Abstract
This article aims to analyze the factors influencing belief in a flat Earth. We focus on Spain, a country that sadly has some of the most relevant figures on this topic in the Spanish-speaking world. After a qualitative analysis of YouTube videos on the main channels on the subject, a survey was carried out on 1252 individuals. The results point to two conclusions. First is the presence of a considerable Dunning–Kruger effect among flat-earthers. There is a significant negative correlation between science literacy—overall and practically all its dimensions—and overconfidence in science in this group. The second, evaluated through a regression tree, confirms that the interaction of low scientific literacy and overconfidence is highly relevant in explaining the belief in a flat earth. Neither factor alone is determinant, but the combination of the two (low scientific literacy and high overconfidence) leads to high levels of flat earth belief.
1. Introduction
When we started this research project, other academics we discussed the project with were skeptical we would find any individuals who believed the earth to be flat. It simply was not possible, especially considering the sample we planned to work with, composed mainly of highly educated individuals. A few months later, when we shared with them that, out of a sample of 1252 people, 16 (1.3%) were fully convinced that the earth was flat, skepticism gave way to incredulity. Especially when we pointed out that only 82.5% of the sample were fully convinced that the earth is not flat. Nonsense? Perhaps not so much if we put it in context. A YouGov poll conducted in February 2018 concluded that only about 82.5% (precisely the same percentage obtained in our study) of North American millennials agreed with the statement: “I have always believed the world is round” (Foster and Branch, 2018). The fact that both figures are the same is most likely a coincidence, but it illustrates the extent to which the percentages obtained in our sample are not far from what happens in other collectives and countries. It is indeed possible that some respondents are simply being ironic or expressing doubts regarding the earth’s shape for other reasons: “being a young flat earther is a very edgelord thing. Or maybe [millennials] are just sick of being asked dumb survey questions” (Kahn, 2018). Some respondents surely fall into this category. However, if they accounted for most doubters, the statistical analysis would reveal a random pattern in their profiles and not the consistent one we have found and described below.
Some celebrities have openly supported—and continue to support—the flat earth theory. Basketball players Kyrie Irving (see Weedon, 2020 for a detailed analysis of his discourse on the subject) and Shaquille O’Neal or singers like Bobby Ray Simmons (B.o.B.) are examples of this support. In Spain, there is no shortage of defenders of this theory either: in 2019, former soccer player Javi Poves created Flat Earth FC, “the first club in the world that fights to spread the truth about the world we live in,” although this initiative lasted barely 2 years. One of the main disseminators of the flat earth theory in Spanish is the Spaniard Oliver Ibañez, author of the book “The biggest conspiracy in history,” whose YouTube account had around 540,000 subscribers as of November 2022.
We have reached a point where the popularity of flat earth theory is such that articles have been published to help instructors refute flat earth arguments put forward by their students (Břízová et al., 2018), and scientists have written papers explaining why the earth cannot be flat (Arif et al., 2019; De Marez and Corre, 2020; Kuzii and Rovenchak, 2019). It is hard to believe this should be necessary for the twenty-first century.
2. State of research on flat-eartherism
Research on flat-eartherism is not as extensive as on other unjustified beliefs. Nevertheless, several papers have reached various conclusions about this phenomenon. First, as Mohammed (2019 88) pointed out, “[a] wave of flat-eartherism emerged with the increasing popularity of social media and the accessibility of video treatises on the subject in the early decades of the 21st century” and “[f]or the past 10 years, the idea that the earth may be flat has experienced a revival” (De Marez and Corre, 2020: 1). Certainly, there is evidence that Flat earth videos in YouTube are a key element in the dissemination of this theory (Olshansky, 2018; Olshansky et al., 2020; Paolillo, 2018), that takes hold in 2014 (Paolillo, 2018). Nowadays, we can find thousands of videos about it and specialized YouTubers whose contents are almost exclusively focused on the flat earth theory.
Second, several studies have demonstrated that flat earth believers show high levels of conspiracy ideation (Landrum and Olshansky, 2019) interacting with science intelligence 1 (Landrum et al., 2021: 136): “whereas people with lower conspiracy mentality do not find the [flat earth] arguments compelling at any level of science intelligence, among those with conspiracy mentality, perception of argument strength decreases as science intelligence increases.” So, low science intelligence and high conspiracy mentality seem to increase susceptibility to flat earth arguments.
