Communicating highly divergent levels of scientific and social consensus: its effects on people’s scientific beliefs

ABSTRACT

This study examined the impact of presenting scientific and social consensus information together on people’s scientific beliefs when the two types of consensus information contradict each other. Japanese adults (N = 1,518) received information about high scientific consensus and low social consensus on the safety of genetically modified (GM) foods, information only about high scientific consensus, information only about low social consensus, or no consensus information. The presentation of only scientific or social consensus information had no effect on participants’ beliefs about the safety of GM foods, whereas the simultaneous presentation of scientific and social consensus information improved their beliefs in some degree. The effect of presenting scientific and social consensus information together was mediated by perceived scientific consensus.

KEYWORDS:

• Scientific consensus
• social consensus
• scientific beliefs
• gateway belief model

Introduction

There are many cases where the general public’s opinion about a scientific issue considerably diverges from a consensus among scientists on the issue. For example, a survey by the Pew Research Center (2015) found wide opinion gaps between the general American and scientists regarding several scientific issues, such as climate change (50% of people and 87% of scientists agreed that human activity is a primary cause of climate change), human evolution (65% of people and 98% of scientists agreed that humans have evolved), and GM foods (37% of people and 88% of scientists agreed that GM foods are safe to eat). These gaps exemplify a failure in science communication (Kahan, 2015) and may contribute to the difficulty of making informed judgments on scientific issues and changing the behavior of large populations (Akter, Bennett, & Ward, 2012; Kennedy, LaVail, Nowak, Basket, & Landry, 2011).

One effective and efficient method for narrowing the wide gap between scientific and social consensus is to inform people about the actual level of scientific consensus. In recent years, an increasing number of studies have demonstrated that, at least under certain conditions, providing individuals with information about the actual scientific consensus successfully changes their scientific beliefs (e.g. Cook & Lewandowsky, 2016; Lewandowsky, Gignac, & Vaughan, 2013) through the improvement in their perceptions of scientific consensus (e.g. Dixon, 2016; Kerr & Wilson, 2018; Kobayashi, 2018; van der Linden, Leiserowitz, Feinberg, & Maibach, 2015).

Remarkably, the previous studies have mainly focused on the impact of scientific consensus information from an authoritative source (e.g. scientists) under the assumption that counter-information is absent or negligible. However, this assumption is too optimistic, considering that in real-world situations, people often encounter conflicting scientific information from different sources (Carpenter et al., 2016). Particularly when there is a wide opinion gap between scientists and the general public, people may be exposed to contrary opinions that reflect scientific and social consensus respectively. The effectiveness of communicating a level of scientific consensus may vary according to whether information indicating the opposite level of social consensus exists. The present study explores this possibility.

Effects of communicating scientific and social consensus

A consensus among group members often acts as a trigger for conformity (Cialdini & Goldstein, 2004). Social influence research has shown that people have a strong tendency to square their opinions with those of the majority (e.g. Ash, 1955; Deutsch & Gerard, 1955). Of particular importance here is that conformity to group consensus occurs not only because the consensus of opinion puts pressure on each group member to affiliate with his or her group but also because it works as a heuristic cue to infer reality (Cialdini & Goldstein, 2004; Deutsch & Gerard, 1955). In addition, conformity does not necessarily require directly observing what others say and do. Merely informing people about the majority opinion can stimulate the conformity process (e.g. Klucharev, Hytönen, Rijpkema, Smidts, & Fernández, 2009).

Group expertise is an important factor that influences the level of conformity to a consensus among group members. All other things being equal, people conform to a group of members with expertise more than a group of members without expertise (Costanzo, Reitan, & Shaw, 1968; Crano, 1970). Also, when the informed minority and the uninfomed majority are inconsistent in their judgments and behaviors, even young children prefer to learn from the informed minority over the uninformed majority (Bernard, Proust, & Clément, 2015; Einav, 2014). In general, the general public perceives scientists to be competent in handling scientific issues (Fiske & Dupree, 2014; Pew Research Center, 2015) and the most reliable source of scientific information (Lang & Hallman, 2005). Therefore, it is not surprising that communicating a level of consensus among scientists to people has an impact on their judgments of what is scientifically acceptable.

