The Dunning-Kruger effect, antivaxers, and the arrogance of ignorance

There’s a famous saying by Alexander Pope from 300 years ago that applies to antivaccination activists. It’s commonly truncated to: “A little learning is a dangerous thing.” However, I don’t like that short version of the quote, because it implies that learning is a dangerous thing. The full quote gives a much better idea: “A little learning is a dangerous thing; drink deep, or taste not the Pierian spring: there shallow draughts intoxicate the brain, and drinking largely sobers us again.” The basic idea is that a superficial knowledge (“a little learning” of a subject can be dangerously intoxicating. We think we know more than we really do and as a result are far more confident in our knowledge and abilities than we have a right to be. “Drinking deep” of knowledge “sobers us up” by forcing us to acknowledge the limits of our knowledge and abilities. In essence, even though we know a lot, we appreciate how little we know. In modern times, we have a term for this phenomenon, namely the Dunning-Kruger effect.

Basically, the Dunning-Kruger effect is named after (who else?) the social psychologists who first proposed it, David Dunning and Justin Kruger, and describes a form of cognitive bias in which people with low knowledge of a topic and/or ability in at a task develop illusory superiority and mistakenly assess their knowledge and ability as greater than it actually is. Basically, people who know a little about a topic often think they know a lot more about it than they, in fact, do, and people who are knowledgeable or even expert in a topic tend to underestimate their knowledge and abilities. Before I had ever heard of the Dunning-Kruger effect, I had already started using a term that sort of describes the effect, but not quite. Specifically, I’m referring to the “arrogance of ignorance,” particularly with respect to the antivaccine movement and people such as the “Thinking Moms’ Revolution.” Then, after I discovered the term “Dunning-Kruger effect,” I started using it to describe antivaxers. Basically, in the antivaccine movement, the Dunning-Kruger effect tend to take the form of parents who think that their University of Google knowledge trumps the knowledge of physicians and scientists who have dedicated large swaths of their lives to the rigorous study of conditions such as autism and the question of how vaccines work. But is the Dunning-Kruger effect really associated with antivaccine views?

Last week, I learned of a new study that sought to examine that very question. The study, by Matthew Motta of the University (Pennsylvania’s Annenberg Public Policy Center), Timothy Callaghan (Department of Health Policy and Management, Texas A&M University), and Steven Sylvester (History & Political Science Department, Utah Valley University) is hot off the presses in Social Science & Medicine and entitled Knowing less but presuming more: Dunning-Kruger effects and the endorsement of anti-vaccine policy attitudes. The basic idea is that the safety, efficacy, and benefits of vaccines are widely recognized by medical experts but public opinion towards vaccine policy is becoming less supportive. Their hypothesis is rather interesting. The authors refer to it as their Overconfidence Thesis:

Pursuant with the discussion above, we break down the Overconfidence Thesis into two sets of testable hypotheses. Hypothesis 1 suggests that individuals low in autism awareness – that is, low in autism knowledge (Hypothesis 1A), and high in autism misinformation endorsement (Hypothesis 1B) – should be more likely to express over-confidence in their own knowledge about autism, relative to medical experts. This hypothesis represents our attempt to integrate Dunning-Kruger effects into the study of citizens’ attitudes about vaccines and autism, as well as to provide an empirical test of Camarago and Grant’s theorizing on the subject.

As a result of Hypothesis 1:

We also suspect that overconfidence has important policy consequences. People who believe that medical professionals know less than they do about autism should exhibit more reservations toward mandatory childhood vaccinations against MMR – an expert endorsed policy (Hypothesis 2a) – and the role that experts play in the vaccination policymaking process more generally (Hypothesis 2b). Although there are many paths by which experts can influence the policymaking process, we focus on two here; their indirect role in informing the public and policymakers about technical issues, and their direct influence in constructing public policy. We refer to negative attitudes toward these roles as “expert discounting.”

To test these hypotheses, the authors used an online survey in December 2017 via Survey Sampling International. The survey asked respondents to answer questions on several topics relevant to health policy and current events, including the safety of genetically modified food, attitudes about plans to combat prescription opioid abuse, and climate change. The company invited 1,529 adults to take part in the study, and the survey had an 85% completion rate, for a total of 1,310 participants. Data were weighted to reflect population benchmarks on gender, race, income, and education, and autism awareness was estimated using two sets of questions: knowledge about autism and misinformation about the link between vaccines and autism. First, knowledge about autism was assessed using a ten question true/false test developed by the National Alliance for Autism Research. Next, they assessed autism misinformation based on responses to a survey asking, “Can vaccines administered to children at young ages cause them to become autistic?” Responses were given on a four-point scale, coded to range from “They definitely can” to “They definitely cannot.”

The next part of the survey examined overconfidence. The first item in this section asked whether the respondents think they know more or less than “medical doctors” about the causes of autism. A second item did the same for “scientists.” Response options were numerically coded to range from “I know a lot less” to “I know a lot more” on a six-point scale and averaged together for the two groups. Then, antivax policy attitudes were assessed based on the extent to which respondents agreed or disagreed (using a standard five-point Likert scale) with the following statement: “Parents should be able to decide NOT to vaccinate their children against measles, mumps, and rubella.” This part of the survey, combined with the part before it, provides an estimate of the Dunning-Kruger effect in this population, and this provides a basis for the next part of the study, whether the Dunning-Kruger effect impacts views towards vaccine policy and who should make it.

