Fake news and conspiracy theories spreading over the internet are a major challenge to public debate. How can we address this challenge? I focus on the dispositions of individuals, as there is some evidence that there are strong individual differences in the propensity to endorse and spread fake news and conspiracy theories (Lazer et al. 2018). My focus is on intellectual virtues and vices … [please read below the rest of the article].
Meyer, Marco. 2019. “Fake News, Conspiracy, and Intellectual Vice.” Social Epistemology Review and Reply Collective 8 (10): 9-19. https://wp.me/p1Bfg0-4tp.
The PDF of the article gives specific page numbers.
- “Trust in a Social and Digital World,” Mark Alfano and Colin Klein
- “Fake News, Conspiracy, and Intellectual Vice,” Marco Meyer
- “Beyond Testimony: When Online Information Sharing is not Testifying,” Emily Sullivan
- “Vices of Distrust,” J. Adam Carter and Daniella Meehan
- “Richly Trustworthy Allies,” William Tuckwell
- “The Trouble With ‘Fake News’,” David Coady
- “Algorithm-Based Illusions of Understanding,” Jeroen de Ridder
- “Is Conspiracy Theorising Irrational?” Neil Levy
Fake news and conspiracy theories spreading over the internet are a major challenge to public debate. How can we address this challenge? I focus on the dispositions of individuals, as there is some evidence that there are strong individual differences in the propensity to endorse and spread fake news and conspiracy theories (Lazer et al. 2018). My focus is on intellectual virtues and vices. Intellectual virtues are character traits that support their bearers in gaining knowledge and understanding. Intellectual vices are deficits in intellectual virtue, undermining the ability to gain knowledge and understanding.
I present initial findings from a survey experiment measuring intellectual virtue and vice, fake news endorsement, and conspiracist beliefs. I collected data on the intellectual virtues of 975 people from the United States, eliciting their intellectual virtues using a validated survey instrument. Early analysis shows that intellectually vicious people are more likely to endorse conspiracy theories. This finding supports claims by virtue epistemologists that conspiracy theorists suffer from intellectual vice (Quassim Cassam 2016). Yet the current experiment looks beyond conspiracy theories by showing that endorsement of fake news is associated with epistemic vice also.
The experiment also makes a contribution to vice epistemology. Vice epistemology is the branch of epistemology that concentrates on the nature, identity, and epistemological significance of intellectual vices (Cassam 2016). Quassim Cassam has suggested that some intellectual virtues and vices may be “stealthy”. A trait is stealthy if possessing the trait stands in the way of knowing that you have the trait (Quassim Cassam 2015). The current experiment uses a self-assessment approach to measure intellectual virtue (Alfano et al. 2017). That this measure is associated with questionable beliefs suggests that intellectual vices are not fully stealthy. People seem to have some knowledge about their intellectual character traits.
I study epistemic virtue and vice in an online environment. Participants for the study were recruited via Amazon Mechanical Turk, an online crowdsourcing platform. This is an appropriate setting to study conspiracy theories and fake news spread over the internet. The experiment suggests that epistemic virtue appears to influence whether people place trust intelligently online.
Section 1 introduces the experiment and describes the sample in terms of its intellectual virtues and their propensity to endorse conspiracy theories and fake news. Section 2 shows regression results of the experiment. Section 3 discusses implications for virtue and vice epistemology. I focus on whether vices are stealthy and on implications for trust on the internet. I conclude with reflections on next steps for the analysis.
1. Overview of the Experiment
Participants were recruited using Amazon Mechanical Turk. The eligibility criteria were living in the United States and being 18 or older. Respondents were paid $2 for participation. 1,357 people participated, of which 975 passed an attention check. Participants who failed the attention check were excluded from analysis. The participants were between 24 and 74 years old, with an average age of 40. 52% of the sample was female. 79% of the sample was White/Caucasian, 9% was African American/Black, 5% was Asian or Pacific Islander, and 5% was Hispanic; the remaining 2% were other or did not disclose ethnicity. 53% have obtained a bachelor’s or a higher degree. Mean household income is 57,000 USD per year. Table 4 in the appendix contains full descriptive statistics.