Yet, this does not mean that the influence of social consensus information on scientific beliefs is negligible. In fact, evidence suggests that the presentation of social consensus information has an impact on people’s scientific beliefs. For example, Stangor, Sechrist, and Jost (2001) found that individuals who were informed about a level of social consensus on stereotypic traits of African American tended to change their beliefs about the issue in line with the consensus level. Puhl, Schwartz, and Brownell (2005) found that providing individuals with information about social consensus in favor of obese persons improved their beliefs about stereotypic traits of obese persons and the cause of obesity. In a study by Kobayashi (2018), information about social consensus on nuclear power and Japan’s whaling research influenced individuals’ beliefs about the issues through their perceptions of the current levels of social consensus on the issues.

Although researchers have quite recently begun to explore whether the impact of communicating the actual scientific consensus is reduced by subsequent exposure to counter-information or misinformation (Benegal & Scruggs, 2018; Cook, Lewandowsky, & Ecker, 2017; van der Linden, Leiserowitz, Rosenthal, & Maibach, 2017), no research attention has been directed to the question how people deal with scientific consensus information in the presence of social consensus information. In our society, the opportunities for exposure to information indicating and reflecting social consensus on a scientific issue are rich. For example, the results of a survey of public opinion about a scientific issue are often made public (see e.g. Pew Research Center, 2015). It is also common for ordinary people to rely on non-scientists (e.g. family members, friends) as a source of scientific information (Haynes, Barclay, & Pidgeon, 2008). Given that communicating a level of social consensus influences people’s scientific beliefs, the interactive effects of scientific and social consensus information are worthy of attention. The first goal of the present study is to investigate whether and how exposure to scientific and social consensus information influences people’s scientific beliefs when the two types of consensus information contradict each other.

Perceived scientific and social consensus as mediators

The gateway belief model (van der Linden et al., 2015) provides a useful framework for examining a psychological mechanism that underlies the impact of scientific and social consensus information. According to this model, ordinary people’s perceptions of (or their beliefs about) the actual consensus among scientists on a scientific issue play a key role in changing or maintaining their beliefs about the issue. The impact of scientific consensus information on their scientific beliefs is mediated by their perceptions of scientific consensus. Although the main concern of the gateway belief model is the role of perceived scientific consensus, prior findings suggest that the presentation of social consensus information affects people’s perceptions of social consensus, which in turn are used to adjust their scientific beliefs (Kobayashi, 2018; Puhl et al., 2005; Stangor et al., 2001). The basic tenet of the gateway belief model may extend to the processing of social consensus information. That is, exposure to social consensus information may influence scientific beliefs through perceived social consensus. The second goal of the present study is to investigate whether perceived scientific and social consensus mediate the impact of scientific and social consensus information on scientific beliefs.

The present study

The present study was conducted using a sample of Japanese people and adopting the safety of GM foods as a scientific issue. Although no studies have directly investigated Japanese scientists’ and ordinary people’s beliefs about this issue, evidence suggests that there is a wide gap between scientific and social consensus in Japan. For example, several Japanese scientific societies (e.g. the Japan Society for Bioscience, Biotechnology, and Agrochemistry) explicitly support the active use of GM foods on scientific evidence (Okada, 2005). On the other hand, a survey conducted by the Council for Biotechnology Information Japan (2016), using a sample of Japanese female adults, indicated that only 4.8% of the respondents had positive attitudes toward GM foods.

In the present study, participants receive information about a high level of scientific consensus and a low level of social consensus on the safety of GM foods, information only about high scientific consensus, information only about low social consensus, or no consensus information. The consensus information is new to them, given that, as mentioned above, levels of scientific and social consensus on the safety of GM foods in Japan have never been investigated and that people tend to misperceive the actual levels of scientific and social consensus (e.g. Leviston, Walker, & Morwinski, 2013; Myers, Maibach, Peters, & Leiserowitz, 2015). Next, participants estimate the levels of scientific and social consensus and report their beliefs about the safety of GM foods. Based on the above discussion, it is hypothesized that presenting only scientific consensus information will increase recipients’ GM safety beliefs, whereas presenting only social consensus information will decrease their beliefs. Regarding the interactive effects of scientific and social consensus information, the following research question is posed. Will the simultaneous presentation of scientific and social consensus information increase, decrease, or not change recipients’ GM safety beliefs? Furthermore, we hypothesize that perceived scientific and social consensus will mediate the impact of scientific and social consensus information on GM safety beliefs.

