Research on awareness of the environmental toll of meat in high-income countries has found that, in general, consumers tend to underestimate or ignore the negative impact that meat has on the environment (de Boer et al., 2016; Hartmann et al., 2021; Pohjolainen et al., 2016; Slotnick et al., 2022). Individuals in high-income countries often perceive reducing plastic use and saving energy at home as more effective at combating climate change than plant-based diets (Campbell-Arvai, 2015; Slotnick et al., 2022; Wynes and Nicholas, 2017). However, research indicates that recycling is four times less effective than following a plant-based diet, while changing household lightbulbs is eight times less effective (Wynes and Nicholas, 2017). Additionally, switching from plastic to reusable shopping bags is less than 1% as effective as a year without eating meat (Wynes and Nicholas, 2017).

Regarding awareness of red meat’s health risks, consumers seem similarly misinformed. Grummon et al. (2021) found that 46% of US parents were not aware of any health risks associated with red meat, including heart disease, stroke, type 2 diabetes, and cancer. Awareness was highest for the association between red meat and heart disease (28%) and lowest for its link to prostate cancer (8%) (Grummon et al., 2021). Misconceptions about nutrition also contribute to resistance toward vegan and vegetarian diets, with some consumers believing that meat provides essential nutrients that cannot be adequately replaced by plant-based foods (Collier et al., 2021; Corrin and Papadopoulos, 2017; Loiselle et al., 2024). However, research has demonstrated that well-planned vegan and vegetarian diets can achieve nutritional adequacy comparable to that of omnivore diets, with significantly lower carbon footprints (Alcalá-Santiago et al., 2025). Moreover, vegan and vegetarian diets are associated with reduced risk of illness, including type 2 diabetes, ischemic heart disease, cancer, and all-cause mortality (Oussalah et al., 2020; Selinger et al., 2023).

It is possible that these knowledge gaps regarding meat’s environmental and health risks stem from more than a simple lack of information. Individuals often avoid messages that challenge their existing attitudes, behaviors, or beliefs, a tendency known as selective exposure bias (Festinger, 1957). This bias may limit the effectiveness of information-based interventions, as discussed further in Section 1.4.

Meat attachment refers to an individual’s positive bond toward meat consumption (Graça et al., 2015a; Graça et al., 2015b). According to Graça et al. (2015a), the construct consists of four dimensions: hedonism, affinity, entitlement, and dependence. A person with high meat attachment is likely to find pleasure in, have a liking towards, feel entitled to, and be dependent on eating meat (Graça et al., 2015a). Consequently, stronger attachment is associated with a lower willingness to reduce meat consumption, a pattern which has been supported by several studies (Circus and Robison, 2019; Mei-Fang, 2024; Wang and Scrimgeour, 2021).

The roots of meat attachment can be traced back through humanity’s relationship to meat and other animal products, which, for the majority of human history, have been vital to maintaining the health and survival of populations (Aboelenien and Arsel, 2022). Today, technological advances in high-income societies allow for widespread access to alternative protein sources, supporting balanced vegetarian and vegan diets. However, due to persistent associations between meat and health, many consumers believe that eating meat is natural, normal, or necessary (Joy, 2010; Piazza et al., 2015; Šedová et al., 2016), thereby reinforcing and contributing to meat attachment (Roozen and Raedts, 2022). In addition, meat has historically been linked to masculinity, social status, and celebration (Aboelenien and Arsel, 2022), associations which continue to foster an emotional connection toward meat consumption and make it difficult to shift normative eating habits toward more sustainable diets (Markowski and Roxburgh, 2019; Mertens and Oberhoff, 2023; Pohjolainen et al., 2015).

Being highly attached to meat does not necessarily mean that one cannot recognize its negative consequences. While omnivores may love animals and care about the environment, they simultaneously ignore the ethical and environmental consequences of eating meat (Bouwman et al., 2022; Rothgerber and Rosenfeld, 2021; Šedová et al., 2016). However, confronting the negative impacts of one’s actions can cause cognitive dissonance, a mental discomfort arising from inconsistencies among cognitions (Festinger, 1957). The existence of dissonance will motivate a person to reduce this psychological discomfort through selective exposure, which involves actively avoiding situations and information which may worsen the experience (Festinger, 1957). For example, for a meat eater, engaging with information related to animal agriculture’s negative environmental impact may induce stress, guilt, or anxiety. To avoid these negative feelings, the individual might selectively expose themselves to information that suggests their dietary preferences are healthy or suitable, while avoiding information that suggests they are harmful or wrong. This avoidance of information prevents meat-related risk information from reaching its target audience, reducing the potential for attitude or dietary shifts, challenges discussed in the next section.

Participants began by completing the baseline survey, which included multiple-choice, open-ended, rank-order, graphic slider, and Likert-scale questions. The first set of questions related to participant demographic information, including age and living location (province). The descriptions and ranges of all demographic items followed those used by Statistics Canada. Age group and province were collected first as low-sensitivity, low-burden items intended to orient participants and ease them into the survey before responding to more identity-salient questions related to meat consumption behaviors, attitudes, and attachment. Following these demographic items, participants were asked about their meat consumption behaviors, including how often they eat meat (multiple choice), the type of meat they eat most often (multiple choice), the main reason they eat meat (rank-order and text-entry), and their willingness to reduce the amount of meat they eat (5-point Likert-type item).