Each participant answered demographic questions about age, gender, race, education, household income, religion, and political affiliation. I measure religiosity by asking respondents how important religion is to them, on a five-point scale from “not at all important” to “extremely important.” I measure political affiliation by asking participants whether they “consider themselves a Republican, a Democrat, an Independent, or what?” Responses are “Strongly Democratic”, “Weakly Democratic”, “Independent (Lean toward Democratic party)”, “Independent”, “Independent (Lean toward Republican party)”, “Republican (Weakly Republican)”, “Republican (Strongly Republican)”.
In the following I discuss the instruments used to measure intellectual virtue, conspiracy theory, and fake news endorsement.
1.1 Measuring Intellectual Virtue
Intellectual virtues are character traits that support their bearers in gaining knowledge and understanding. Intellectual vices are deficits in intellectual virtue. Intellectual vices undermine the ability of their bearers to gain knowledge and understanding. I measured intellectual virtues using a validated survey instrument (Alfano et al. 2017). The scale provides a measure of intellectual humility. Intellectual humility is only one among many intellectual virtues. However, Alfano et. al have worked with an extensive definition of humility. Using 23 items, the scale measures four virtues: Open-mindedness, intellectual modesty, engagement, and corrigibility.
The constructs are defined in Table 1 below. While the four dimensions do not provide a comprehensive measure of intellectual virtue, the measure is broad enough for the purposes of this experiment. Responses were scaled on a five-point agree-disagree scale. The measure has high internal reliability. For a full list of items, please refer to the validation paper by Alfano et al. (2017).
|Intellectual Virtue||Definition||Example item|
|Acknowledgment of the limitations of one’s knowledge, especially relative to others, and a desire to gain knowledge irrespective of status.||I don’t take people seriously if they’re very different from me. (R)|
|Low concern for how one’s intellect is perceived, and for one’s intellectual reputation.||I like to be the smartest person in the room. (R)|
|Motivation to investigate things one doesn’t understand, particularly in response to encountering ideas different from one’s own.||A disagreement is like a war. (R)|
|Resilience in emotional response when confronted with challenges to one’s knowledge or intellectual abilities.||I appreciate being corrected when I make a mistake.|
Table 1: Overview of Alfano et al.’s Intellectual Humility Scale
For the initial analysis of the data, I constructed a summary measure of intellectual virtue rather than analysing each virtue separately. I obtained the measure by summing the weighted responses to all items. Weights are the factor loadings of the first factor from a principal component analysis of all items.
1.2 Measuring Conspiracist Thinking
Conspiracy theories are explanations for (assumed) phenomena that invoke a conspiracy. Some conspiracy theories are true (Dentith 2016; Harris 2018). I am interested in conspiracy theories that are not inferences to the best explanation (Harman 1965). Some of these conspiracy theories may still turn out to be true. But I maintain, in line with Quassim Cassam, that believing such conspiracy theories is a sign of intellectual vice (Quassim Cassam 2016).
To elicit the propensity to endorse conspiracy theories, I used an established measure from political science (Oliver and Wood 2014). Participants were presented with five conspiracy theories in random order and asked whether the statements presented were true or false, on a five-point scale (“definitely true”, “probably true”, “do not know”, “probably false”, “definitely false”). Table 2 below presents the items and proportion of respondents endorsing each of the statements. I took respondents to endorse a statement if they replied “true” or “definitely true”. More than one third of respondents endorsed at least one conspiracy theory.
|The US invasion of Iraq was not part of a campaign to fight terrorism but was driven by Jews in the U.S. and Israel.||10%|
|Certain U.S. government officials planned the attacks of September 11, 2001, because they wanted the United States to go to war in the Middle East.||18%|
|President Barack Obama was not really born in the United States and does not have an authentic Hawaiian birth certificate.||13%|
|The financial crisis of 2008/09 was secretly orchestrated by a small group of Wall Street bankers to extend the power of the Federal Reserve and further their control of the world’s economy.||19%|
|Billionaire George Soros is behind a hidden plot to destabilize the American government, take control of the media, and put the world under his control.||16%|
Table 2: Measure of Conspiracist Thinking
1.3 Measuring Fake News Endorsement
Fake news refers to content that presents (typically) false or misleading claims as news (Lazer et al. 2018; Gelfert 2018).
I developed a new instrument to elicit the propensity of respondents to endorse fake news. Each respondent was presented with ten screenshots of articles from news and fake news websites in random order. Participants were asked to rate the screenshots according to how credible they find the article, on a five-point scale. Figure 1 gives examples of a news and a fake news item as presented to participants of the study.