Method

Participants

An Internet-based experiment was conducted on Japanese adults in October 2017. A total of 6,461 potential participants were invited through a survey research company (Intage Inc.) to participate and 1,518 (50.1% females; M = 44.60 years old, SD = 14.26) completed the experiment. Participants were randomly assigned to one of four conditions: scientific and social consensus condition (n = 383), scientific consensus condition (n = 377), social consensus condition (n = 378), and no consensus condition (n = 380). A power analysis using G*Power 3.1.3 (Faul, Erdfelder, Buchner, & Lang, 2009) indicated that given the present sample size and α = .05, an effect size of η_p² = .012, which corresponds to f = 0.11, a small effect as judged by Cohen’s (1988) criteria, would be detectable with 95% power.

Stimulus materials

Information about scientific and social consensus on the safety of GM foods was presented with pie charts (see the supplemental online material) as the results of a (fictitious) survey. The levels of scientific and social consensus were created using the results of the Pew Research Center (2015) and the Council for Biotechnology Information Japan (2016), respectively. Participants in the scientific and social consensus condition received the following information: ‘In 2016, the Science Council of Japan conducted a survey of opinions about genetically modified foods, with a random sample of 1,018 scientists and 2,173 ordinary people. Results indicated that 88% of the scientists and 5% of the ordinary people agreed that genetically modified foods are safe to eat.’ Participants in the scientific consensus condition and the social consensus condition received information only about scientific and social consensus, respectively. Participants in the no consensus condition did not receive any information about scientific or social consensus.

Measures

Familiarity, personal importance, competence, and knowledgeability

Participants rated their familiarity with the health risk of GM foods and their personal importance of the safety of GM foods, using 7-point Likert-type scales ranging from not familiar at all (1) to very familiar (7) and not important at all (1) to very important (7), respectively. They also rated the extent to which scientists and ordinary people are competent in judging the safety of GM foods accurately and knowledgeable about the issue, using 7-point Likert-type scales ranging from not competent at all (1) to highly competent (7) and not knowledgeable at all (1) to highly knowledgeable (7), respectively. The latter two ratings were averaged (for scientists, r = .53, and for ordinary people, r = .74) so that a composite index, perceived expertise, was created.

Prior attitudes toward GM foods

Using a 7-point Likert-type scale ranging from strongly disagree (1) to strongly agree (7), participants rated the extent to which they agreed with the following statements: ‘It is better to avoid purchasing genetically modified foods,’ ‘The import of foreign-produced genetically modified foods should be increased if they are cheaper than non-genetically modified foods,’ ‘Genetically modified crops should not be used for children’s food,’ and ‘Genetically modified crops should be more widely used for various kinds of processed food.’ The first and third ratings were reversed, and then the four ratings were averaged (α = .85). Higher scores indicated more favorable attitudes toward GM foods.

Perceptions of scientific and social consensus

Participants estimated the percentages (0–100%) of scientists and ordinary people who would agree with the following belief statements: ‘Genetically modified foods are safe for adults to eat,’ ‘Genetically modified foods are safe for children to eat,’ and ‘Scientific evidence favors the safety of genetically modified foods.’ The three estimated percentages were averaged for scientists (α = .94) and ordinary people (α = .92).

GM safety beliefs

Participants rated the extent to which they agreed with the above-mentioned three belief statements, using a 7-point Likert-type scale ranging from strongly disagree (1) to strongly agree (7). The three ratings were averaged (α = .88) so that higher scores indicated stronger beliefs about the safety of GM foods.

