefs about a group and affect towards them). In parallel, we also created and tested several videos that attempted to target affective pathways. In particular, these videos were intended to generate affective warmth and empathy for the FARC by presenting compelling personal narratives that made them appear sympathetic52. Specifically, one video (‘FARC-Daughter’) presented the daughter of a FARC ex-combatant who told her story of being born in prison, growing up in urban Colombia and now living reunited with her mother. A second video presented the story of an individual ingroup member—a non-FARC Colombian police officer—who described how he was treated charitably by the FARC members in the camp (‘Colombian Policeman’). Finally, we tested the effectiveness of a video developed previously by the journal Nature, which highlights the trauma faced by FARC ex-combatants (‘FARC-Trauma’53; for more information about the videos, see Supplementary Table 3).

Study 1

In Study 1, we tested the causal effects of each of these media interventions on non-FARC Colombians’ beliefs about and attitudes towards the FARC, as well as their support for peace and reintegration. We recruited a large sample of non-FARC Colombians broadly resembling Colombian society in terms of age and gender (see Supplementary Table 4 for sample demographic details and a comparison to national demographics) and randomly assigned them to a no-video control condition or one of the 11 media interventions (our key edit of the FARC-Integration video and three alternative edits, four edits of the FARC-Peace video, the FARC-Daughter video, the Colombian Policeman video and the FARC-Trauma video developed by Nature). This intervention tournament approach54 (see also refs. 29,55) allows for the simultaneous comparison of several candidate interventions against the same control group. Despite its efficiency, one potential downside of this approach is the increased risk of false positives with an increased number of comparisons. To verify the robustness of our results, we tested the effects of the interventions not just immediately after exposure to a media intervention but also in a second wave, 10–12 weeks later. Moreover, we ran replication (Study 2) and extension (Study 3) studies on independent samples of participants.

As we describe below, the main edit of the FARC-Integration video (v.1) proved particularly effective and robust when compared with the control. It was therefore the video we chose to build on in subsequent replication studies; thus, we focus primarily in the main text on the comparison between the FARC-Integration video (v.1; n = 172) and the control condition (n = 209; for the complete results and descriptive statistics for all interventions versus the control using univariate analysis, see Supplementary Tables 6 and 11).

We conducted a multivariate path analysis using the lavaan package in R56, testing the effect of condition on the outcome measures (that is, dehumanization, support for peace and support for reintegration) via cognitive beliefs about FARC members’ malleability (that is, their willingness and ability to change) and the affective variables, empathy and prejudice. We first considered this (fully saturated) model at Wave 1. Then, consistent with recent research emphasizing the importance of examining whether interventions promoting tolerance have lasting effects57, we considered whether the effects of our intervention persisted over time by testing the model again at a second wave (that is, 10–12 weeks later; specifically, we tested the effect of experimental condition on the Wave 2 measures of the same mediators and outcome variables we had originally assessed at Wave 1). Our multivariate path analysis allowed us to examine the effects of the manipulation on the outcome measures while accounting for the intercorrelations between them. For analyses in both waves, we addressed missing values and attrition with full information maximum likelihood estimation58. For analysis in Wave 2, we included age, gender and socio-economic status (that is, the demographic variables that were associated with attrition; Supplementary Information) as auxiliary variables, a common strategy to improve power and reduce attrition bias59. Note that we obtained similar findings in both waves when conducting analyses to examine the effect of the manipulation on each of our dependent variables (DVs) separately in an analysis of variance framework with listwise deletion (Supplementary Tables 6, 11 and 16).

At Wave 1, we found that exposure to the FARC-Integration media intervention significantly reduced beliefs about FARC members’ unwillingness and inability to change relative to control participants (unstandardized b regression coefficient = −0.64; s.e. = 0.10; Z = −6.30; P < 0.001; 95% CI = [−0.84, −0.44]; for complete information, see Table 1 and Supplementary Figs. 1 and 2). This effect persisted even 10–12 weeks later (that is, at Wave 2), with non-FARC Colombians in the FARC-Integration condition (n = 110) continuing to more strongly reject beliefs about FARC members’ lack of malleability relative to controls (n = 141) (b = −0.40; s.e. = 0.13; Z = −3.16; P = 0.002; 95% CI = [−0.64, −0.15]). Turning next to the affective mediators, we observed at Wave 1 that those in the FARC-Integration condition reported more empathy towards FARC members (b = 0.57; s.e. = 0.10; Z = 5.59; P < 0.001; 95% CI = [0.37, 0.77]) and lower levels of prejudice than the controls (b = −0.41; s.e. = 0.10; Z = −3.95; P < 0.001; 95% CI = [−0.61, −0.20]). Whereas the effect on empathy towards FARC members persisted 10–12 weeks later (that is, at Wave 2; b = 0.27; s.e. = 0.13; Z = 2.13; P = 0.033; 95% CI = [0.02, 0.52]), the effect on prejudice did not (b = −0.14; s.e. = 0.13; Z = −1.08; P = 0.280; 95% CI = [−0.38, 0.11]). Finally, the FARC-Integration media intervention had a significant effect on all focal outcome measures. At Wave 1, those who watched the video dehumanized FARC members less (b = −0.45; s.e. = 0.10; Z = −4.38; P < 0.001; 95% CI = [−0.65, −0.25]) and expressed stronger support for peace (b = 0.22; s.e. = 0.10; Z = 2.12; P = 0.034; 95% CI = [0.02, 0.42]) and reintegration (b = 0.24; s.e. = 0.10; Z = 2.32; P = 0.021; 95% CI = [0.04, 0.44]). In all three cases, these effects persisted at Wave 2 (dehumanization: b = −0.39; s.e. = 0.13; Z = −3.07; P = 0.002; 95% CI = [−0.63, −0.14]; support for peace: b = 0.29; s.e. = 0.13; Z = 2.28; P = 0.023; 95% CI = [0.04, 0.54]; support for reintegration policies: b = 0.35; s.e. = 0.13; Z = 2.81; P = 0.005; 95% CI = [0.11, 0.60]).

