Food for Thought: Investigating Communication Strategies to Counteract Moral Disengagement Regarding Meat Consumption

Eating less meat would benefit environmental sustainability, human health, and animal welfare. Providing information about this, however, does often not lead people to adopt according beliefs, attitudes or behaviors. In fact, dietary changes are often prevented by dissonance reduction (i.e., moral disengagement) if information elicits a conflict regarding meat. In the present investigation we thus aimed to address moral disengagement via a communication strategy that consisted of two stages: In Stage I, we presented information by showing distressing scenes from animal agriculture. In Stage II, we then counteracted moral disengagement in a dialog. Two studies indicate that, following the dialog, people’s evaluations of meat changed and their willingness to eat meat decreased; this seemed to result from lowered moral disengagement. By providing an empirically tested communication strategy for addressing moral disengagement on the exemplary conflict regarding meat, we hope to inspire research and interventions that intend to communicate (environmental) issues.

sized German city. Every member of the group wears black or dark clothing 5 , and four to eight activists stand in a square formation, one or two on each side. While standing there, they are holding laptops in their hands, showing scenes from animal agriculture, such as baby chickens that are ground up alive, cows that are slaughtered, and pigs in small enclosures with injuries (see example videos on https://www.youtube.com/channel/UCihWowxl98t9CECAtsn-RIQ; SOKO Tierschutz, n.d.).
In Stage I, the activists waited in this formation for pedestrians to stop and watch the videos. Usually, if pedestrians seemed to be interested, one activist, who does not carry a laptop, approached them and started a dialog in Stage II. In regard to our investigation, however, two procedures were possible if a pedestrian was interested: In the pre-dialog condition, an investigator approached the interested pedestrians first and they were asked to give informed consent and to fill out a survey before they talked to an activist; in the post-dialog condition, the pedestrians talked to an activist first before they were approached by an investigator who invited them to participate in the study, i.e., give informed consent and fill out the same survey.
In Stage II, the dialog was conducted by experienced activists in a semi-standardized way.
The activists usually started a dialog with three specific questions: First off, they asked whether the participants had seen such footage before; second, they asked what the bystanders felt when viewing the videos; lastly, they asked whether they knew a solution for these issues. Following the questions, the dialog roamed freely, typically addressing people's reservations about reducing their meat consumption, i.e., moral disengagement strategies (e.g., "eating meat is part of a healthy diet", "meat tastes so good, I could never give it up"; see Piazza et al., 2015).
To engage in these dialogs, activists are trained to talk to pedestrians, and identify and counteract moral disengagement strategies (see supplemental materials for a catalog providing scientific arguments to counteract specific moral disengagement strategies, which was assembled prior to Study 2). That is, prior to reaching out to the pedestrians, new activists are informed about the intervention techniques and arguments to counter reservations. Subsequently, they have to observe experienced activists during the dialog (e.g., while presenting videos). If they engage in their first dialogs, experienced activists will supervise them. Later on, there also exists the possibility to get supervision, and there is an online forum in which the activists give advice to each other.