The next set of questions addressed participants’ meat attitudes and attachment. To gauge general attitudes toward meat, participants were asked to indicate their feelings about meat using a five-point graphic slider. Graphic scales convey levels of a particular affective domain in a format that can be easier for participants to process and understand than text scales, making them especially beneficial for low-literacy participants (Stange et al., 2018; Toepoel et al., 2019). To determine participants’ levels of meat attachment, this study used a shortened version of the Meat Attachment Questionnaire (MAQ) developed by Graça et al. (2015a, 2015b). The original MAQ includes 16 Likert-type items across four dimensions: hedonism, affinity, entitlement, and dependence. The current study used only 12 items across the four dimensions to reduce survey fatigue. The 12 items were selected based on their corresponding dimensions and factor loadings reported by Graça et al. (2015a, 2015b), with the top three items from each dimension selected. Internal consistency reliability for the shortened MAQ was high (Cronbach’s α = 0.89), indicating that the retained items reliably measured the intended construct.

The last set of questions addressed participants’ perception of the risks and benefits of meat in health and environmental contexts. Using a Likert scale, participants indicated the extent to which they agreed or disagreed that (1) “meat production is harmful to the environment,” (2) “meat production is beneficial to the environment,” (3) “meat-rich diets increase risk of disease,” and (4) “meat is an important part of a healthy diet.” Responses to these questions were averaged to create a composite measure of perceived meat-related risk, with higher scores indicating greater perceived risk (Cronbach’s α = 0.74). Finally, participants were asked to indicate the effectiveness of eating less meat and eating a vegan diet for mitigating climate change, demonstrating their diet-related environmental knowledge (Likert scale).

The baseline measures described above (meat-related behaviors, attitudes, attachment, and perceived risk) were collected prior to the experimental task to establish pre-exposure characteristics relevant to selective exposure. The experimental task followed these baseline measures.

In the experiment section of the questionnaire, participants were randomly assigned to the control or experimental group. In both groups, participants were provided with a list of statements and asked to select one they would like to learn more about. The statements addressed the risks and benefits of meat consumption in health and environmental contexts. After selecting a statement, participants were provided with information about their chosen topic.

Participants in the control group were given four statements to choose from, none of which were framed using information utility. The statements provided to the control group were:

Participants in the experimental group were given the same four statements to choose from, plus two additional statements framed using information utility. The two additional statements provided to the experimental group were:

The four (control group) or six (experimental group) statements listed above were presented in a multiple-choice format and appeared in a randomized order for all participants. Additionally, all of the statements were presented in a plain, written format. Although visuals are often used alongside words in the online media environment, this study did not include images or colors to control for factors that may influence statement selection beyond the statements themselves. After selecting the statement they were most interested in, participants were provided with a short article on the topic, written by the researchers (see Supplementary materials for full articles).

The environment and individual health were selected as the focal topics for the statements above as they are two commonly cited reasons people choose to forgo meat in their diet (Ghaffari et al., 2022; Verain and Dagevos, 2022; North et al., 2021; Janssen et al., 2016). While there are other reasons individuals may choose not to eat meat, such as animal welfare, the topics of environment and health can be discussed from divergent viewpoints, which was an important characteristic needed in this study. For example, eating meat can be argued as a benefit or a risk to health. Because one objective of the study was to determine whether participants engage in selective exposure to meat-related information that aligns with their pre-existing attitudes and awareness, it was important that for each statement in the control group, another existed with an opposing view. This enabled exploration of whether participants selected a pro-meat or anti-meat statement and to what extent this selection aligned with their meat-related attitudes, attachment, and risk perception (as measured in the baseline survey).

In the experimental group, two additional statements were displayed: one related to environmental risks and one related to health risks, each framed using information utility. Information utility increases with the individual’s perceived (1) magnitude of challenges, (2) likelihood of threat occurrence, (3) immediacy of threat occurrence, and (4) efficacy to influence outcomes (Knobloch-Westerwick and Kleinman, 2012). To test whether information utility may be an effective strategy for encouraging meatless diets, the framing technique was applied to meat risk statements rather than to meat benefit statements. Information utility was employed by emphasizing the magnitude of the challenge (“Eating meat plays a big role in global warming”) and the reader’s efficacy in influencing outcomes (“how shifting your diet can reduce your impact”).

The post-experiment survey included multiple-choice and open-ended questions. The first question asked participants why they selected the statement they did (text entry). This question was included as a quality check of participants’ engagement in the experimental task, as well as to provide descriptive insight into the reasoning underlying information choice. The next set of questions asked participants whether they learned anything new (dichotomous item), whether what they read would change how they make dietary decisions in the future (dichotomous item), whether the information altered their attitudes toward eating meat (three-point Likert-type item), and how willing they were to reduce the amount of meat they eat (five-point Likert-type item). These questions were designed to determine whether participants’ attitudes or behavioral intentions changed post-experiment.

The final set of questions addressed participant demographic variables, including urban versus rural residence, gender, and education. These demographic items were collected at the end of the study rather than at the beginning, as they gathered potentially sensitive information that may have influenced participants’ responses to later questions or motivated them to leave the study early.