Table 1 The effect of the FARC-Integration (v.1) video (versus the control) on dependent measures in a multivariate analysis in Study 1 at Waves 1 and 2
Full size table

To gain more theoretical insight into the psychological mechanism behind the effectiveness of the FARC-Integration video, we tested the indirect effects in the context of our saturated multivariate path analysis using bootstrapping estimation with 5,000 replications. We first considered this at Wave 1 (Supplementary Fig. 1) and next considered the persistence of these effects by re-examining the model using the Wave 2 measures of the mediators and outcome variables (Supplementary Fig. 2). As predicted, we found that the effects of the intervention on the outcomes at Wave 1 were mediated through beliefs that FARC members are unwilling and unable to change, controlling for effects on positive affect towards FARC members (assessed with empathy and affective prejudice; Supplementary Table 7a). That is, those exposed to the FARC-Integration video were more likely to believe that FARC ex-combatants were willing and able to change, which mediated decreased dehumanization of FARC members (indirect effect = −0.17; s.e. = 0.03; Z = −5.70; P < 0.001; 95% CI = [−0.22, −0.11]), increased support for peace (indirect effect = 0.28; s.e. = 0.05; Z = 6.06; P < 0.001; 95% CI = [0.19, 0.36]) and increased support for reintegration policies (indirect effect = 0.26; s.e. = 0.04; Z = 6.06; P < 0.001; 95% CI = [0.17, 0.34]). We also observed significant indirect effects via increased empathy for all three outcome measures (dehumanization: indirect effect = −0.09; s.e. = 0.02; Z = −4.66; P < 0.001; 95% CI = [−0.13, −0.05]; support for peace: indirect effect = 0.11; s.e. = 0.02; Z = 4.83; P < 0.001; 95% CI = [0.07, 0.15]; support for reintegration policies: indirect effect = 0.18; s.e. = 0.04; Z = 5.32; P < 0.001; 95% CI = [0.12, 0.25]) and via prejudice on decreased dehumanization (indirect effect = −0.14; s.e. = 0.04; Z = −3.87; P < 0.001; 95% CI = [−0.22, −0.07]) but not on support for peace (indirect effect = 0.01; s.e. = 0.01; Z = 1.20; P = 0.232; 95% CI = [−0.01, 0.03]) or support for reintegration policies (indirect effect = −0.01; s.e. = 0.01; Z = −0.72; P = 0.471; 95% CI = [−0.02, 0.01]).

This pattern of results persisted 10–12 weeks later. Those in the FARC-Integration video condition continued to express greater perceptions of FARC willingness and ability to change (as assessed by our Wave 2 measures), which continued to mediate the intervention’s effects on dehumanization (indirect effect = −0.13; s.e. = 0.04; Z = −3.08; P = 0.002; 95% CI = [−0.21, −0.05]), support for peace (indirect effect = 0.20; s.e. = 0.06; Z = 3.12; P = 0.002; 95% CI = [0.07, 0.32]) and support for FARC reintegration policies (indirect effect = 0.16; s.e. = 0.05; Z = 3.11; P = 0.002; 95% CI = [0.06, 0.26]; for the complete details, see Supplementary Table 12a). The same was true with respect to the indirect effects via our Wave 2 measure of empathy (although there was some evidence that the indirect effects via empathy were weaker overall than those via malleability) (dehumanization: indirect effect = −0.04; s.e. = 0.02; Z = −2.04; P = 0.042; 95% CI = [−0.08, −0.002]; support for peace: indirect effect = 0.05; s.e. = 0.02; Z = 2.04; P = 0.042; 95% CI = [0.002, 0.09]; support for reintegration policies: indirect effect = 0.10; s.e. = 0.05; Z = 2.11; P = 0.035; 95% CI = [0.01, 0.20]). In contrast, there were no indirect effects at Wave 2 via reduced prejudice (dehumanization: indirect effect = −0.05; s.e. = 0.05; Z = −1.08; P = 0.281; 95% CI = [−0.14, 0.04]; support for peace: indirect effect = 0.003; s.e. = 0.004; Z = 0.65; P = 0.514; 95% CI = [−0.01, 0.01]; support for reintegration policies: indirect effect = 0.001; s.e. = 0.003; Z = 0.36; P = 0.722; 95% CI = [−0.01, 0.01]).

Study 2

In the next preregistered experiment, we aimed to replicate the effects of the FARC-Integration media intervention on outcomes (the same three measures considered in Study 1), as well as the indirect effect pathways through beliefs about FARC members’ unwillingness and inability to change, and the affective variables empathy and prejudice. (Although we focus in the main text on the FARC-Integration video versus the control, we also tested three other promising videos from Study 1; see Supplementary Tables 13, 16 and 17b–d.) In this experiment, we separated the assessment of the belief that FARC members are unwilling to integrate and give up violence (eight-item measure; α = 0.92) from beliefs about FARC members’ inability to integrate and give up violence (for which we employed a widely used eight-item measure assessing perceived malleability; α = 0.91; refs. 36,60). Our initial intention, as preregistered, was to test these scales separately, to consider whether there might be any difference between perceptions about the unwillingness61,62,63,64 versus inability to change36,65. However, factor analysis indicated that the items from both measures collectively formed a unidimensional scale (eigenvalue, 9.32; percentage of variance explained, 58.25%). We therefore continued to focus on participants’ composite beliefs about FARC unwillingness and inability to change (16 items; α = 0.95).