Materials.
The complete survey for participants 6 consisted of visual analogous scales and participants were asked to make a cross on a 100 mm long line to indicate their approval of the different statements; thus, all scales ranged from 0 to 100.
Ambivalence. First, we assessed objective ambivalence (OA), which refers to a structural existence of conflicting evaluations regarding one attitude object (van Harreveld, Nohlen, & Schneider, 2015), by asking participants to indicate how positive (P) meat consumption is for them, and how negative (N) meat consumption is for them on two separate scales; a unified OA score can be computed from these two scales using the formula (P + N)/2 -| P -N | (Povey, Wellens, & Connor , 2001). Following, we assessed subjective ambivalence (SA), which refers to the actual experience of the conflicting evaluations (van Harreveld et al., 2015), by asking for agreement on the following statement "I feel torn about the two sides of eating meat" (Berndsen & Van der Pligt, 2004); in Study 2 we added a second item ("I am uncertain if I should eat meat, because I can't decide between one side and the other"), and computed a mean SA score out of both items (Berndsen & Van der Pligt, 2004).
Willingness for Dietary Change. Subsequently, we assessed whether participants were willing to change their diets via four statements: "Instead of eating a dish containing meat, I would choose a vegetarian (vegan) dish" (Kunst & Hohle, 2016), as well as "I could imagine to adopt a vegetarian (vegan) diet in the future" (adopted from Berndsen & Van der Pligt, 2004).
Out of these statements, we computed a mean score (α = .73), indicating participants' willingness to change their diets.
Emotions (Only Study 1). In the following, we asked participants about certain emotions aroused by the slaughtering for meat production, and they indicated their agreement on the following statements "When I think of the animals that were killed for our consumption, I feel compassion (anger, indifference, disgust, dismay)".
Moral Disengagement (Only Study 2). Instead of asking for emotions, we added two questionnaires on specific moral disengagement strategies. At first, we asked people if they believe that animals would possess certain mental (i.e., self-control, morality, memory, and planning, α = .67) and emotional (i.e., happiness, melancholy, excitement, guilt, and panic; α = .72) capacities (Rothgerber, 2014); and their agreement with different statements rationalizing meat consumption as nice, necessary, normal and natural (α = .85; Piazza et al., 2015). These questionnaires were inserted into the survey prior to the ambivalence measures.
Demographics. Finally, we assessed demographic variables (age, sex, profession, consumption of meat and fish, and consumption of other animal products), and we asked participants whether they had a pet (and if so, which pet).

Results
All analyses were conducted with SPSS for Windows (Version 25); mediation and moderation analyses were calculated via PROCESS macro 3.3 by Andrew Hayes (2018), which is a modeling tool for different statistical moderation and mediation models. As some participants did not complete every single item in the survey, they were excluded from analyses including these items if no mean score could be computed (Study 1: One participant did not complete the evaluations for OA. Study 2: Two participants did not complete the evaluations for OA from which one participant did also not complete the SA and willingness to change items; wo participants did not fill out the questionnaire on mind and emotion attribution).

Study 1: Investigating the Role of Emotions
Ambivalence. First, we tested whether people's ambivalence towards meat increases following the dialog by conducting two separate one-way analyses of variance (ANOVA) on SA and OA. An ANOVA allows to detect statistical differences between the means of two or more conditions (see Tabachnick  . Consequently, the indirect effect (ab = 6.15, 95% CI = [1.54, 11.81]) suggests that the change in evaluations, due to the dialog, led to heightened willingness to eschew meat; in fact, independent of the indirect effect, the dialog did not seem to affect people's intentions (c' = 6.15, 95% CI = [-3.02, 15.32]). This implies that people were more willing to alter their diet after Stage II because they changed their evaluations of meat.
Emotions. Finally, we assessed whether people who experienced stronger emotions when watching the videos were more likely to be willing to reduce their meat consumption after the dialog. To test this hypothesis, we used moderated mediation analyses (Model 7; Hayes, 2018), which also examine the direct and indirect effects, but they further test whether an indirect effect depends on a moderator variable (see Hayes, 2018). We did not find a significant moderation effect that would indicate that the effectiveness of the dialog is conditional on the experienced emotions of compassion, anger, indifference, disgust, or dismay (all ps > .128). On the contrary, the indirect effect of the dialog on the willingness to change mediated via evaluations of meat remained constant in all of these analyses. Thus, the effectiveness of the intervention was not restricted to people who experienced a higher degree of emotions aroused by the videos.