Religiosity could also play a role in the conversion to flat-eartherism. Mohammed (2019) concluded that flat-earth videos were more likely to include conspiracy ideation and science denial than debunking videos, which is to be expected, but they were also more likely to include religious content. However, the effect of religion is far from clear, and Olshansky’s (2018) work concludes that “Religiosity [. . .] may not be as motivating a factor as previously thought [. . .] [but] high religiosity combined with high conspiracy mentality, could be” (p. 40). That is, we again find a possible interaction effect with conspiracy mentality.
Finally, flat-earthers are very skeptical of authority and doubt the legitimacy of proven facts (Olshansky, 2018). As Erlaine (2020) pointed out, “the total rejection of science in favor of opinion is today a feature of the discursive landscape, and nowhere it is more poignant than in flat earth belief” (p. 173). Nevertheless, flat-earthers consider themselves more logical than the general population (Olshansky, 2018). This feature of flat-earthers is extremely interesting. Imhoff and Lamberty (2017) and Lantian et al. (2017) found an association between the self-attributed need for uniqueness and a conspiracy mentality. It is worth asking to what extent this view of themselves, as more logical than the general population, is related to a need for uniqueness. In any case, and whatever the causes, we note a possible overconfidence bias, which has not yet been sufficiently studied in the literature. In fact, after viewing all the flat-earth theory videos and related user comments on Oliver Ibañez’s channel mentioned above and other popular Spanish-language channels on the subject, we identified certain patterns that could indicate overconfidence. This possible trait is what we found most striking and became the research question of this article: is there a relationship between overconfidence bias and belief in a flat earth? Our research hypothesis is that scientific literacy and overconfidence are interacting variables that significantly impact the belief in a flat earth, at least in the case of Spanish speakers.
3. Method
Data
The data were obtained through a survey collected in two waves and distributed online through Google Forms. In the first wave, conducted between 26 January and 10 February 2021, the participants were undergraduate students at Comillas Pontifical University (Spain). A total of 666 responses were received. It was necessary to eliminate 18 due to various problems detected in the preliminary analysis, such as incomplete information or errors in the sanity check—which requested the sum of 5 plus 6. The second wave was conducted between 1 September and 8 October, 2021, and included Spanish individuals from all backgrounds using the snowball procedure. A total of 618 responses were obtained in this wave, 14 of which had to be eliminated. Thus, the final sample consisted of 1252 valid records. All participants were of legal age and gave their consent for their responses, which were treated anonymously, to be used for research purposes.
The sample has a clear bias toward individuals with a high socioeconomic level (37.1% declared a monthly family income of over €5000) and a high level of education (71.7% of university students). Thus, rather than reflecting the Spanish population as a whole, this sample has a bearing on a specific group of great relevance for the country’s future: people who, as a result of the considerable direct effect of social origin on success in the labor market in Spain (Bernardi, 2012), are likely to occupy relevant positions in the corporate world or public administration.
The sample combines two characteristics that make it distinctive and appropriate for examining the weight of overconfidence in anti-scientific beliefs. First, most participants have a high education level, as they are university students and have completed more than 12 years of formal education. Second, a significant number of participants belong to the upper classes. This extraction has been linked to higher rates of overconfidence, a characteristic that makes people seem more prepared, competent, and likely to achieve high social status (Belmi et al., 2020).
Procedure
The R programming environment has been used for the empirical study. After a preliminary analysis, a regression tree was constructed for the target variable (belief in a flat earth) using CART, classification, and regression trees (Breiman et al., 1984). To prevent overfitting, we have used the rule of a minimum of 50 elements to split a node and a minimum of 20 to create a node (pre-pruning), and cross-validation based on the minimum relative error to estimate the correct complexity parameter (post-pruning). We chose this econometric approach over ordinary least squares because it is particularly useful to carry out an exploratory analysis focusing on evaluating each variable’s relevance and, especially, their interactions. It also makes fewer assumptions about the data than alternative techniques. The procedure creates a model that predicts the value of the variable “belief in flat earth” based on several input factors. The dependent variable, “belief in a flat earth,” is measured using a 0–10 scale based on the question: “Consider the following statement: There is evidence that the earth is flat. Indicate your opinion, 0 meaning strongly disagree, and 10 strongly agree.”
Regarding the predictors, first, we have included three control variables: “age” (expressed in years), gender (variable “female”: 1 for female and 0 for male), and “income level” (measured on a scale ranging from “very low income” (category 1), corresponding to a household’s family income below 1100€ per month, to “very high income” (category 6), corresponding to household’s family income over 5000€ per month).