Consistent with Study 1, in the context of our (fully saturated) multivariate path analysis using lavaan with full information maximum likelihood estimation, we found that those exposed to the FARC-Integration media intervention (n = 452) were less likely than those in the control condition (n = 505) to endorse the specific belief that FARC members are unwilling and unable to change (b = −0.46; s.e. = 0.06; Z = −7.29; P < 0.001; 95% CI = [−0.59, −0.34]). Also consistent with the results of Wave 2 of Study 1, empathy towards FARC members was significantly greater among those who saw the media intervention than in the control group (b = 0.15; s.e. = 0.06; Z = 2.26; P = 0.024; 95% CI = [0.02, 0.27]). Levels of anti-FARC prejudice were not significantly lower than in the control group (b = −0.06; s.e. = 0.07; Z = −0.97; P = 0.334; 95% CI = [−0.19, 0.06]). With respect to the outcome measures, consistent with Study 1, we observed that participants in the FARC-Integration condition evidenced less dehumanization of FARC members (b = −0.16; s.e. = 0.07; Z = −2.43; P = 0.015; 95% CI = [−0.28, −0.03]; for the complete information, see Table 2). The intervention’s effects on support for peace (b = 0.12; s.e. = 0.07; Z = 1.87; P = 0.062; 95% CI = [−0.01, 0.24]) and support for reintegration policies (b = 0.12; s.e. = 0.07; Z = 1.85; P = 0.065; 95% CI = [−0.01, 0.25]) were not significant, but were marginally significant (that is, the 90% CIs did not include zero). Of note, in Study 2, as a secondary outcome measure, we also gave the participants a bonus for completing the study and gave them three options: (1) keep it for themselves, (2) donate half to an organization promoting FARC reintegration through extreme sports (specifically, rafting) or (3) donate the whole sum to the organization. We did not find a significant difference in participants’ willingness to donate between the FARC-Integration condition and the control in a univariate analysis (t(953) = 0.82; P = 0.813; d = 0.02; 95% CI = [−0.08, 0.10]; for the complete information, see Supplementary Table 16). As described further below, we considered the effects of the intervention on willingness to donate again, using an improved and more sensitive measure, in Study 3.

Table 2 The effect of the FARC-Integration (v.1) video (versus the control) on outcome measures in a multivariate analysis in Study 2
Full size table

We tested the same indirect effect pathways as in Study 1 (Fig. 1 and Supplementary Table 17a). As predicted, and as in Study 1, we found that decreases in the belief that FARC members are unwilling and unable to change significantly mediated decreased dehumanization (indirect effect = −0.19; s.e. = 0.03; Z = −6.86; P < 0.001; 95% CI = [−0.25, −0.14]) and increased support for both peace (indirect effect = 0.29; s.e. = 0.04; Z = 7.12; P < 0.001; 95% CI = [0.21, 0.38]) and reintegration policies (indirect effect = 0.25; s.e. = 0.04; Z = 7.08; P < 0.001; 95% CI = [0.18, 0.32]), even when accounting for potential alternative affective pathways via empathy and prejudice. Empathy also independently mediated the effects of the FARC-Integration media intervention (dehumanization: indirect effect = −0.01; s.e. = 0.01; Z = −2.00; P = 0.046; 95% CI = [−0.02, 0.000]; support for peace: indirect effect = 0.02; s.e. = 0.01; Z = 2.16; P = 0.031; 95% CI = [0.002, 0.04]; support for FARC reintegration policies: indirect effect = 0.04; s.e. = 0.02; Z = 2.24; P = 0.025; 95% CI = [0.01, 0.08]). Of note, however, these indirect effects via empathy were significantly weaker than those via the belief about FARC members’ unwillingness and inability to change (dehumanization: indirect effect diff. = −0.18; s.e. = 0.03; Z = −6.88; P < 0.001; 95% CI = [−0.23, −0.13]; support for peace: indirect effect diff. = 0.28; s.e. = 0.04; Z = 7.33; P < 0.001; 95% CI = [0.20, 0.35]; support for FARC reintegration policies: indirect effect diff. = 0.21; s.e. = 0.03; Z = 7.10; P < 0.001; 95% CI = [0.15, 0.27]). Again, prejudice did not significantly mediate the effects of the intervention on any of the outcome measures (dehumanization: indirect effect = −0.02; s.e. = 0.02; Z = −0.97; P = 0.335; 95% CI = [−0.06, 0.02]; support for peace: indirect effect = 0.001; s.e. = 0.002; Z = 0.81; P = 0.419; 95% CI = [−0.002, 0.01]; support for FARC reintegration policies: indirect effect = 0.001; s.e. = 0.001; Z = 0.58; P = 0.561; 95% CI = [−0.002, 0.003]). The results of Study 2 therefore replicate the results of Study 1, demonstrating that the FARC-Integration media intervention reduces the dehumanization of FARC members and (marginally) increases support for peace and reintegration, primarily through a specific psychological mechanism: by challenging the belief that FARC members are unwilling and unable to change—a cognitive barrier to conflict resolution—more so than by improving affect.

Fig. 1: Multivariate path analysis showing that beliefs about FARC members’ unwillingness and inability to change mediate the effect of condition (FARC-Integration v.1 versus the control) on dehumanization, support for peace and support for FARC reintegration policies in Study 2, controlling for empathy and prejudice.
figure 1

The model is fully saturated, and the coefficients are standardized. Coefficients in parentheses represent total effects. Pathways in grey are not significant. Intercorrelations between mediators and between outcome variables are not shown. *P < 0.05; **P < 0.01; ***P < 0.001.