Study 2: Investigating the Role of Moral Disengagement
Manipulation Check. First, we checked whether the dialog successfully changed participants' use of moral disengagement strategies. Therefore, we screened for outliers on these variables via Mahalanobis distance with p < .001; this way, one outlier was identified and thus excluded from subsequent analyses 7 . Looking separately at the moral disengagement strategies . Adopting a two-stage intervention strategy, we found in two studies that counteracting moral disengagement increased participants' willingness to reduce their meat consumption. Thus, our findings highlight the incremental value of counteracting moral disengagement and stress that two-stage intervention strategies are superior to mere information campaigns (e.g., by confronting people with animal welfare issues).
These findings align with research that shows that people do not just adapt their attitudes, beliefs and behavior if they are informed about scientific evidence (Hornsey et al. 2016;Hart & Nisbet, 2012). Going beyond mere information, it seems to be necessary to also account for processes of motivated reasoning, for example, in the form of moral disengagement (Bastian & Loughnan, 2017;Hart & Nisbet, 2012). Thus, by providing a two-stage template of how to address moral disengagement, the current investigation complements many studies that investigate how to inform people about scientific evidence and motivate them to alter their behavior (e.g., Bianchi et al., 2018;de Boer, Schösler, & Boersema, 2013). In fact, our research suggests that it is necessary to discuss information and consider people's perspectives as well as reservations-ideally in a face-to-face setting with an expert (Nisbet & Scheufele, 2007). For this endeavor, we hope that the assembled argumentation catalog may be a starting point for researchers, lay people, and activists who want to study and counteract moral disengagement regarding meat consumption.
Going beyond meat consumption, moral disengagement may also be highly important regarding other harmful behaviors. In fact, moral disengagement seems to play an important role underlying and maintaining many forms of unsustainable actions (Bandura, 2007). Atkinson and Kim (2015) showed, for example, that people diffuse their responsibility and denied harm when purchasing plastic packed products. People argued, for instance, that they had no choice but to buy the product; or they reason that a purchased product is not as detrimental for the environment compared to a worse one. Such statements reflect similarities with reasons people produce to resolve the dissonance resulting from the meat paradox (Bastian & Loughnan, 2017).
The usefulness of our dialogue-based approach is supported by the fact that we gathered representative and heterogeneous samples for an industrialized, western population, including a similar amount of men and women with a wide age range, and a diverse professional background (e.g., students, academics, craftsmen, service providers, and pensioners). This is essential because people in industrialized western countries typically eat a high amount of meat (Tilman & Clark, 2014). Similarly, many environmental issues are mainly driven by these countries (Wei et al., 2016), which makes their inhabitants an important target group for intervention campaigns. For this target group, our findings suggest that the used intervention strategy is effective, and counteracting moral disengagement therefore should not be restricted to specific demographics.
Thus, researchers and activists can confidently target a variety of recipients using this template.

Limitations and Future Studies
It is important to note that the arguments seemed to be mostly effective in increasing emotional capacities that people attribute to animals; in fact, there were only descriptive differences between the conditions regarding attributions of animals' mental capacities and rationalizations towards meat. It may be speculated that the profound effect of the dialog on attributions of animal emotion may result from a synergy between the videos and the strategy of denying harm. In fact, this specific moral disengagement strategy revolves around animals' capacity to suffer. That is, when seeing these animals suffer, people might be more motivated to deny inflicted harm than to diffuse their responsibility. This may have rendered specific arguments during the dialog more effective that revolve around animals' emotional capacities than around rationalizations of meat consumption. Although this does not explain why we did not find an effect on the attributions of animal mental capacities, previous research also showed that providing information on the intelligence of livestock (i.e., pigs) does not increase moral concern towards these animals (Piazza & Loughnan, 2016). Thus, informing about emotional rather than mental capacities of animals seems to be a more promising lever for interventions. Nonetheless, we aimed to improve the catalog, and included additional arguments to address rationalizations of meat consumption more effectively (i.e., grey-colored arguments; see supplemental materials).
Future studies should examine the effectiveness of these arguments, and scrutinize whether arguments that intend to counter rationalizations of meat consumption are especially effective after raising people's awareness about environmental or health instead of animal welfare issues.
In addition, it has to be acknowledged that we assessed participants' willingness to change only at one point in time, and that we did not assess actual meat consumption. Although these intentions do not necessarily translate into behavior (Sheeran, 2002), recent research showed that people who intend to reduce their meat consumption after an intervention often achieve this goal (Amiot, Boutros, Sukhanova, & Karelis, 2018). Nonetheless, researchers may aim to overcome this intention-behavior gap by using intervention tools, like mental contrasting and implementation intentions, which have already been used successfully to reduce meat consumption (e.g., Loy, Wieber, Gollwitzer, & Oettingen, 2016;Rees et al., 2018). These tools work by clearly defining individual goals and courses of action to initialize behavior more automatically. This individualized approach could also help to take the different situations of recipients into account, for example, by aiming at realizing meat-free days instead of eating smaller portions of meat or vice versa (de Boer, Schösler, & Aiking, 2014). Future studies may show that combining these tools and simultaneously addressing moral disengagement could improve the effectiveness of intervention strategies that aim at reducing meat consumption. To do so, it would be advisable to conduct longitudinal studies on people's moral disengagement strategies, willingness to reduce their meat consumption, and their dietary behavior to scrutinize if the effects last over time.
Lastly, it is important to communicate information serving as pull factors that motivate people to eat plant-based food next to push factors for eschewing meat (de Boer & Aiking, 2017).
Thus, the activists provided recipes and point of sale information during the dialog if participants reasoned that meat eating is too nice to quit. These pull factors should frame plant-based food as savory, tasty, and pleasurable (de Boer & Aiking, 2017). For instance, research showed that it may be highlighted that novel culinary experiences may be derived from plant-based dishes of other cultures being as savory and enjoyable as meat dishes (Schösler, de Boer, & Boersema, 2012). Going beyond mere communication, however, opportunities have to be created so that people make explicit positive sensory experiences with meatless food (de Boer & Aiking, 2017).
Therefore, the availability of vegetarian options in everyday life can be increased. Recent studies showed, for instance, that increasing the proportion of vegetarian dishes (i.e., offering 50% instead of 25% of total dishes without meat) leads people to choose more of those dishes without affecting overall sales (Garnett, Balmford, Sandbrook, Pilling, & Marteau, 2019). Similarly, one might encourage restaurants, cafeterias, or hospitals-and individuals-to implement at least one meat-free day per week (Laestadius, Neff, Barry, & Frattaroli, 2013). In this sense, it is promising to see that, for example, the Meatless Monday campaign has been highly successful (Euromonitor International, 2011), being implemented already in 40 countries (Meatless Monday Campaigns, 2020). Nonetheless, more research is necessary to examine how to best communicate information that motivates people to eat plant-based food so that they experience it as positive.