Procedure

First, participants rated their familiarity with the health risk of GM foods, their personal importance of the safety of GM foods, the competence and knowledgeability of scientists and ordinary people, and their attitudes toward GM foods. Second, participants in the consensus treatment conditions received information about scientific and/or social consensus and then rated their surprise at the given consensus information, using a 7-point Likert-type scale ranging from not surprised at all (1) to very surprised (7). This rating was used to make them pay attention to the consensus information (Stangor et al., 2001). Third, participants in all the conditions estimated the levels of scientific and social consensus on the safety of GM foods. Finally, they answered their GM safety beliefs. After answering all the questions, participants were debriefed.

Results

Preliminary analyses

The mean score for prior attitudes toward GM foods was 3.18 (SD = 1.00). Of the 1,518 participants, 7.8% had positive attitudes toward GM foods (over the score of 4.00) and 64.7% had negative attitudes (below 4.00). The mean ratings of familiarity with and personal importance of the issue were 3.21 (SD = 1.27) and 4.54 (SD = 1.28), respectively. Separate one-way analyses of variance (ANOVAs) revealed no significant differences among the consensus conditions in prior attitude, F(3, 1514) = 1.48, MSE = 1.00, η_p² = .003, familiarity, F < 1, or personal importance, F < 1.

A mixed model ANOVA was conducted on perceived expertise scores, with consensus condition as a between-participant factor and group (scientists, ordinary people) as a within-participant factor. Scientists (M = 4.51, SD = 1.02) were perceived to be more expert than ordinary people (M = 3.15, SD = .99), F(1, 1514) = 1666.73, p < .001, η_p² = .52. There were no significant main effect of consensus condition, F(3, 1514) = 1.24, MSE = 1.17, η_p² = .002, or consensus condition × group interaction, F(3, 1514) = 1.07, MSE = .85, η_p² = .002.

Effects of scientific and social consensus information on GM safety beliefs

A one-way analysis of covariance (ANCOVA) on GM safety belief scores, with consensus condition as a between-participant factor and prior attitude as a covariate, revealed a significant difference among the consensus conditions, F(3, 1513) = 5.63, MSE = .58, p < .001, η_p² = .01. Bonferroni post-hoc tests at p < .05 indicated that participants in the scientific and social consensus condition had stronger GM safety beliefs (M = 3.55, SE = .04) than participants in the no consensus condition (M = 3.38, SE = .04) and the social consensus condition (M = 3.35, SE = .04; see Table 1). That is, the simultaneous presentation of scientific and social consensus information (compared to only social consensus and no consensus information) increased participants’ GM safety beliefs. No other differences reached significance, indicating that, contrary to our hypotheses, neither scientific nor social consensus information substantially influenced their GM safety beliefs.

Table 1. Adjusted means, standard errors, and effect sizes for GM safety beliefs and perceived scientific and social consensus as a function of consensus condition.

	GM safety beliefs		Perceived scientific consensus		Perceived social consensus
Condition	M (SE)	Cohen’s d	M (SE)	Cohen’s d	M (SE)	Cohen’s d
Scientific and social consensus	3.55 (.04)a, b	0.22	61.6 (1.3)a, b	0.64	24.5 (1.1)a, b	–0.33
Scientific consensus	3.47 (.04)	0.12	63.0 (1.3)c, d	0.70	35.1 (1.1)a, c	0.17
Social consensus	3.35 (.04)a	–0.04	38.4 (1.3)a, c, e	–0.28	28.2 (1.1)c	–0.15
No consensus	3.38 (.04)b	–	45.4 (1.3)b, d, e	–	31.5 (1.1)b	–

Note. Same subscripts within a column indicate that the two means are significantly different at p < .05. Cohen’s ds are reported for the consensus treatment versus no consensus conditions.

Mediating effects of perceived scientific and social consensus

Two separate ANCOVAs, with consensus condition as a between-participant factor and prior attitude as a covariate, revealed that the consensus conditions significantly differed in perceived scientific consensus, F(3, 1513) = 91.61, MSE = 607.87, p < .001, η_p² = .15, and perceived social consensus, F(3, 1513) = 18.03, MSE = 433.36, p < .001, η_p² = .04. Bonferroni post-hoc tests (p < .05) indicated that, as shown in Table 1, perceived scientific consensus was the highest for the scientific and social consensus condition (M = 61.6, SE = 1.3) and the scientific consensus condition (M = 63.0, SE = 1.3), followed by the no consensus condition (M = 45.4, SE = 1.3) and the social consensus condition (M = 38.4, SE = 1.3). Perceived social consensus was lower for the scientific and social consensus condition (M = 24.5, SE = 1.1) than for the scientific consensus condition (M = 35.1, SE = 1.1) and the no consensus condition (M = 31.5, SE = 1.1). Participants in the social consensus condition (M = 28.2, SE = 1.1) also perceived social consensus to be lower than participants in the scientific consensus condition.