Full size image

Study 3

Studies 1 and 2 provided evidence for the robustness of the FARC-Integration video, its effectiveness over time and the mechanism by which it operates. Still, the outcome measures in those studies were primarily attitudinal and focused on broader social policy support rather than practical actions or decisions that everyday non-FARC Colombians might themselves take to promote the reintegration of FARC ex-combatants. In our final preregistered experiment (Study 3), we further investigated the robustness of the FARC-Integration video and extended our consideration of outcome measures to include (1) a more concrete measure of support for reintegration (that is, expressed willingness to hire job candidates whose résumés indicated a prior affiliation with FARC) and (b) a measure of behaviour (that is, an improved version of the donation measure assessed in Study 2). We focused our data collection exclusively on the FARC-Integration video and a control condition to maximize the sample size available to us (and thus our power to test effects). Moreover, whereas the control condition in prior studies was an ‘empty’ control (that is, the participants watched no video), in Study 3, we included a neutral video (about eco-tourism in Colombia) to further increase the comparability of the experimental design across conditions.

The willingness to hire FARC ex-combatants is a key element of successful reintegration efforts and therefore of a sustainable peace. To assess this, we provided the participants with résumés of six potential ex-combatant job applicants, each for consideration for a different job (for example, taxi driver, bank teller or nurse; Methods). The participants rated these applicants’ suitability as potential hires for each job (that is, the extent to which the participants were willing to hire them and would trust them to do a good job; α = 0.85). The donation measure we used in Study 2 had only three response options, limiting variability. In Study 3, we gave the participants the opportunity to donate whatever exact percentage of their bonus (from 0% to 100%) they were willing to donate to an organization supporting FARC ex-combatants (the participants could keep any money not donated). We also changed the charity to Tejiendo Paz (Weaving Peace), an organization of FARC ex-combatants that manufacture clothes to promote peace. The participants were asked to donate to a specific cause that Tejiendo Paz was working on during the coronavirus pandemic (that is, manufacturing masks to donate to those in need).

Consistent with our preregistered hypotheses, and in line with the conclusions of Studies 1 and 2, we found using a (fully saturated) multivariate path analysis that the FARC-Integration media intervention (n = 705) increased participants’ assessment of suitability of FARC ex-combatants to the jobs compared with the control group (n = 701; mean = 5.02 and s.d. = 1.01 versus mean = 4.57 and s.d. = 0.91, respectively; Z = 8.84; P < 0.001; d = 0.47; 95% CI = [0.36, 0.56]). Moreover, participants in the FARC-Integration condition donated more money to the FARC ex-combatant organization, Tejiendo Paz, than those in the control group did (mean = 48.41 and s.d. = 36.33 versus mean = 43.63 and s.d. = 36.61, respectively; Z = 2.46; P = 0.014; d = 0.13; 95% CI = [0.03, 0.24]; note that the distribution of donations was non-normal, with peaks at 0%, 50% and 100%). Repeating our analysis using bootstrapping procedures, which do not assume normality, confirmed our conclusions. The results of Study 3 therefore conceptually replicate and extend the results of Studies 1 and 2, showing that the FARC-Integration media intervention increases support for a concrete action (hiring ex-combatants) critical to successful reintegration, as well as shaping individuals’ actual financial support of an organization trying to promote peace and reintegration in Colombia.

Discussion

We demonstrate that exposure to a media intervention developed in collaboration between scientists and filmmakers reduces psychological barriers to peace among a populace riven by generations of violent internal conflict. Adapting a bottom-up paradigm used in research on persuasion in the context of health32, we identified a prime psychological target for intervention that was commonly held among non-FARC Colombians and closely associated with opposition to peace: the belief that members of the FARC are unwilling and unable to renounce violence and reintegrate into Colombian society. Building on a literature highlighting the benefits of parasocial contact21,26, we created a video that provided non-FARC Colombians—most of whom have never directly interacted with members of the FARC—with information highlighting demobilized FARC members’ strong desire and demonstrated ability to integrate. We combined this with corroborating perspectives from ingroup members highlighting their own experiences and observations.

The results from three studies confirmed the effectiveness of this intervention. We found that exposure to this video reduced the dehumanization of FARC ex-combatants and increased support for peace and reintegration policies among non-FARC Colombians. Confidence in the robustness of the intervention is bolstered by the fact that its effects persisted 10–12 weeks post-exposure and were broadly replicated in two independent samples of non-FARC Colombians. In Study 3, this included evidence that the intervention changed behaviour—significantly (albeit slightly) increasing donations to an organization that promotes the reintegration of FARC ex-combatants—and increased expressed willingness to hire ex-FARC job applicants, a critical aspect of a successful and sustainable peace. (Of note, it remains unclear whether the intervention’s effects were specific to FARC ex-combatants or also increased willingness to hire non-FARC individuals, a question that future work could consider.) The persistence of the intervention’s effects is particularly striking given that intergroup interventions are rarely shown to endure beyond the intervention period57, although future work should test whether our intervention’s effects last over still-longer periods.

Research on conflict resolution has emphasized the value of both changing cognitive beliefs about an ‘enemy’ outgroup and intervening to improve affect and increase empathy43,44,45. Our intervention builds on prior work on the importance of cognitive beliefs about malleability35,36 by targeting the specific belief that FARC members are willing and able to change, and we therefore expected our intervention to exert its effects primarily through this cognitive mechanism. Still, media interventions inherently include a mixture of content, and even interventions targeting specific beliefs might actually work for reasons beyond those presumed to be manipulated. Here we provide theoretical support for the mechanism we posit: in both Studies 1 and 2, we found that the effects of the focal FARC-Integration video are significantly mediated by reductions in the belief that FARC members are unwilling and unable to change, even when controlling for alternative mediation pathways via increased empathy (for which we find some, but significantly weaker, support) and reduced prejudice (for which we find no support).

We note that we modelled beliefs about FARC unwillingness and inability to change as mediators affecting downstream attitudes and behaviour such as support for FARC reintegration policies. Our data do not allow us to definitively rule out alternative causal orders; nevertheless, it is worth noting from a theoretical perspective that the intervention specifically highlights evidence of FARC ex-combatants being willing and able to change, and it therefore stands to reason that our intervention would operate most proximally by influencing this variable (indeed, our intervention generally had stronger effects on beliefs about unwillingness and inability to change than on the outcome variables, more consistent with our proposed causal order than the reverse). Still, future work should further consider the causal relationship between variables such as the unwillingness and inability to change and empathy on the one hand, and variables such as support for peace and reintegration policies on the other.