Pre-experiment analyses were descriptive in nature and were conducted to characterize the sample and baseline measures; no hypotheses were tested at this stage. Here, descriptive statistics were used to explore participants’ demographic information, meat consumption behaviors, attitudes toward meat, meat attachment, awareness of meat risks and benefits, and diet-related environmental knowledge. Chi-square goodness-of-fit tests were conducted to assess whether the sample demographics differed significantly from those in the Canadian census. A Friedman test was also conducted to analyze the primary reasons participants eat meat, as collected via a rank-order question.

Furthermore, thematic analysis was used to examine an open-ended question asking participants to specify the main reason they eat meat. All participants provided at least one reason for eating meat, while others listed two or three. A single researcher coded 514 reasons into the following categories: taste, health, culture, social pressures, and other.

To determine participants’ meat attachment levels, responses to the meat attachment questions were translated into scores using a methodology similar to that described by Dowsett et al. (2018). For each item, participants received a score out of five depending on their answer (with “strongly disagree” equaling a score of one and “strongly agree” equaling a score of five for all items except for those in the affinity dimension, which were reverse-scored). Item scores were then summed to create a total score for each participant, with the lowest possible score being 12 (i.e., 1 point × 12 questions) and the highest possible score being 60 (i.e., 5 points × 12 questions). Individual scores were then grouped into low (12–27), medium (28–44), and high (45–60) attachment categories.

The experimental analysis included both confirmatory tests of selective exposure hypotheses (H1–H2) and exploratory analyses examining individual differences in statement selection (RQ1). To test H1, chi-square goodness-of-fit tests were conducted within the control condition to assess whether benefit-focused and risk-focused statements were selected at equal rates, with effect size estimated using Cohen’s w. To test H2, a chi-square test of independence was used to examine whether the distribution of statement selections differed between the control and experimental groups. Standardized residuals were inspected to identify which categories contributed most to any significant associations, and Cramer’s V was calculated to estimate effect size.

To address RQ1, an exploratory binary logistic regression analysis investigated whether selective exposure to meat-related information (risk vs. benefit) varied as a function of participants’ meat-related attitudes, attachment, and perceived risk, adjusting for relevant demographic variables. Predictor variables included participants’ affect toward meat (graphic slider scale), meat attachment score, and perceived meat-related risk (composite score). Age (grouped by generation), gender (man vs. woman), education level, and urban versus rural residence were included as control variables. These items were included because of their potential to influence statement selection independent of the main predictors. For instance, women tend to score higher on the trait of openness than men (Weisberg et al., 2011), potentially making them more likely to select statements that challenge their pre-existing attitudes and beliefs.

The post-experiment analysis used descriptive statistics to assess whether participants learned anything new and whether the information provided affected their dietary intentions. Additionally, thematic analysis was used to examine an open-ended question about participants’ reasons for selecting their statements. One researcher assigned codes to each response that best reflected its content, allowing responses to be grouped by theme. Responses were coded into the following categories of “general interest in topic/statement,” “lack of knowledge/interested in learning more,” “confirmation of prior knowledge attitudes, and/or beliefs,” “skeptical of information,” “topic appeared legitimate, scientific, or convincing,” “chose randomly/no specific reason,” and “reason unclear.” “Reason unclear” often included responses that summarized the information provided for a specific statement, making it difficult to determine whether participants were providing a reason for their selection or summarizing what they had learned. Overall, examining the descriptive patterns related to statement selection, information learned, and future dietary decisions provided additional context regarding participants’ motivations and self-reported intentions following exposure to information.

To investigate post-exposure attitude and willingness outcomes (RQ2–RQ3), ordinal logistic regression models (proportional odds models) were estimated using a logit link function. Willingness to reduce meat consumption and post-exposure attitude change were treated as ordered categorical outcomes. Information utility framing (no vs. yes), topic domain (environmental vs. health risk), and their interaction were entered as predictors. For the willingness model, baseline willingness was included as a covariate to account for pre-existing differences. Model fit was assessed using likelihood ratio chi-square tests, and the proportional odds assumption was evaluated using the test of parallel lines. These analyses were restricted to participants exposed to meat risk information in the experimental condition, as information utility framing was present only in this condition and was the focal predictor of interest.

Multiple steps were taken to improve the study’s construct validity, including a pilot test, an assessment of convergent validity, and an experiment engagement quality check. In the pilot test, 45 participants were asked to provide feedback on the survey questions and experiment in terms of clarity and experience. None of the pilot study participants provided critical feedback on the study, and 10 provided positive feedback. Furthermore, the present study assessed convergent validity by comparing participants’ willingness to reduce their meat consumption and their meat attachment levels, two variables expected to be strongly correlated. Spearman’s Rho correlation coefficient indicated a significant association between the two variables (p < 0.01), and crosstab analysis confirmed this finding. The strong relationship between the variables provides evidence that the MAQ by Graça et al. (2015a, 2015b), while shortened, still accurately measured participants’ willingness to reduce their meat consumption. Lastly, a quality check was conducted to ensure that participants perceived, understood, and responded to the statement selection process as expected. Following the experiment, an open-ended question asked participants to explain why they selected the statement they did. Only 3% of participants reported selecting a statement at random or having no specific reason for their decision, indicating that most participants were actively engaged in the task.