To examine the mediating effects of perceived scientific and social consensus, a mediation analysis was conducted, using the PROCESS macro version 3 (Model 4) for SPSS by Hayes (2017). The analysis focused on the indirect effect of the scientific and social consensus versus no consensus conditions because only the simultaneous presentation of scientific and social consensus information significantly influenced participants’ GM safety beliefs. The mediation model included GM safety belief scores as a dependent variable, consensus condition (0 = no consensus condition, 1 = scientific and social consensus condition) as an independent variable, perceived scientific and social consensus as mediators, and prior attitude, familiarity, and personal importance as covariates. Percentile bootstrap 95% confidence intervals (CIs) for the mediating effects of perceived scientific and social consensus were estimated with 10,000 bootstrap samples. As shown in Figure 1, the mediating effect of perceived scientific consensus was significantly greater than zero, 95% CI [.08, .18], whereas the mediating effect of perceived social consensus was not significant, 95% CI [−.05, .00]. In short, scientific and social consensus information influenced participants’ GM safety beliefs through their perceptions of scientific consensus, supporting our hypothesis partly.

Figure 1. Results of mediation analysis (0 = no consensus condition, 1 = scientific and social consensus condition). Unstandardized regression coefficients (b), standard errors (SE), percentile bootstrap 95% confidence intervals (95% CI), and effect sizes (f²) are reported.

Discussion

The present study was the first attempt to examine whether and how presenting scientific and social consensus information together influences people’s scientific beliefs when the two types of consensus information contradict each other. Unexpectedly, participants who received information about either high scientific consensus or low social consensus on the safety of GM foods did not differ from those who received no consensus information in their beliefs about the issue.¹ However, participants who received both scientific and social consensus information had stronger GM safety beliefs than those who received only social consensus information or no consensus information. This finding suggests that in the presence of information indicating a low level of social consensus, the impact of communicating a high level of scientific consensus on people’s scientific beliefs does not decrease, but rather increases.

One possible reason for the effectiveness of presenting scientific and social consensus information together is that the two types of consensus information may have served as contrasting cases. A set of contrasting cases has been shown to help individuals differentiate distinctive features of target information from non-distinctive features (e.g. Gibson, 1969; Gick & Paterson, 1992). Similarly, participants who received scientific and social consensus information together may have noticed and realized a marked discrepancy between scientific and social consensus. Consistent with this possibility, participants in the scientific and social consensus condition perceived a wider gap between scientific and social consensus (M = 37.1, SD = 33.8) – which was computed by subtracting perceived social consensus from perceived scientific consensus – than participants in the scientific consensus conditions (M = 27.9, SD = 25.0), t(703.48) = 4.26, p < .001, d = 0.31. Furthermore, there was no significant correlation between perceived scientific and social consensus only for the scientific and social consensus condition (r = −.06; for the other conditions, rs = .31 to .49, ps < .001), suggesting that the simultaneous presentation of scientific and social consensus information helped participants clearly distinguish scientific consensus from social consensus. Given that group expertise is a determinant of conformity (Costanzo et al., 1968; Crano, 1970) and that in the present study, scientists were perceived to be more expert in the GM safety issue than ordinary people, the perceived consensus discrepancy may have encouraged participants in the scientific and social consensus condition to adjust their GM safety beliefs to the level of perceived scientific consensus.

However, it is important to note that even when social consensus information was presented together, scientific consensus information had only a small effect (d = 0.22) on participants’ GM safety beliefs. Although small effects can be meaningful at a population level, particularly considering that the effects are obtained by the simple method of consensus messaging (e.g. van der Linden, Leiserowitz, & Maibach, 2019), one must be cautious not to exaggerate the effectiveness of communicating highly divergent levels of scientific and social consensus.