Figure 2 shows the proportion of respondents who find fake news items credible. Items were deemed credible by between 11% and 50% of respondents. I took respondents to find an article credible if they replied “agree” or “strongly agree” to the question whether the article was credible. Almost four of five respondents found at least one of the fake news items presented to them credible.
It is noteworthy that endorsement of conspiracy theories and endorsement of fake news are correlated. The number of fake news items deemed credible and the number of conspiracy theories endorsed correlates with a coefficient of 0.6. This result suggests that there might be an underlying factor explaining both types of pernicious beliefs.
This section presents the results of a regression analysis to test whether endorsement of conspiracy theories or fake news is associated with intellectual vice. A regression approach goes beyond showing mere correlations between intellectual vice and the endorsement of questionable beliefs. Such correlations could be caused by some underlying third factor. Other explanations that have been suggested in the literature appeal to education, socio-economic background, political orientation, religion, or news consumption (Brotherton, French, and Pickering 2013; Oliver and Wood 2014; Hagen 2018; Lazer et al. 2018; Allcott and Gentzkow 2017).
To vindicate intellectual vice, it should explain questionable beliefs over and above other, established explanations. Regression analysis allows to test associations between outcomes and our measure of intellectual vice while controlling for these other factors.
Table 3 shows regression results for conspiracy theories and fake news endorsement. Columns 1 and 2 concern conspiracy theories measured as the number of conspiracy theories endorsed. Columns 3 and 4 concern fake news endorsement measured as the number of fake news items deemed credible. I have normalized both outcome measures by calculating the z-score for each observation.
Columns 1 and 3 show regression results using only control variables. Controls used are age, household income, sex, education, ethnicity, political affiliation, religion, and news consumption. Coefficients can be compared with one another because all discrete variables have been normalized by computing their z-scores.
The reason to show these results is that they give us a baseline for how much of the variance in endorsement of conspiracy theories and fake news is accounted for by control variables. These regressions account for 16% and 26% of variance as measured by R2, respectively. This result implies that our control variables account can explain 16% of variance between respondents in endorsing conspiracy theories. Controls explain 26% of variance in endorsing fake news.
Political affiliation and religion explain most variance, followed by obtaining news from social media. For conspiracy theories, household income and news consumption explain some variance as well. People who get their news mainly from printed newspapers are somewhat less likely to believe in conspiracy theories, as are people who get their news from news aggregators such as Google news. People who get their news mainly from social media are slightly more likely to believe in conspiracy theories.
Other variables show no statistically significant association with conspiracy or fake news. Age, education and ethnicity are not significantly associated with outcomes. Female respondents are slightly more likely to endorse conspiracy theories, but the coefficient is significant only at the 10% level.
Columns 2 and 4 show regression results for the same set of controls plus the summary measure of intellectual virtue. For both outcomes, the association with intellectual virtue is statistically significant at a 1% level. The coefficient of intellectual virtue is larger than the coefficient of any of the controls. The proportion of variance we can explain jumps by 10 and 8 percentage points, respectively. This result suggests that intellectual vice explains endorsement of conspiracy theories and fake news over and above alternative explanations as measured by controls.
Political affiliation and religion remain statistically significant, with coefficients at the same order of magnitude as intellectual virtue. It is noteworthy that sex becomes statistically significant at the 1% level for fake news endorsement once we account for intellectual virtue. Female respondents were somewhat more likely to endorse fake news.
|Conspiracy Theories||Conspiracy Theories||Fake news||Fake news|
|High school diploma or equivalent||-0.429||-0.563||0.0308||-0.0919|
|Some college but no degree||-0.460||-0.610||-0.0959||-0.233|
|Asian or Pacific Islander||-0.190||-0.0931||-0.232||-0.143|
|Black or African American||-0.172||-0.0223||-0.130||0.00734|
|White / Caucasian||-0.202||-0.108||-0.0982||-0.0120|
|TV and Radio||0.00782||-0.00663||0.0322||0.0189|
|Standard errors in parentheses|
|*** p<0.01, ** p<0.05, * p<0.1|
Table 3: Regression Results
3. Implications for Vice Epistemology
What can vice epistemology learn from these results? The descriptive statistics show that questionable beliefs in fake news and conspiracy theories are widespread. Some established explanations account for some of the variance in outcomes. Republicans are more likely to endorse both conspiracy theories and fake news. Religion is also associated with conspiracist thinking and endorsement of fake news.