After that, expert discounting was examined using two different sets of measures. First, authors asked respondents how much they trust several different groups to provide them with “information about vaccinations.” Responses were given on a four-point numeric scale ranging from “Not at all” to “A great deal.” Trust in “medical experts” combines responses to two items (“The Centers for Disease Control” and “your personal doctor”), while “non-experts” combines response to two other prompts (“Friends/Family” and “Celebrities”). Finally, respondents were asked to state what role (if any) various groups should have in “making decisions about policy issues related to the vaccination of children under the age of 16.” Responses were given on a three-point numberic scale ranging from “no role at all” to “a major role” for two groups: “medical scientists” and “the general public.” As you would expect, the authors also tried to control for relevant potential confounders. These included political ideology; religious service attendance; gender; educational attainment; race; total yearly household income; and age.

Depressingly, the first result shows that there are a lot of people out there who think they know more than scientists and doctors about autism:

When we decomposed the overconfidence measure into its two constituent parts (described in the methods section), we found that more than a third of our sample believed that they knew as much as or more than medical doctors (36%) and scientists (34%) about the causes of autism. The idea that a substantial number of U.S. adults think that they know more than medical experts about the causes of autism is well-reflected in our sample. The statistics further suggest that many respondents trust information from experts and think that they should play a major role in the policymaking process (Table 1). However, many also place high levels of trust on information from non-experts (42%) and feel that non-experts should play a major policymaking role (38%). We also noted that while our sample was moderately well informed about the causes of autism, misinformation endorsement was also fairly high.

The authors also found that both knowledge and misinformation were correlated with increased confidence in one’s own knowledge relative to experts. Using a conservative estimate, moving from low to high levels of autism knowledge was associated with a 39% decrease in overconfidence, while moving from low to high levels of misinformation endorsement was associated with a 17% in overconfidence. (Only 17%, I ask? That seems low.) Basically, people lowest in autism knowledge, on average, viewed themselves as somewhat better informed than experts, a number that dropped precipitously as knowledge about autism increases. Other interesting findings related to this part of the study were that religiosity was associated with increased overconfidence. The authors note that this finding is consistent with previous studies showing that more religious people are more likely to hold anti-expert attitudes. Oddly enough, the authors also found that wealthier and older individuals were less likely to exhibit overconfidence in their models. This latter result surprised me a bit, because it seems to go against my anecdotal experience.

Another interesting finding is that overconfidence didn’t seem to affect attitudes towards the role that experts play in formulating vaccine policy, with no statistically significant effects of overconfidence on support for the role that experts play in vaccine policymaking. I wouldn’t have guessed that. I would, however, have guessed this result:

However, we did find positive and statistically significant effects of overconfidence on support for the role that non-experts play in constructing public policy about vaccines in both the knowledge and misinformation models. In the former, moving from the minimum to maximum observed values of overconfidence translates to an 18% increase in the likelihood of respondents thinking that non-experts should play a “major role” in policy-making decisions about vaccination; from 32% at low levels of overconfidence to 50% at high levels. We found a similar pattern of effects in the misinformation model, with overconfidence shifting support from 31% to 53%.

And:

We uncovered a strikingly similar pattern in our final set of analyses, focused on who citizens trusted to provide them with information about vaccines. We tested this aspect of Hypothesis 2b in Columns 7–10. Again, we found no significant effects of overconfidence on trust in information provided by medical professionals (in fact, the parameter estimates were nearly zero in both cases). However, overconfidence was associated with large increases in trust in non-experts. In the knowledge model, overconfidence boosted trust by 21%, from a predicted trust level of 36% at low levels of overconfidence to 57% at high levels of overconfidence. We found an even stronger pattern of results in the misinformation model, with overconfidence associated with a 26% increase in trust (from 35% to 61%, respectively). Again, both models reveal strong evidence of partial mediation.

Thinking about this more, I’m starting to see the sense of it. If this study reflects what is really going on, overconfidence in one’s own knowledge engendered by the Dunning-Kruger effect doesn’t necessarily have to mean that one discounts the expertise of real experts. It just means that one now considers oneself as knowledgeable (or nearly as knowledgeable) as an expert and thereby provides the confidence (or, as I like to call it, the arrogance or overconfidence) to contradict what experts say about the science showing no correlation between vaccines and autism. It also makes sense that such a willingness to challenge experts, even if the person doing the challenging doesn’t know what he’s talking about, would correlate with more trust in the knowledge of non-experts.

Of course, as with all studies of this type, this study has significant weaknesses. For one thing, it’s a single cross-sectional study. That means its one measure at one point in time of one population. It could be a fluke. More importantly, it doesn’t really shed much light on the origins of the misinformation about autism that leads to antivaccine views.

But what does this all mean? First, I was surprised when the authors pointed out that their analysis is the first to find evidence that people’s lack of awareness of what they don’t know about autism (i.e., the Dunning-Kruger effect) and their acceptance of misinformation regarding vaccines and autism can “alter how citizens’ view their own knowledge, compared to that of medical experts.” That seems like such a potentially obvious conclusion that I’m surprised there isn’t more evidence for it.The authors also note that their results “indicate that overconfident individuals in our sample are less supportive of mandatory vaccination policy and tend to elevate the role that non-experts should play in the policymaking process.” That’s a conclusion that I had made myself based on anecdotal experience dealing with antivaxers, but it’s nice to see some evidence in favor of it.

Finally, in an op-ed, the authors note that their results point to an “uphill battle” being faced by the scientific community. (Tell me something I didn’t already know!) They note that, even though the “mountain of evidence” exist supporting the safety and efficacy of vaccines, many Americans think they know more than the experts trying to correct the misinformation. Unfortunately, these attitudes seem to be prevalent about more than just vaccines and autism. Trust in nonexperts and distrust of experts seems to be a feature of the age we are living in, fed by the ease with which everyone can be an expert on social media and the proliferation of “fake news.” Countering the assaults on science and reason resulting from humans in the thrall of the Dunning-Kruger effect is the central challenge of this age.

Bertrand Russell on the Dunning-Kruger effect