Notwithstanding the limits of mediation analyses66, our findings add to conflict resolution research in providing more evidence for the distinct importance of changing cognitive beliefs about outgroups (see the Supplementary Discussion for more discussion of our mediation model). Relatedly, our work adds to growing research suggesting the importance of growth (versus fixed) mindsets and perceived outgroup malleability in intergroup conflict35,36. In particular, our work extends this research beyond the Israeli/Palestinian context, showing the importance of cognitive beliefs such as malleability in internal conflicts (that is, between co-nationals), often marked by especially high rates of recidivism1. This extension is valuable, in part because internal conflicts have certain unique dynamics; for example, individuals might feel additionally aggrieved by the perception that a co-national has ‘betrayed’ the nation or fought their ‘own people’. Moreover, our work extends evidence for the role that malleability can play beyond active conflict to post-conflict settings in which the challenge is moving from a state of peace in principle to a sustainable and lasting peace in practice.

Our work also contributes to intergroup contact research. Whereas most research on contact has identified its effectiveness via mechanisms such as increased trust, reduced anxiety, increased empathy and more inclusive categorization, our work advances the contact literature in suggesting that parasocial contact can be effective through changing cognitive beliefs about an outgroup’s ability to change (that is, malleability)20,21,44,67. Of course, we note that although our intervention worked (as designed) primarily by changing cognitive beliefs about an outgroup’s ability to change, different interventions might well exert equally positive effects via more affective routes. That is, we view cognitive and emotional pathways as complementary mechanisms for increasing support for peace and reintegration.

The FARC-Integration intervention worked by targeting negative beliefs about the outgroup’s perceived intentions and capacities, providing information that challenged the participants’ existing negative views about the contents of FARC members’ minds. Our work thus adds to the literature on meta-perceptions (beliefs about what another group thinks), which shows that meta-perceptions are subject to a negativity bias31,64,68,69 and therefore may be particularly amenable to correction, with positive downstream consequences for intergroup relations31,69. Still, negative meta-perceptions may not always be exaggerated, and it is unclear what effect our approach would have in contexts where negative meta-perceptions are more accurate70. Trying to convince a group that the other side is able and willing to give up violence and reintegrate when they are in fact not stretches the bounds of both plausibility and ethical research; it could also backfire. This highlights the need for diagnostic studies in different contexts to determine the nature of specific meta-perceptions and their accuracy before intervening.

Our work poses several additional interesting questions for future research. For one, it would be valuable for future research to build on our work by exposing non-FARC Colombians to our intervention (and/or other similar interventions) in the field, outside the context of a survey experiment. A particularly compelling test would be to expose non-FARC Colombians in one part of the country but not others to our media campaign, and test whether this intention-to-treat manipulation improved conciliatory attitudes and behaviour (for example, the hiring of ex-FARC combatants) in the treated region. Beyond further increasing the applied value of our work by examining behaviour in a naturalistic setting, this would also bolster our theorizing by showing that an intervention like ours rooted in parasocial contact can ‘withstand’ the risk that those who might at baseline disagree with the message could simply tune it out outside the context of an opt-in experiment (of note, however, we found no evidence that there was differential dropout in our experiment across conditions or that baseline attitudes predicted dropout from the intervention condition; Supplementary Information pp. 13–14).

Future work could also further explore the effect of message order in driving attitude change. We found that the enduring and replicable peace-promoting effects of the FARC-Integration video were directionally the most pronounced for the primary version of the video, which was edited to first include responses by FARC members and then corroborating responses by non-FARC Colombians affirming FARC members’ willingness and ability to change. We thought this version might be the most effective because corroborating information from the ingroup would help resolve any scepticism after exposure to the FARC members’ responses49,50,51. Still, differences between alternative edits of the FARC-Integration video were slight and non-significant (encouragingly, this was because of the strong performance of several of the alternative edits, not the weak performance of the focal edit; Supplementary Tables 6 and 11). Future work could consider more specifically investigating the optimal order and combination of exposure to information intended to change specific beliefs from outgroup versus ingroup sources.

Limitations of our work include the fact that the participants were exposed to our manipulation in the context of a research study as opposed to using a more naturalistic and unobtrusive field design; that we did not examine the effects of our intervention over periods exceeding three months; that we examined only a limited spectrum of behaviour relevant to support for peace and reintegration; and that our samples, although relatively representative, were not random probability samples of the Colombian population.

To conclude, combining methodological techniques from psychology and communications and leveraging scientist–filmmaker collaboration, we developed a media intervention and demonstrate its effectiveness in promoting peace after decades of violent conflict. We show that the intervention succeeds by generating a lasting change in beliefs about the intentions and abilities of outgroup members. Our work thus complements previous peacebuilding interventions in Colombia71,72,73 and in other conflictual contexts that are focused on inducing positive affect12 or reducing conflict-escalating norms20. Practically, this intervention can be scaled up relatively easily and thus has the potential to nudge Colombian society, as well as other societies immersed in conflict, towards more lasting peace.

Methods

The research was approved by the University of Pennsylvania Institutional Review Board (IRB Protocol No. 823959). Informed consent was obtained from all participants. We used an α level of 0.05 for all statistical tests. All test statistics were two-sided. Informed consent was obtained at the beginning of the each of the studies.