Previous works have reported a significant correlation between a belief in conspiracy theories and a belief in the paranormal (Brotherton et al., 2013; Darwin et al., 2011). For this reason, we have incorporated the variable “belief in astrology” that takes values from 0 (no belief in astrology) to 10 (strong belief in astrology).
Regarding the first variable of interest in this article, science intelligence, Landrum et al. (2021) pointed out that it predicts acceptance of scientific issues. However, this acceptance is sometimes moderated by other factors, such as religiosity and political orientation (Kahan, 2017). Therefore, it seems necessary to incorporate these variables into the model. “Scientific literacy” was measured using the OSI 2.0 scale (Kahan, 2017), a 0–19 scale, where 0 represents very low scientific literacy, and 19 represents very high. The original scale was used with a slight modification: in questions regarding earth movements (“does the earth go around the Sun, or does the Sun go around the earth?” and “How long does it take for the earth to go around the Sun?”) the text “according to official science” was added. As the objective was to study flat earth theory, it was necessary to distinguish between an individual’s belief (measured in another question) and knowledge of established scientific facts (what these questions aim to measure). “Religious orientation” was measured using five categories: practicing Catholic, non-practicing Catholic, believer of another religion, indifferent or agnostic, and atheist. Finally, “political orientation” was measured on a 0–10 scale, where 0 represents far-left, and 10 represents far-right.
The second variable of interest, overconfidence, has been divided into two items. The first refers to the level of general overconfidence. It was measured on a scale from 1 (low overconfidence) to 5 (high overconfidence). Factor 4 (lack of intellectual overconfidence) of the questionnaire proposed by Krumrei-Mancuso and Rouse (2016) was used. It consists of six items rated on a 5-point Likert-type scale ranging from 1—strongly disagree to 5—strongly agree. It includes questions such as “My ideas are usually better than other people’s ideas” or “When I am really confident in a belief, there is very little chance that belief is wrong.” The level of overconfidence was calculated as the average of these six items. The second one, “Overconfidence in science,” refers to a person’s level of overconfidence in his or her own scientific knowledge. It ranges from 1 (low overconfidence in one’s scientific knowledge) to 5 (high overconfidence in one’s scientific knowledge). Measurement is based on an adaptation of the question proposed by Motta et al. (2018): “There are probably some scientific topics that interest you. In relation to these topics, do you think you know more or less than scientists?.” Options were coded from 1 (“I know a lot less”) to 5 (“I know a lot more”).
4. Results and discussion
Preliminary analysis
When analyzing independently the group of individuals who are confident of the spherical shape of the earth (Belief in flat earth = 0) and those who express some doubt about the possibility that it is flat (Belief in flat earth > 0), we find substantial differences in the level of science literacy. Figure 1 shows the distribution of both groups, with vertical lines representing average values. The Anderson–Darling test confirms that the two distributions are significantly different (p-value < .001). The same applies to overconfidence in scientific knowledge (Figure 2). Among flat-earth believers, 24.7% claim to know the same or more than scientists (Overconfidence in science ⩾ 3), while this percentage is only 2.2% among skeptics.
Smoothed density distribution of scientific literacy (OSI_2.0, Kahan, 2017) measured on a 0–19 scale. The figure was generated using a smoothing kernel set to “Gaussian.”
Bar plot of overconfidence in science measured on a 1–5 scale.
The science literacy scale we used (Kahan, 2017) distinguishes between four different dimensions: Basic Facts (knowledge of certain basic scientific facts), Methods (items relating to the understanding of how science works), Quantitative Reasoning (proficiency in reasoning with quantitative information), and Cognitive Reflection (based on items which require respondents to resist crediting an intuitive answer that can be shown to be incorrect by logical analysis). Table 1 shows the correlations between overconfidence in science and science literacy at the aggregate level (OSI 2.0 Global) and split along the four dimensions mentioned above. None of the correlations are significant among skeptics, with a confidence level of 1%. While all correlations are significant among believers except with “Basic Facts.” The most interesting issue is that correlations are negative. The lower the actual scientific knowledge, the higher the level of overconfidence in one’s scientific knowledge. Indeed we are dealing with a Dunning–Kruger effect (Dunning, 2011): the less competent a person is, the more he or she overestimates him/herself, which manifests itself both at the aggregate level and in practically all dimensions of scientific literacy. Imhoff and Lamberty (2017) and Lantian et al. (2017) concluded that the endorsement of conspiracy theory might be partially motivated by the need for uniqueness, which could perhaps explain why flat-earth believers exhibit such abnormally high levels of overconfidence: they need to believe they are unique.
Spearman correlations between scientific literacy and overconfidence in science. Correlation coefficient (p value).