Several design choices were implemented to increase the study’s internal validity, including varied question types, neutral language, a cover story, careful placement of sensitive demographic questions, random assignment, controls to limit external influences, online administration, and anonymity. To begin, the baseline and follow-up surveys incorporated a variety of question formats to reduce repetition and survey fatigue. Additionally, questions were worded in neutral, nonjudgmental language to avoid moral cues that could elicit social desirability bias. Similarly, a cover story was created to minimize demand characteristics, describing the research as focused on communicating the risks and benefits of meat consumption, without explicit value-laden framing. Furthermore, sensitive demographic information was collected at the end of the survey rather than at the beginning to avoid influencing participants’ responses to later questions or motivating participants to leave the study early.

Randomly assigning participants to the control and experimental groups helped ensure that any observed effects could be attributed to the variables of interest rather than pre-existing differences between participants. The experimental materials were also designed to control for factors unrelated to the wording of the statements, such as visuals or sentence structure, that could otherwise shape participants’ choices. Moreover, conducting the experiment online allowed participants to respond to stimuli in their homes, reducing their vulnerability to systematic biases arising from unfamiliar settings or the experimenter’s presence. Relatedly, administering the study anonymously further mitigated potential social desirability bias by limiting concerns about evaluation or judgment.

The study’s external validity was supported by recruiting a diverse participant sample and by incorporating design features intended to enhance ecological validity. As discussed previously, the sample was identified through a third-party recruitment service, which enabled a wide reach of participants in terms of age, gender, location, and education level, thereby increasing the generalizability of the findings to the Canadian adult population. Additionally, the experiment portion of the study allowed participants to choose statements from a list of several options, resembling free-choice media environments, such as the internet, where individuals can select and avoid content based on their attitudes and preferences.

To increase the study’s reliability, the research designs and methods, as well as survey questions, were based on similar research by Lueders et al. (2022), Gaspar et al. (2016), Knobloch-Westerwick and Kleinman (2012), Graça et al. (2015a, 2015b), and Dowsett et al. (2018). This ensured that the study’s design, instrumentation, and measurements were consistent with those of previous studies.

Additionally, to protect the trustworthiness of the study’s qualitative analysis, reflexivity was applied during coding of open-ended data to acknowledge personal biases and preconceptions that may have influenced coding decisions. While one researcher conducted the study’s qualitative analysis, they frequently consulted other members of the research team when interpreting open-ended responses, thereby minimizing personal bias and enhancing the credibility of the results.

This study expected that participants in the control group would be more likely to engage with statements about the benefits of meat than with those about its risks (H1). This held true for health topics but not environmental topics. Fifty-three percent of participants read about the health benefits of meat, while 21% read about the health risks, indicating a preference for health benefit information over health risk information. This preference was not observed for environmental statements, as an equal number of participants (13%) read about both environmental benefits and environmental risks.

A chi-square goodness-of-fit test confirmed that selection of health information was not equally distributed (χ²(1, N = 132) = 25.49, p < 0.001, w = 0.44). Participants were substantially more likely to select health benefit information than health risk information (OR = 2.57, 95% CI [1.76, 3.76]), indicating a strong preference for benefit-focused health content.

In the experimental group, it was expected that participants would be more likely to engage with information related to meat risks when the statement was framed using information utility (H2). This held true for both environmental and health topics. Eleven percent of participants chose to read about environmental risks framed using information utility, while 8% chose to read about environmental risks without information utility framing. Seventeen percent of participants selected the health risks statement framed using information utility, while 16% selected the statement without information utility framing.

In comparing statement selection between the control and experimental groups, a chi-square test of independence showed a significant association between condition and statement choice (χ²(5) = 59.05, p < 0.001, Cramer’s V = 0.41). Examination of standardized residuals indicated that the four original statements did not differ substantially by condition (all |z| < 2). However, both information-utility statements introduced only in the experimental condition were selected more often than expected statistically (environmental statement: z = 3.1; health statement: z = 3.9). Together, these results suggest a directional advantage for information utility framing in a free-choice context, consistent with H2.

Exploratory logistic regression analyses examined whether selective exposure varied as a function of meat-related attitudes, attachment, and perceived risk, controlling for demographic factors. The model significantly distinguished between participants selecting risk- versus benefit-framed information in the control group (χ²(12) = 29.18, p = 0.004). Higher perceived risk was associated with lower odds of selecting benefit-framed information (OR = 0.47, p = 0.046, 95% CI [0.23, 0.99]). Meat attachment (OR = 1.06, p = 0.11, 95% CI [0.99, 1.4]) and affective attitude (OR = 0.81, p = 0.54, 95% CI [0.43, 1.56]) were not independently associated with message selection.

The most common reason participants gave for selecting the statement they did was to confirm prior knowledge, attitudes, or beliefs. Examples of this reason included believing the statement to be true or preferring to read about information that supported their eating habits. The second most common reason for statement selection was to learn more. Responses coded into this category ranged from those who stated having no prior knowledge of the topic to those who wanted to be reminded of specific details or facts.

The majority of participants (61%) did not learn anything new from the information they read in the experiment portion of the study. Descriptively, the proportion of participants reporting new learning varied across statements, but no single article clearly stood out as substantially more informative than the others.