Study 1, Wave 1

Participants

We recruited 2,644 non-FARC Colombian participants through an online survey company, Netquest. We removed 326 participants who failed an attention check question, leaving 2,318 participants (mean age = 39.58; s.d. = 14.23; 54.1% women, 45.8% men, 2 participants did not disclose) in the final sample, including at least 169 per group. The participants were assigned to one of 12 conditions at the beginning of the study (see Supplementary Tables 3 and 4 for the number of participants and participant demographics for each condition, as well as links to the videos and video lengths). The survey company rewards participants with credit points, called ‘Korus’, that they can then exchange for products in Netquest’s online shop. Each Korus is equivalent to roughly five to eight US cents. The participants were compensated for their time by the survey company with 22 Korus (roughly US$1.50).

Procedure and measures

The participants were randomly assigned to one of 11 video conditions or to the no-video control condition. Participants in the video conditions watched a five- to seven-minute video at the beginning of the study. The present research focused on a specific video that was created as a collaboration between the researchers and documentary filmmakers (in Supplementary Tables 5 and 6, we provide the complete details on all other conditions and DVs) and that involved interviews with demobilized FARC members and the villagers, police and members of NGOs who had direct contact with the demobilized FARC members (FARC-Integration, 5.5 minutes). To ensure broad representation of FARC members, we obtained permission from the FARC leadership to interview any consenting member in the camp, and we chose interviewees across a range of ages and from both genders. To ensure that interviews were as authentic and independent from each other as possible, interviewees were not allowed to watch other interviews prior to their own and were not provided with the interview questions beforehand. The video aimed to directly counter beliefs commonly held by non-FARC Colombians that FARC members are unwilling and unable to integrate into Colombian society.

This video had four versions that we tested. The first version (v.1) included first interviews with FARC ex-combatants (3 minutes), followed by responses from non-FARC Colombians (that is, ingroup members; 2.5 minutes). As with the FARC interviews, non-FARC interviewees were not allowed to watch other interviews before their own and were not given the interview questions in advance. Again, we ensured a broad demographic spread across interviewee age and gender. The second version (v.2) included the exact same interviews but in the reversed order (that is, non-FARC and then FARC members). The third version (v.3) included only interviews with FARC ex-combatants. The fourth and final version (v.4) included only interviews with non-FARC Colombians.

A second video (FARC-Peace, 5.5 minutes) was created using different sections from the same interviews as the first video, and aimed to directly counter another belief commonly held by non-FARC Colombians about FARC ex-combatants’ honesty, and specifically the sincerity of their commitment to peace. This video also had the same four versions as FARC-Integration, determined by which interviews were presented and in which order (that is, FARC and then non-FARC (v.1), non-FARC and then FARC (v.2), FARC only (v.3) and non-FARC only (v.4)). The third video (FARC-Daughter, 5.5 minutes) was an interview with the 18-year-old daughter of a FARC member who was born in prison, raised from ages 4 to 16 by family in an urban centre and then reunited with her mother after her mother demobilized following the peace treaty in 2017. We interviewed the young woman while we were creating the FARC-Integration and FARC-Peace videos. A fourth video (Colombian Policeman, 5.5 minutes) was an interview with a Colombian policeman who was posted to the region, about how he was welcomed by the FARC members, who helped him complete his studies and pass his high school equivalency exam. The final video (FARC-Trauma, 7 minutes) was created previously by a documentary film crew commissioned by the journal Nature to accompany an article on the peace process in Colombia and the role scientists have in helping the former fighters and victims to heal53. The video, which focuses on two former FARC members who spoke about the trauma they had experienced and remorse for some of their past actions, was never previously tested as an intervention to promote support for peace and the reintegration of FARC ex-combatants. After watching one of the videos, or not for those in the control condition, all participants completed the same survey.

Beliefs about FARC unwillingness and inability to change were assessed by asking the participants to rate how much they agree or disagree with statements endorsing beliefs about the FARC with a five-item scale (α = 0.89) using Likert scales anchored at 1 = not at all to 7 = extremely so. This was a revised version of the items used in the Preliminary Survey (that is, ‘Members of FARC want to impose their norms and laws in the places where they will demobilize’, ‘Members of FARC are unwilling to get along with others in Colombian society’, ‘Members of FARC are unable to get along with others in Colombian society’, ‘FARC is not really interested in peace’ and ‘When demobilized, FARC members will be unwilling to give up their violence and criminality’).

Empathy towards FARC members was assessed using a four-item scale (α = 0.91) derived from Bruneau et al.27 (that is, ‘To what extent do you feel empathy towards ex-combatants who lost their partners due to the conflict’, ‘To what extent do you feel empathy towards ex-combatants whose family reject them because of their past’, ‘To what extent do you feel empathy towards ex-combatants who were disabled by landmines’ and ‘To what extent do you feel empathy towards ex-combatants who cannot find work to support their families’).

Prejudice against FARC members was assessed with feeling thermometers, which asked the participants to report how ‘cold/unfavourable’ (0) to ‘warm/favourable’ (10) they felt towards different groups in random order. Specifically, the participants evaluated 12 groups, including Colombians, FARC members and 10 filler groups (for example, Europeans or Venezuelans in Colombia). A prejudice score was calculated by subtracting the feeling thermometer rating of FARC members from the rating of Colombians.

Dehumanization was assessed with two measures, presented in randomized order. First, we used the ‘Ascent of (Hu)Man’ scale39. The participants were provided with the popular ascent of (hu)man scale and given the following instructions: ‘Some people think that people can vary in how human-like they seem. According to this view, some people seem highly evolved whereas others seem no different than lower animals. Using the image below, indicate with the sliders how evolved and civilized you consider the average member of each group to be.’ The participants evaluated the same 12 groups as in the prejudice measure. The target groups were presented in random order, and the participants made judgements using a sliding scale anchored at the ‘least evolved’ image (0) and the ‘most evolved’ image (100). The ascent dehumanization score was calculated by subtracting the ascent rating of FARC members from the rating of Colombians. Second, trait dehumanization was assessed with the participants’ ratings of how well a series of six traits characterize Colombians (α = 0.80) and FARC members (α = 0.81), with the target group order randomized across participants, using a scale of 1 = not at all to 7 = extremely so. These traits included backward and primitive, savage and aggressive, lacking morals, barbaric and cold-hearted, refined and cultured (reverse coded), and rational and logical (reverse coded). A trait dehumanization score was calculated by subtracting the rating of Colombians from the rating of FARC members. A composite dehumanization score was created by first Z-scoring each measure and then averaging them (r = 0.48, P < 0.001; for a similar approach, see refs. 17,37).