Intellectual vice appears to be an additional ingredient to the explanation. Intellectually vicious people are more likely to endorse conspiracy theories. This result holds up when we control for political orientation, religion, and a range of other factors. This finding supports claims by virtue epistemologists that conspiracy theorists suffer from intellectual vice (Quassim Cassam 2016). Intellectual vice is associated with questionable beliefs other than conspiracy theories. The experiment shows that fake news is associated with epistemic vice as well.
Most people encounter conspiracy theories and fake news primarily online. This study suggests that intellectual virtue and vice influence epistemic conduct in an online environment. Epistemic virtue appears to influence whether people place trust intelligently online (O’Neill 2002). As a next step, it would be interesting to compare the influence of intellectual virtue and vice in an offline setting.
I administered the scale as a self-assessment questionnaire. Self-assessment has two advantages. First, data gathering is relatively unproblematic. Through online services like Amazon Mechanical Turk researchers have easy access to a large pool of participants (Buhrmester, Kwang, and Gosling 2011; Paolacci and Chandler 2014). Second, participants retain a high degree of autonomy over how they are described and rated. But this latter feature also gives rise to a challenge to self-assessment.
Can we know our own vices? Ignorance about one’s vices is a challenge because people can only overcome their epistemic flaws once they recognize them. The self-assessment methodology is premised on the assumption that people have at least some insight into their own character traits (Vazire 2010). The method does not require that people have a sophisticated conceptual understanding of intellectual virtue and vice. Rather, each intellectual trait is measured by aggregating responses to a number of agree-disagree items related to concrete behaviours, attitudes, motivations, and skills. Still, respondents may lack the self-knowledge necessary to respond to items adequately (Dunning, Heath, and Suls 2004).
The philosophical correlate of the methodological problem with self-assessment is the problem of stealthy virtues (Quassim Cassam 2015). Traits are stealthy if possessing the trait stands in the way of knowing that you have the trait (Quassim Cassam 2015). Self-knowledge about intellectual humility in particular may be tenuous. One feature of the truly humble may be that they do not think about themselves as particularly humble. The boastful, on the other hand, are unlikely to fully appreciate their lack of intellectual humility. In effect, the pretentious as well as the self-depreciatory may well lack the self-knowledge necessary to answer questions on intellectual humility correctly. Since intellectual virtue is however associated with epistemic outcomes as expected, people appear to have some knowledge about their intellectual character traits. This finding is consistent with one of the validation studies for the survey used in this experiment. Alfano et al. conducted a study comparing self-ratings with ratings by informants and found positive correlations (Alfano et al. 2017, 12ff.). This result suggests that the scale picks up on some trait of subjects that they and informers judge similarly.
Will participants respond to the items in the Intellectual Virtue Scale truthfully, even if they have self-knowledge? Participants who want to appear intellectually virtuous can easily do so by selecting socially desirable items. The transparency of the scale limits its application to cases where respondents do not have strong incentives to answer in socially desirable ways. But in the absence of strong incentives to appear virtuous, the motive of self-discovery gives respondents a reason to answer truthfully. Since the scale relates to outcomes as predicted, the challenge of deception appears to be limited.
The initial results from the survey experiment suggest that intellectual virtue and vice are associated with fake news and conspiracist thinking. Intellectual virtue explains variance in the endorsement of conspiracy theories and fake news among respondents. Moreover, it has explanatory power over and above established explanations appealing to religiosity or political orientation.
The analysis of the data from the experiment is not complete. One next step in analysing the data from the experiment is to investigate the associations of individual virtues with conspiracist thinking and fake news. Another is to demonstrate the robustness of the results presented here. One element is to build and test various models of how the independent variables used in the regression are related to one another using structural equation modelling.
The experiment makes a methodological contribution by showing that intellectual virtue and vice can be measured by a self-assessment scale. This result is supported by the finding that the self-reported measures of intellectual virtue and vice are related to epistemic outcomes in expected ways. People appear to have a good sense whether they manifest the behaviours, attitudes, and motivations that reflect intellectual virtues or vices. Survey methodology can transform this knowledge into insights about intellectual vices.