Conclusion
In the face of man-made crises like global warming, the loss of biodiversity, the rise of civilization diseases like cancer or diabetes, and severe animal welfare issues, it is mandatory to understand how people maintain harmful behaviors that contribute to these issues. Drawing on meat consumption as prime example of harmful behavior, we outlined that moral disengagement strategies help people to maintain their behavior. This suggests that merely providing information about problematic behavior is not sufficient to elicit behavioral change (Bastian & Loughnan, 2017). Indeed, our research shows that intervention strategies need to specifically counteract moral disengagement to effectively motivate people to change. We hope that the current investigation will thereby improve our knowledge about these strategies, and outline courses of action in the domain of meat consumption and other harmful behaviors.

Foot Notes
1 We use the term animals if we do not refer to human animals.
2 Note that other food of animal origin is also linked to environmental, health and animal welfare issues.
3 Six adolescents (age 13-16) also wanted to take part in the study; however, due to data privacy protection, we only analyzed the data of adults. 4 We also conducted an a priori power analysis via G-Power 3.1 (Faul, Erdfelder, Lang, & Buchner, 2007) for Study 2 considering the effect size of the significant two-way interaction in Study 1. This analysis resulted in a total sample size of N = 18 (effect size: η p ² = .09; power: 1-β = .95; correlation for repeated measures: r = -.781). Thus, we decided to increase the power based on the feasibility of data collection.
5 For the majority of Study 1, the activists were associated with the activist group Anonymous for the Voiceless and wore white masks to cover their faces as the protocol for the "cube of truth" (Anonymous for the Voiceless, 2019) suggests. At a certain point, though, the activists separated from Anonymous for the Voiceless and stopped wearing masks. Consequently, the activists wore no masks at all in Study 2. The rest of the procedure stayed the same. 7 The inclusion of this person did not severely affect the results; however, the one-way ANOVA on attributions of animal emotion failed to reach conventional criterions of significance, F(1, 96) = 3.05, p = .084. 8 Note that the 95% CI does not contain zero for most bootstrapping samples. Due to the nature of bootstrapping, however, the lower level of the 95% CI may fluctuate and become negative in some cases.    Error bars denote the 95% confidence interval.