In line with the findings of prior research on the mediating effect of perceived scientific consensus (e.g. Dixon, 2016; Kerr & Wilson, 2018; van der Linden et al., 2015), the present study found that the presentation of information about high scientific consensus substantially raised the level of perceived scientific consensus, which in turn contributed to the improvement of GM safety beliefs. This result reinforces the gateway belief model’s contention that perceived scientific consensus operates as the gateway to scientific belief change when a consensus message is communicated (van der Linden et al., 2015).

Contrary to our expectations, though, participants’ perceptions of social consensus did not act as a mediator. This result is inconsistent with prior findings that communicating a level of social consensus influenced people’s scientific beliefs through their perceptions of social consensus (Kobayashi, 2018; Puhl et al., 2005; Stangor et al., 2001). In the present study, information about low social consensus slightly lowered the level of perceived social consensus (d = −0.33) only when information about high scientific consensus was presented together, suggesting that participants’ sensitivity to the level of communicated social consensus was not as high as their sensitivity to the level of communicated scientific consensus. Additionally, the effect of perceived social consensus on GM safety beliefs was negligibly small (f² < 0.01). It seems that participants attached little importance to what was the consensus among ordinary people on the issue of GM safety.

There are some limitations to the present study. First, in the present study, the level of scientific consensus was manipulated in one direction (high) only, whereas the level of social consensus was manipulated in the opposite direction (low) only. This was because information about scientific and social consensus was created to reflect the (probable) opinion gap between scientists and the general public regarding the safety of GM foods in Japan. Additionally, to our knowledge, there is no evidence suggesting that the effectiveness of communicating the scientific and/or social consensus is asymmetrical between the two directions (see e.g. Klucharev et al., 2009). Nevertheless, future work should include independent manipulation of consensus direction and type. Second, we did not measure or manipulate participants’ perceptions of the extent to which they identify themselves with or are similar to ordinary people. Previous work has shown that information about social consensus has a greater impact on people’s perceptions of social consensus and beliefs when the information comes from an ingroup source than when it comes from an outgroup source (e.g. Puhl et al., 2005; Stangor et al., 2001). In cases where individuals see ordinary people as ‘us’ or like-minded others, the presentation of social consensus information may substantially influence the individuals’ scientific beliefs through their perceptions of social consensus. This possibility should be considered in future research. Finally, the present study focused on samples of Japanese people and the issue of GM safety, limiting the generalizability of the present findings. Indeed, evidence suggests that the impact of communicating a level of scientific or social consensus may differ according to country (Cook & Lewandowsky, 2016) and issue (Kerr & Wilson, 2018; Kobayashi, 2018). It will be important to replicate the results of the present study, using samples coming from different countries and other scientific issues.

Despite these limitations, the present findings have practical implications for science communication. Our results suggest that presenting scientific and social consensus information together helps people distinguish scientific consensus from social consensus and adjust their consensus perceptions to the levels of communicated scientific and social consensus. Using the two types of consensus information as contrasting cases would serve to highlight the actual level of scientific consensus. Most importantly, the present study showed that informing participants about both the high scientific consensus along with the low social consensus increased their GM safety beliefs in some degree. This implies that when there is a wide gap between scientific and social consensus on a scientific issue, the simultaneous presentation of scientific and social consensus information may even slightly improve people’s beliefs about the issue and thereby contribute toward narrowing the wide gap.

Data availability statement

The data that support the findings of this study are available from the corresponding author, Keiichi Kobayashi, upon reasonable request.

Disclosure statement

No potential conflict of interest was reported by the author.

Supplementary material

Supplemental data for this article can be accessed here.

Additional information

Funding

This work was supported by the Japan Society for the Promotion of Science [Grant-in-Aid of Scientific Research (C) / No. 15K0].

Notes

1. An additional analysis found that the effect size for the impact of only scientific consensus information (d = 0.12) observed in the present study was within the rage of variability of those found in previous studies on the issue of GM safety (see the supplement online material).