Most participants (78%) reported that the information they read during the experimental portion of the study would not influence their future dietary decisions. Even among those who read about meat risks (without information utility), the majority (62%) still reported that the information would not change their dietary decisions. This pattern persisted among participants who read risk information framed using information utility, with 58% reporting that the information would not influence their future dietary decisions. Notably, participants who read health risk information framed using information utility were more likely to report that it influenced their future dietary decisions: 55% indicated a change, 15 percentage points higher than those who read health risk information without information utility framing (40%).

For 57% of participants, the information they read did not change their self-reported attitude toward eating meat. Twenty-seven percent said that the information made them less favorable toward eating meat, and 16% said the information made them more favorable toward eating meat. An ordinal logistic regression examined whether information utility framing, topic domain (environment vs. health), or their interaction predicted changes in attitudes toward meat. The overall model was statistically significant (χ²(3) = 26.98, p < 0.001), and the proportional odds assumption was satisfied (test of parallel lines, p = 0.18). Information utility did not significantly predict post-exposure attitudes toward meat (OR = 2.98, p = 0.055, 95% CI [0.98, 9.11]), providing no evidence that the framing effect reduced favorable attitudes toward meat. Moreover, neither the topic domain (OR = 0.91, p = 0.82, 95% CI [0.42, 2.00]) nor the interaction between information utility and topic (OR = 2.73, p = 0.16, 95% CI [0.68, 10.90]) significantly predicted attitude change.

An ordinal logistic regression was conducted to examine whether information utility framing or topic domain predicted post-exposure willingness to reduce meat consumption, controlling for baseline willingness. The overall model was significant (χ²(4) = 133.62, p < 0.001), and the proportional odds assumption was met (test of parallel lines: p = 0.98). Baseline willingness was a strong positive predictor of post-exposure willingness (OR = 4.27, p < 0.001, 95% CI [3.11, 5.87]), indicating that participants with higher initial willingness were substantially more likely to report greater willingness after exposure to risk information. Topic domain also significantly predicted willingness, with participants who engaged with health risk information having higher odds of greater willingness than those who engaged with environmental messages (OR = 4.43, p < 0.001, 95% CI [2.15, 9.14]). In contrast, information utility framing was not significantly associated with post-experiment willingness (OR = 1.40, p = 0.52, 95% CI [0.51, 3.86]), nor was the interaction between information utility and topic (OR = 0.89, p = 0.84, 95% CI [0.30, 2.61]).

This study predicted that participants would preferentially select information about the benefits of eating meat over information about its risks (H1). The results partially support this hypothesis. In the control group, participants were substantially more likely to select information emphasizing the health benefits of meat compared to health risks. In contrast, no such preference emerged for environmental information, where selection of benefit- and risk-framed statements was evenly distributed.

This pattern aligns with the theory of selective exposure, which posits that individuals prefer information that aligns with their pre-existing attitudes, behaviors, or beliefs (Festinger, 1957). Because all participants in this study were omnivores, health-benefit information provides a personal and immediate justification for meat consumption, thereby affirming the value of participants’ current dietary practices. Environmental benefit information may have been perceived as more abstract and less directly tied to individual outcomes, reducing or eliminating selective exposure behaviors for this type of material. However, previous research has demonstrated selective exposure to meat-related environmental information when such content is evaluated in isolation (Lueders et al., 2022), suggesting that the current study’s higher choice context may have influenced selective exposure effects. In other words, health-related benefit information may have dominated attention not because environmental considerations were irrelevant, but because health content offered greater personal salience in a context in which participants had multiple options. This pattern aligns with prior research, which has found that individuals tend to favor self-related information over information unrelated to the self (Cosme et al., 2023; Green and Peloza, 2014; Kwong et al., 2021).

Taken together, these findings align with existing literature demonstrating that people favorably seek information that endorses rather than challenges their preferred diets (Gaspar et al., 2016; Lueders et al., 2022). Additionally, the results indicate that selective exposure to meat-related information is shaped not only by the content (benefits vs. risks) but also by perceived personal relevance and environmental complexity.

The study predicted that information utility framing would increase participants’ likelihood of selecting meat-related risk information compared with standard framing (H2). The results for this hypothesis support a modest directional advantage for content framed using information utility compared to content without information utility. The chi-square test confirmed that both utility-framed risk statements were chosen more frequently than expected, whereas selection of the original statements remained stable across conditions. This pattern suggests that information utility may enhance the perceived relevance or value of risk information without substantially displacing interest in other options. Importantly, the observed effects were directional rather than large in magnitude, highlighting that while information utility framing may help draw attention to risk-related content for some individuals, it is unlikely to substantially increase engagement among the majority.

Prior research has demonstrated that information utility framing can mitigate selective exposure biases, particularly in political contexts where individuals are motivated to acquire information that supports effective decision-making (D’Alessio, 2015; Knobloch-Westerwick and Kleinman, 2012; Wagner, 2017). The modest effects observed in the present study may be explained by differences in motivational contexts. Unlike political information, which is often tied to imminent, externally imposed decisions, dietary choices are habitual and largely self-regulated. As a result, the perceived value of risk information may be lower, limiting the extent to which information utility framing can override existing preferences. Together, the study’s results suggest that information utility framing may be less effective for overcoming selective exposure biases in the context of meat consumption than politics.