The experiment demonstrates that empirical research can contribute to vice epistemology. Much remains to be done. Vice epistemologist have discussed intellectual vices including gullibility, dogmatism, prejudice, closed-mindedness, negligence, intellectual pride, idleness, cowardice, conformity, and rigidity. One important task is to develop a taxonomy of intellectual virtues and vices. Psychometric techniques provide compelling methods for developing such a taxonomy. We have only just begun to develop scales measuring individual intellectual virtues and vices. Eventually, experiments can contribute to answering the question which intellectual vices matter most. We need an empirical approach to investigate which intellectual vices are most harmful to gaining knowledge and understanding.
|High school diploma or equivalent||974||0.10||0.30||0.00||1.00|
|Some college but no degree||974||0.22||0.41||0.00||1.00|
|Asian or Pacific Islander||971||0.05||0.22||0.00||1.00|
|Black or African American||971||0.09||0.28||0.00||1.00|
|White / Caucasian||971||0.79||0.41||0.00||1.00|
|TV and Radio||975||3.38||1.24||1.00||5.00|
Table 4: Summary Statistics
Contact details: Marco Meyer, University of York, firstname.lastname@example.org
Alfano, Mark, Kathryn Iurino, Paul Stey, Brian Robinson, Markus Christen, Feng Yu, and Daniel Lapsley. 2017. “Development and Validation of a Multi-Dimensional Measure of Intellectual Humility.” PLOS ONE 12 (8): 1–28.
Allcott, Hunt, and Matthew Gentzkow. 2017. “Social Media and Fake News in the 2016 Election.” Journal of Economic Perspectives 31 (2): 211–36.
Brotherton, Robert, Christopher C. French, and Alan D. Pickering. 2013. “Measuring Belief in Conspiracy Theories: The Generic Conspiracist Beliefs Scale.” Frontiers in Psychology 4.
Buhrmester, Michael, Tracy Kwang, and Samuel D. Gosling. 2011. “Amazon’s Mechanical Turk: A New Source of Inexpensive, Yet High-Quality, Data?” Perspectives on Psychological Science 6 (1): 3–5.
Cassam, Quassim. 2015. “Stealthy Vices.” Social Epistemology Review and Reply Collective 4 (10): 19-25.
Cassam, Quassim. 2016. “Vice Epistemology.” The Monist 99 (2): 159–80.
Dentith, Matthew R. X. 2016. “When Inferring to a Conspiracy Might Be the Best Explanation.” Social Epistemology 30 (5–6):572–591.
Dunning, David, Chip Heath, and Jerry M. Suls. 2004. “Flawed Self-Assessment: Implications for Health, Education, and the Workplace.” Psychological Science in the Public Interest 5 (3): 69–106.
Gelfert, Axel. 2018. “Fake News: A Definition.” Informal Logic 38 (1): 84–117.
Hagen, Kurtis. 2018. “Conspiracy Theories and the Paranoid Style: Do Conspiracy Theories Posit Implausibly Vast and Evil Conspiracies?” Social Epistemology 32 (1): 24–40.
Harman, Gilbert H. 1965. “The Inference to the Best Explanation.” The Philosophical Review 74 (1): 88–95.
Harris, Keith. 2018. “What’s Epistemically Wrong with Conspiracy Theorising?” Royal Institute of Philosophy Supplements 84 (November):235–57.
Lazer, David M. J., Matthew A. Baum, Yochai Benkler, Adam J. Berinsky, Kelly M. Greenhill, Filippo Menczer, Miriam J. Metzger, et al. 2018. “The Science of Fake News.” Science 359 (6380): 1094–96.
Oliver, J. Eric, and Thomas J. Wood. 2014. “Conspiracy Theories and the Paranoid Style(s) of Mass Opinion.” American Journal of Political Science 58 (4): 952–66.
O’Neill, Onora. 2002. A Question of Trust: The BBC Reith Lectures 2002. Cambridge University Press.
Paolacci, Gabriele, and Jesse Chandler. 2014. “Inside the Turk: Understanding Mechanical Turk as a Participant Pool.” Current Directions in Psychological Science 23 (3): 184–88.
Vazire, Simine. 2010. “Who Knows What about a Person? The Self–Other Knowledge Asymmetry (SOKA) Model.” Journal of Personality and Social Psychology 98 (2):281.