In addition, comparing the average scores of both groups in each of the dimensions of scientific literacy, we confirmed that the believers obtained lower scores in all of them. The cognitive reflection dimension is particularly striking, showing the starkest differences. Believers in the flat earth theory score an average of 0.56 (on a 0–3 scale) compared with 1.08 for skeptics. Confirming the idea, already noted in previous works, that belief in conspiracy theories is associated with lower analytic thinking (Swami, 2018; Swami et al., 2014) and a lack of critical thinking (Lantian et al., 2021). Moreover, among flat earth believers, there is a significant (p = .007) and positive (rho = .18) correlation between belief in a flat earth and belief in astrology. It is fascinating that such a correlation exists when the two beliefs are entirely incompatible, pointing again to a lack of analytic thinking.
Regression tree
The regression tree (Figure 3) and the weights of the different variables (Figure 4) confirm the hypothesis of this article. The most important explanatory variable for flat earth belief is overconfidence in science, followed closely by scientific literacy. Moreover, both variables interact, and the highest levels of flat-earth belief are obtained by the combination of low scientific literacy and high overconfidence. Generic overconfidence plays a similar–if less relevant–role. Individuals with low scientific literacy and overconfidence in science, but high general overconfidence, also have high levels of flat-earth belief, especially if they are men.
Regression tree (dependent variable: belief in a flat earth in a 0–10 scale).
Standardized importance of the variables in the regression tree model.
Concerning the control variables incorporated into the model, it seems religiosity is not as important as suggested by some previous research, which coincides with the results of Olshansky (2018). Nor does political ideology seem to play a determinant role, coinciding with the findings of Mohammed (2019). Belief in astrology, on the contrary, is of relative importance. Although there seems to be some relationship between magical thinking and belief in a flat earth, the effect is smaller than other variables.
5. Conclusion
In this article, we aimed to identify which factors influence the belief in a flat earth, focusing on the level of scientific literacy and overconfidence in general and related to scientific literacy. We hypothesized that scientific literacy and overconfidence are interacting variables that significantly impact the belief in a flat earth.
Using a sample of 1252 individuals, we confirmed that both variables are highly relevant. Neither factor alone seems to be determinant, but the combination of the two is. Thus, the highest levels of flat earth belief are obtained when both low scientific literacy and high overconfidence are present.
Two conclusions can be drawn from these results. The first is that the importance of this issue should not be underestimated. The fact that 17.5% of respondents expressed some degree of doubt about the earth’s shape is very worrying. If we doubt something so indisputable, science becomes a matter of opinion, subject to the whims of the individual, who can decide what is or is not true according to his or her tastes (Weedon, 2020). Nevertheless, as it appears harmless, we may be tempted to play down the importance of a belief like flat-earthism. However, other similar beliefs, such as accepting vaccines cause autism and refusing to vaccinate children, are not harmless. Both ideas are arrived at through the same path, which seems related to the Dunning–Krueger effect (Motta et al., 2018).
The second conclusion is that more research is needed to devise ways to reduce the Dunning–Kruger effect among particularly susceptible individuals. This effect is present in several pseudoscientific and conspiracy beliefs, such as the flat earth, the anti-vaccine movement (Motta et al., 2018), or COVID-19 conspiracy theories (Vranic et al., 2021).
The first limitation of this study is that the design is not suitable for drawing causal conclusions but can only show correlations. A second limitation is that the sample is not representative of the Spanish population, as it is biased toward individuals with a high socioeconomic profile and level of education. This bias is a weakness, but at the same time, it is of interest in that it offers us insight into a group of individuals destined to occupy positions of relevance in the business world and public administration. It would be dramatic to end up with a president who believed that the earth is flat, and, given our results, it is not something we can completely rule out.
Supplemental Material
sj-docx-1-pus-10.1177_09636625231166255 – Supplemental material for Dunning–Kruger effect and flat-earthers: An exploratory analysis
Supplemental material, sj-docx-1-pus-10.1177_09636625231166255 for Dunning–Kruger effect and flat-earthers: An exploratory analysis by Jose Luis Arroyo-Barrigüete, Carlos Bellón Núñez-Mera, Jesús Labrador Fernández and Victor Luis De Nicolas in Public Understanding of Science
Footnotes
Acknowledgements
The authors would like to thank the two anonymous reviewers for their helpful comments on the manuscript, which have contributed to substantially improving the final version of this article.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
ORCID iDs
Supplemental material
Supplemental material for this article is available online.
Notes
Author biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