RQ1 explored whether selective exposure to meat-related information varied as a function of participants’ pre-existing meat-related characteristics, including attitudes, attachment, and perceived risk. Consistent with selective exposure theory (Festinger, 1957), individuals’ prior beliefs and perceptions showed meaningful associations with information selection patterns. In particular, a higher perceived risk associated with meat consumption was linked to higher odds of selecting risk information, suggesting that participants were inclined to engage with content that confirmed their prior beliefs. Importantly, this association emerged even after controlling for sociodemographic factors, indicating that perceived risk plays a distinct role in shaping selective exposure beyond demographic differences. While meat attachment and affective attitudes did not emerge as independent predictors of statement selection in the regression model, their influence may operate indirectly through risk perceptions or broader evaluative frameworks. For instance, previous research has shown that attitudes toward red meat are positively linked to information avoidance (Gaspar et al., 2016); thus, the absence of such effects here may reflect measurement limitations rather than a true lack of influence. Specifically, because this study’s attitude question asked participants how they “feel” about meat, responses may have primarily captured affective reactions such as taste preference rather than more nuanced judgments about the food.

Beyond model-based associations, participants’ self-reported reasons for statement selection offer complementary evidence of selective exposure biases. The most common reason given for statement selection was to confirm prior knowledge, attitudes, or beliefs, indicating that many participants were consciously aware of selecting information that aligned with their existing perspectives. Taken together, these quantitative and qualitative findings suggest that perceived risk may be a particularly salient factor shaping selective exposure to meat-related information. These findings should be interpreted cautiously, given the exploratory nature of the analyses, but they nonetheless highlight perceived risk as a promising focus for future research examining selective exposure in dietary and sustainability contexts.

RQ2 examined whether information utility can decrease positive attitudes toward meat consumption among participants. Overall, the results did not provide evidence that information utility framing meaningfully shifted affective evaluations of meat. Moreover, neither health nor environmental risk content independently predicted changes in participants’ attitudes, suggesting that sentiments toward meat may be relatively resistant to change following brief exposure to information. These findings contrast with research by Gaspar et al. (2016), who found that self-selected exposure to health and environmental risk messages reduced positive attitudes toward red meat. One plausible explanation for this discrepancy lies in differences in how attitude changes were operationalized. Whereas Gaspar et al. (2016) employed multi-item attitude scales capable of detecting subtle shifts in evaluative judgments, the present study relied on a single, self-reported post-exposure item. This measure likely captured only consciously recognized attitude change, rather than more modest or implicit shifts in evaluation. Together, these findings indicate that brief exposure to meat-related risk information is unlikely to shift attitudes in large, obvious ways; however, risk information may change attitudes more subtly, potentially requiring several repeated exposures before attitude shifts are more pronounced.

Similar in nature to attitude changes, RQ3 assessed whether information utility can increase willingness to reduce meat consumption among participants. Participants who engaged with health risk information reported greater willingness to reduce meat consumption than those exposed to environmental messages; however, information utility framing did not independently predict willingness, nor did it interact with topic domain. Additionally, baseline willingness emerged as a second predictor of post-exposure willingness, with those already inclined to reduce meat consumption substantially more likely to report greater willingness after exposure to risk information. These results suggest that reframing risk information in terms of its usefulness may be insufficient to shift willingness after brief message exposure. Instead, the findings indicate a relative advantage of health risk information over environmental risk information in promoting reductions in meat consumption, consistent with patterns observed by Taillie et al. (2022). Furthermore, the prevalence of baseline willingness highlights the importance of pre-existing motivations in shaping responses to dietary messaging and suggests that information-based interventions may be most effective when targeted toward individuals who already exhibit some readiness to change.

Descriptive analyses further indicate that, for the majority of participants, risk information did not influence their future dietary choices, consistent with the lack of attitude changes and the influence of baseline willingness reported in RQ2 and RQ3. Notably, however, a descriptive pattern suggested that participants who engaged with health risk information framed using information utility were more likely to report that the information influenced their dietary decisions. This pattern should be interpreted cautiously. Rather than indicating persuasive effects of information utility, it likely reflects the selection processes: individuals already considering dietary change may have been more inclined both to engage with information-utility-framed content and to perceive that information as influential. These results suggest that information utility may serve as a “threshold nudge” for individuals already near the point of change, rather than as a mechanism capable of shifting intentions among resistant consumers.

These results partially contrast with the literature on risk communication, which suggests that attention to and exposure to risk information should increase individuals’ perceptions of risk (Mrkva et al., 2021), thereby increasing the likelihood that individuals respond to the risk with protective behavior (Cheng et al., 2022; Saari et al., 2021; Qi et al., 2022). In the present study, exposure to health risk information was associated with greater post-exposure willingness to reduce meat consumption; however, this effect occurred in the absence of corresponding changes in attitudes or self-reported dietary intentions and was conditioned by baseline willingness. These patterns suggest that health risk messages may have primarily reinforced existing inclinations rather than produced broad motivational change. In the context of meat consumption, pre-existing attitudes, habits, and behavioral intentions may be particularly strong, limiting the extent to which brief exposure to risk information can produce behavioral change. These conclusions are consistent with research by Weingarten et al. (2022), who found that a single exposure to information about the negative effects of meat consumption on personal health or the environment had limited effects on attitudes, intentions, and behavior. Conversely, studies that provided similar information through multiple exposures or over longer durations have been more successful in influencing dietary behaviors (Jalil et al., 2020; Jay et al., 2019; Wolstenholme et al., 2020). These results indicate that risk information may require repeated exposure before having a significant impact on predictors of meat consumption, such as attitudes, intentions, and willingness.