Support for peace was measured with three items (α = 0.73) assessing the extent to which the participants support the peace process with the FARC (that is, ‘I support the peace process’, ‘If we don’t accept a peace deal with FARC, there will be more suffering’ and ‘The peace process must be stopped’ (reverse coded)). In the Preliminary Survey and in Study 1, support for peace and FARC reintegration policies were measured with a few additional items that were removed from the final scales as they lowered the scales’ reliabilities. One of these items was also measured in Study 2 but was excluded for consistency, although including it did not affect the Study 2 conclusions (for the complete details, see the Supplementary Information).

Support for FARC reintegration policies was measured with three items (α = 0.63) assessing the extent to which the participants support policies that help FARC ex-combatants reintegrate into Colombian society (that is, ‘Invest public resources in job training programs for demobilized members of the FARC financed by the government’, ‘Create public scholarships specifically for ex-combatants of FARC and their families with public funds’ and ‘Support policies that stipulate that demobilized FARC members must live in a restricted location outside of urban limits’ (reverse coded)).

Study 1, Wave 2

Participants

Approximately 10–12 weeks after the completion of Wave 1 of Study 1, we attempted to reassess all 2,318 participants who took part in that study. Eventually, we managed to recruit 1,608 non-FARC Colombian participants. We removed 104 participants who failed an attention check question, resulting in 1,504 participants (64.9% retention rate; mean age = 41.30; s.d. = 15.75; 51.3% women, 48.7% men) in the final sample, including at least 110 per group (see Supplementary Tables 8 and 9 for n values and demographics for each condition). The participants were compensated for their time by the survey company with 12 Korus (roughly US$0.80).

Procedure and measures

The participants were asked to complete a survey. They were not explicitly made aware of the connection between the survey and the videos they had been asked to watch 10–12 weeks earlier. Similar to Wave 1, in the analysis we focused on the comparison between the effects of FARC-Integration v.1 and the no-video control condition on key outcome measures of prejudice, empathy, beliefs about FARC unwillingness and inability to change, dehumanization, support for the peace process and support for FARC reintegration policies. In Supplementary Tables 10 and 11, we provide the complete details on all other conditions and DVs.

Beliefs about FARC unwillingness and inability to change were assessed by asking the participants to rate how much they agree or disagree with statements endorsing anti-FARC beliefs using a scale of 1 = not at all to 7 = extremely so with seven items (α = 0.94). This was a revised version of the items we used in Study 1, Wave 1, with two additional items (that is, ‘Members of FARC are unwilling to integrate into Colombian society’ and ‘Members of FARC are unable to integrate into Colombian society’). The rest of the items were identical to the items in Wave 1.

Empathy towards FARC members (α = 0.92) was measured with the same items as in Wave 1.

Prejudice against FARC members was measured and computed in the same way as in Wave 1.

Dehumanization (Colombian trait dehumanization α = 0.81; FARC trait dehumanization α = 0.82; correlation between the ascent difference score and the trait dehumanization difference score: r = 0.48, P < 0.001) was measured and computed in the same way as in Wave 1.

Support for peace was measured with the same three items we used in Wave 1, to which we added four additional items (that is, ‘I think the current Colombian government should maintain the peace process with FARC’, ‘I think the current Colombian government should make the peace process with FARC a high priority’, ‘The peace process with FARC is very important to me’ and ‘The time is right to continue to maintain the peace process with FARC’; α = 0.91).

Support for FARC reintegration policies was measured with the first item from the previous study, to which we added four new items, in an attempt to create a more reliable scale (that is, ’Force demobilized members of the FARC to live in restricted location outside urban limits’ (reverse coded), ‘Provide demobilized FARC members and their families with opportunities for education’, ‘Provide FARC members with access to treatment for psychological trauma’ and ‘Ban FARC members from well-paying jobs’ (reverse coded); α = 0.75).

Study 2

Participants

We recruited 2,429 non-FARC Colombian participants through an online survey company, Netquest. We removed 72 participants who failed an attention check question, resulting in 2,357 participants (mean age = 36.93; s.d. = 13.37; 55.2% women, 44.5% men, 6 participants did not disclose) in the final sample, including at least 451 per group. The participants were assigned to one of five conditions at the beginning of the study (see Supplementary Tables 13 and 14 for n values and demographics for each condition). The participants were compensated for their time by the survey company with 22 Korus (roughly US$1.50).

Procedure and measures

The participants were randomly assigned to one of four video conditions or to the no-video control condition. The videos were those that were shown to be the most effective in Wave 1 of Study 1. Similar to Study 1, in the analysis we focused on the comparison between the effects of FARC-Integration v.1 and the no-video control condition on beliefs about FARC unwillingness and inability to change, the alternative affective mediators of empathy and prejudice, and the key outcome measures of dehumanization, support for the peace process and support for FARC reintegration policies. In Supplementary Tables 15 and 16, we provide the complete details on all other conditions and DVs.

In Study 2, we attempted to separate the measure we used for beliefs about FARC unwillingness and inability to change into its two main components—that is, beliefs about FARC unwillingness to reintegrate and give up violence, and beliefs about FARC (in)ability to reintegrate and give up violence (or perceived malleability).