The majority of participants in this study held positive attitudes toward meat and were highly attached to it. Despite their favorable attitudes, more participants were willing to reduce their meat consumption than those who were unwilling. These findings align with research by Charlebois et al. (2018), who found that just over half of Canadian participants were willing to reduce their meat intake. However, most Canadians in this study reported eating meat daily, with infrequent consumption habits being few and far between. Again, this reflects research by Charlebois et al. (2018), who reported that nearly half of their participants ate meat daily, followed by a substantial portion who ate meat once or twice a week.

In this study, chicken was the most commonly consumed meat, followed by beef. While less popular than chicken, beef is the largest contributor to food-system GHG emissions globally (Poore and Nemecek, 2018) and has an estimated protein-based emissions intensity over 11 times that of chicken (Dyer et al., 2010). These findings indicate a need to discourage meat consumption, particularly beef, among Canadian consumers. As taste was a primary reason for eating meat among participants in this study, strategies aimed at addressing this concern may be beneficial for promoting sustainable diets. Health was another main reason for eating meat, providing support for why health messages performed better than environmental messages in this research.

The study’s results also indicate that, in general, Canadian adults have a poor understanding of the environmental impact of meat production. Overall, the findings suggest that approximately one-third of participants had at least a moderate level of awareness related to the environmental impact of meat production; however, very few had a strong understanding of the subject. According to Wynes and Nicholas (2017), eating less meat is a “moderate-impact” individual action for mitigating climate change, while eating a vegan diet is a “high-impact” action. In the current study, just over a third of participants correctly indicated that eating less meat is a somewhat effective lifestyle change for fighting climate change; however, very few knew that a vegan diet is highly effective. Additionally, about one-fourth of participants harbored misconceptions about the topic, believing that meat production benefits the environment. These findings align with previous research showing that consumers in high-income countries tend to underestimate or ignore the negative environmental impact of meat (de Boer et al., 2016; Hartmann et al., 2021; Pohjolainen et al., 2016; Slotnick et al., 2022). Specifically, this study’s findings reflect those reported by Pohjolainen et al. (2016), who found that approximately one-third of Finnish participants were aware of the environmental issues of the meat sector, almost half were unsure, and close to a quarter opposed the idea that meat production is harmful to the environment.

Regarding awareness of meat-related health risks, the findings suggest that Canadian adults have a limited understanding of this issue. About a third of participants agreed that “meat-rich diets increase the risk of disease,” while a similar number disagreed. Additionally, the majority agreed that “meat is an important part of a healthy diet,” while very few disagreed. These findings support research suggesting that misconceptions about nutrition act as a barrier to adopting meatless diets (Collier et al., 2021; Corrin and Papadopoulos, 2017; Loiselle et al., 2024), indicating that many consumers believe meat provides essential nutrients that cannot be adequately replaced by plant-based foods.

This study’s first limitation relates to its participant recruitment process. Using a third-party recruitment service to identify participants for this study enabled a broad range of participants across age, gender, location, and education level. However, this approach limited participants to registered platform clients, thereby excluding those with lifestyles or personalities unlikely to complete online studies in their spare time. This limitation was accepted as a trade-off for a broader range of participants and faster data collection.

Additionally, although a diverse group of participants was recruited for this study, chi-square tests revealed significant differences between the study sample and 2021 Canadian census data, indicating that the sample was not fully representative of the Canadian population. In particular, geographic location and education level differed greatly from the census benchmarks; however, all other observed discrepancies in age, gender, and urban versus rural living location were generally small. Nonetheless, results should be interpreted with awareness that the sample was not perfectly representative of the Canadian adult population.

Moreover, although the sample size was sufficient to estimate moderate differences in experimental and post-experimental outcomes, some effects were estimated with limited precision due to relatively low base rates of selection. This resulted in wider confidence intervals around odds ratio estimates and reduced sensitivity to detect small effects. Consequently, weak or null findings should be interpreted as reflecting statistical uncertainty rather than definitive evidence of no effect. Future research with larger samples is needed to more precisely estimate small effects and to clarify the conditions under which information utility may influence selective exposure and downstream outcomes.

A second set of limitations concerns the study’s experimental design. In the experimental group, participants were provided with two sets of statements related to meat risks: one set framed with information utility and the other without. This allowed the researchers to observe whether participants preferred risk statements that used information utility or those that did not, when presented with both options. To ensure that sentence phrasing did not influence statement selection between groups, the first half of each information utility statement remained the same as its meat risk counterpart, while the second half was changed. This meant that participants in the experimental group were presented with two sets of very similar risk statements. For example, the environmental risk statements shown in the experimental group were: “Eating meat plays a big role in global warming; here’s why” and “Eating meat plays a big role in global warming; how shifting your diet can reduce your impact.” Having these closely related statement options in the experimental group may have indicated to participants the study’s true purpose and, therefore, influenced their topic choice. However, including these mirrored statements in the experimental group was necessary to create information utility without significantly altering sentence phrasing.

Additionally, the study’s overall research design may limit the generalizability of its findings to all individuals and situations. This study’s experiment asks participants to select one of several statements in a multiple-choice format. However, in the real world, a host of other factors, such as personalized or targeted messaging, associated images, or website recognition, may influence the content that individuals interact with. While creating an online news website may have better simulated real life, as was done in similar studies (Jang, 2014; Knobloch-Westerwick and Kleinman, 2012), using Qualtrics allowed for inexpensive, simple data collection that would not have been possible with other platforms.

While valuable for capturing personal insights and experiences, this study’s use of self-reported data may have unintentionally limited the collection of accurate information due to social desirability bias, memory recall issues, or misunderstandings of the questions. Social desirability bias can lead participants to answer in ways they perceive as more favorable or acceptable rather than truthfully. Moreover, the subjective nature of self-reporting means responses can be influenced by individual perceptions and interpretations, which may not accurately reflect objective reality. These factors can compromise the reliability and validity of self-reported data, necessitating caution in interpreting and using it in research conclusions.

Furthermore, this study measured participants’ short-term attitudes and intentions following message exposure, but did not assess long-term behavior change. As a result, it is unclear whether participants who reported an intention or willingness to shift their future dietary habits actually did so. Future research could examine whether these initial intention shifts persist over time and translate into measurable behavioral change.

Lastly, regarding the specific instruments used in this study, the measure used to examine baseline meat attitudes may have lacked construct validity. Because the study’s results did not reveal the expected relationship between baseline meat attitudes and selective exposure bias, it is assumed that the initial attitude question did not collect the data it was intended to. In an effort to make the study accessible for all participants, a five-point graphic slider was used to collect information on participants’ attitudes toward meat. Although graphic scales can be easier to process and understand than text scales (Stange et al., 2018; Toepoel et al., 2019), a semantic differential scale may have allowed for more nuanced and accurate responses from participants. Similarly, given that this study did not observe attitude change following exposure to risk information, unlike prior research (Gaspar et al., 2016), the single-item post-experiment attitude measure used may have lacked sufficient sensitivity to detect small or implicit shifts in evaluation. Therefore, further research on meat consumption attitudes should use semantic differential scales rather than graphic sliders or single items to ensure data are collected accurately and as intended.

The final limitations of this study to be discussed relate to its boundaries and scope, giving way to concrete recommendations for future research. The current study was limited to exploring meat-related message framing and did not explicitly explore plant-based diet communication strategies. However, vegan diet messaging is an equally important area of investigation, as encouraging these diets is crucial for mitigating global warming (Gerber et al., 2013; IPCC, 2022; Xu et al., 2021). Additionally, because veganism is subject to greater social stigma than vegetarianism (Nezlek et al., 2023; Paslakis et al., 2020; Reuber and Muschalla, 2022), consumers may interact with and respond to plant-based messaging differently than they do with meatless diet messaging. Therefore, future research should consider conducting a similar study focused on the communication of vegan diets.

Future research might also explore alternative ways of framing meat-related risk communication that reduce psychological defensiveness and selective exposure bias. For instance, while there are several reasons individuals may choose not to eat meat, this study focused only on health and environmental motivations. Future research could examine whether framing the financial or animal welfare benefits of sustainable diets in terms of information utility may motivate selective exposure. Additionally, this study identified taste and health as the two primary reasons Canadians eat meat. Because health information was more popular than environmental information overall, future research may benefit from examining taste-focused messaging as a potential pathway for promoting sustainable diets. No studies have investigated whether framing meatless diets as tasty influences selective exposure, highlighting an avenue for future research.

Furthermore, to keep within the boundaries and scope of its objectives, the current study limited participants to only those who eat meat. Because the research aimed to understand selective exposure as a driver of positive bonds toward meat consumption and information utility as a strategy for mitigating this selection bias, individuals who identified as vegan, vegetarian, or pescatarian were excluded from this study. The results, therefore, do not provide insights into the extent to which meat abstainers engage in selective exposure. Future research might conduct a similar experiment using statements focused on the risks and benefits of meat-free diets with vegan or vegetarian participants. As previous studies have found that people who eat a meatless diet are higher in openness and agreeableness than meat eaters (Keller and Siegrist, 2015; Pfeiler and Egloff, 2018; Tan et al., 2021), it is possible that these individuals would not engage in selective exposure to the same degree that omnivores do. These findings would add to the literature on the personalities and behaviors of vegans and vegetarians, providing additional understanding as to why these individuals choose to follow meatless diets.

Lastly, the present study was primarily designed to test selective exposure and information utility framing in a high-choice information environment. Accordingly, analyses examining whether selective exposure varied as a function of participants’ baseline characteristics (RQ1) were exploratory and fell outside the study’s central inferential focus. Although perceived risk emerged as a meaningful correlate of statement selection, these analyses were not intended to provide a comprehensive explanation of individual differences in information-seeking behavior. Therefore, further research, specifically focused on individual differences in selective exposure, could build on these exploratory findings by employing pre-registered hypotheses.