Beliefs about FARC unwillingness to reintegrate and give up violence were assessed by asking the participants to rate how much they agree or disagree with statements endorsing anti-FARC beliefs using a scale of 1 = not at all to 7 = extremely so with eight items pertaining to beliefs that are specifically about FARC members’ unwillingness to reintegrate into Colombian society and give up violence to pursue peace. In addition to four items used in previous studies (that is, ‘Members of FARC want to impose their norms and laws in the places where they will demobilize’, ‘Members of FARC are unwilling to get along with others in Colombian society’, ‘FARC members are not really interested in peace’ and ‘When demobilized, FARC members will be unwilling to give up their violence and criminality’), we added four new items (all reverse coded; that is, ‘Members of FARC are willing to adopt the norms and practices of the rest of Colombian society’, ‘FARC members are motivated to be accepted and integrated into Colombian society’, ‘FARC members are eager to find a peaceful resolution to the conflict’ and ‘FARC members want peace’; α = 0.92).

Perceived malleability of FARC members was measured with an eight-item scale, assessing the extent to which the participants perceive that FARC members have the ability to change over time36,60 (for example, ‘FARC members can do things differently, but the important parts of who they are can’t really be changed’; α = 0.91).

While our initial intention was to test these three scales separately, factor analysis and intercorrelation (r = 0.85, P < 0.001) indicated that they are part of one scale assessing the participants’ beliefs about FARC unwillingness and inability to change (eigenvalue = 9.32; percentage of explained variance = 58.25%; α = 0.95), similar to Study 1.

Empathy towards FARC members (α = 0.91) was measured with the same items as in Study 1, Wave 2.

Prejudice against FARC members was measured and computed in the same way as in Study 1. The only difference was that prejudice was measured on a 0–100 scale.

Dehumanization (Colombian trait dehumanization α = 0.83; FARC trait dehumanization α = 0.84; correlation between the ascent difference score and the trait dehumanization difference score: r = 0.42, P < 0.001) was measured and computed in the same way as in Study 1. The only difference was that trait dehumanization was measured on a 0–100 scale.

Support for peace (α = 0.95) and support for FARC reintegration policies (α = 0.79) were measured with the same items as in Study 1, Wave 2.

Donation to a non-profit organization that promotes the reintegration of FARC ex-combatants was assessed by providing the participants with a bonus at the end of the study (of 12 Korus, roughly US$0.80) and then providing them the opportunity to keep all of the bonus to themselves, donate half or donate all. The organization, Remando por la Paz (Rafting for Peace), is a tourism project located in the Caquetá region in south Colombia. It was formed by a group of FARC ex-combatants that joined forces with local residents to become rafting guides and open this tourist attraction.

Study 3

Participants

We recruited 1,549 non-FARC Colombian participants through an online survey company, Netquest. We removed 143 participants who failed an attention check question, resulting in 1,406 participants (mean age = 40.41; s.d. = 13.75; 55.1% women, 44.9% men, 1 participant did not disclose) in the final sample. The participants were assigned to one of two conditions at the beginning of the study (see Supplementary Table 18 for the sample size and demographics for each condition). The participants were compensated for their time by the survey company with 17 Korus (roughly US$1.00).

Procedure and measures

The participants were randomly assigned to either the FARC-Integration v.1 video condition or a control condition. The FARC-Integration video was identical to the video we used in Studies 1 and 2, with one minor modification. Specifically, whereas the first 20 seconds of the original video—which includes footage of the road to the demobilization camp alongside introductory text—had no audio, we added neutral music to the revised video (https://player.vimeo.com/video/460327543). Participants in the control condition were asked to watch a short (4.75-minute) video about eco-tourism in Colombia (for the control video, see https://player.vimeo.com/video/471499181). We examined the effect of the FARC-Integration video (versus the control) on the participants’ assessment of potential FARC ex-combatant applicants’ suitability to different jobs, and their actual donation to an organization that aims to promote peace and help ex-combatants reintegrate into Colombian society.

Assessment of job applicants’ suitability was assessed as follows. The participants were shown six different job descriptions (bank teller, cashier at a supermarket, taxi driver, nurse, farmer and factory line worker) in a randomized order and were asked to imagine themselves as the person in charge that needs to hire a new employee. They received one résumé for each job description (six résumés altogether). The résumés were modelled around real résumés that we obtained from FARC ex-combatants and used a standard form that is common in Colombia. In addition to some general information about the applicant, the résumés included a sentence indicating that the applicant is a beneficiary of a programme that is run by the Colombian National Agency of Reintegration (an indirect way to inform the participants that the applicant is an ex-combatant). After each résumé, the participants were asked to indicate the extent to which they trust the applicant to do a good job (from 1 = not at all to 7 = very much) and the likelihood that they would hire this applicant (from 1 = not likely at all to 7 = highly likely). These two items were highly correlated for each of the job descriptions (all r > 0.88, all P < 0.001), and thus, as preregistered, we averaged these two items for each of the job descriptions. We then averaged the ratings across all six jobs to create one composite measure for applicant assessment (α = 0.85).

Donation to an organization that promotes the reintegration of ex-combatants into Colombian society was measured as follows. The participants received the following prompt (translated into Spanish): ‘As a thank you for completing this survey, we will be giving you a bonus! It is up to you how you use it. You can either accept the bonus as 12 Korus points [roughly US$0.80], or you can donate your bonus instead to a non-governmental organization (NGO) that helps reintegrate former FARC combatants, Tejiendo Paz…[here we provided the participants with a description of the NGO and the specific project they would be donating to—that is, manufacturing masks to donate to those in need]. Please indicate below what percent of the money you would like donate to the re-integration program, and what percent of the money you would like to keep for yourself—we will then base our contributions on your decision. Please make sure that the choices add up to 100%!’ The responses were made on a scale of 0 to 100.

Reporting Summary

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