Exploring the Dunning-Kruger Effect in Health Communication: How Perceived Food and Media Literacy and Actual Knowledge Drift Apart When Evaluating Misleading Food Advertising

Marketers often advertise products high in sugar, fat or calories as healthy products. With this potentially misleading information, they can influence eating decisions with negative consequences for human health. Consumers need the ability to uncover misleading food advertising. However, individuals’ perceived knowledge and their actual objective abilities often drift apart – a phenomenon which has come to be known as the Dunning-Kruger Effect. Thus, this study set out to uncover the phenomenon’s potential existence in health communication, more precisely in the area of food and media literacy. In a quantitative survey representative of the Austrian population (n = 1000) the Dunning-Kruger Effect (DKE) could be detected: Individuals who were most knowledgeable underestimated their food and media literacy, but – on the positive side – they acted as opinion leaders. Individuals who were least knowledgeable about advertising strategies used to market an unhealthy product as healthy and about the actual nutrition score of the advertised product were most likely to overestimate their own food and media literacy. Worryingly, further concerning consequences emerged, especially for least knowledgeable individuals. The study’s results provide important implications for public health campaigns.

and can also include food and health choices (Truman et al., 2017). 1  Previous findings support the relevance of media literacy, for example, in promoting engagement in protective health behavior during the COVID-19 pandemic (Austin, Austin, Willoughby, Amram, & Domgaard, 2021) and fact-checking behavior in relation to health information (Lee & Ramazan, 2021).Austin, Austin, French, and Cohen (2018) highlight the importance of also strengthening individuals' media literacy in the nutrition context to mitigate the negative effects of unrealistic food marketing.
However, there is a risk that media and food literacy could be seriously compromised.Kruger and Dunning (1999) discovered that those people who, in fact, display the lowest knowledge are those who overestimate their actual abilities most.In contrast, people showing actual high knowledge often underestimate their abilities, which points to a mismatch between actual and perceived abilities -a phenomenon known as the Dunning-Kruger Effect (DKE).People's confidence in their own knowledge can become a problem, if their knowledge is incorrect (Freiling, 2019) and interventions to foster media literacy may be ineffective among those individuals who would need training in media literacy most (Gross & Latham, 2012).Individuals with the lowest actual skills are unlikely to seek help for skills which they believe they have, since they are highly confident about their skills (Gross & Latham, 2012;Sivakumar, Hanoch, Barnes, & Federman, 2016).
To the best of the authors' knowledge research on the DKE's presence in health communication and in particular among food and media literacy is still missing.Thus, the present study aims at empirically detecting the DKE's existence in health communication in the context of misleading advertising on a quantitative and largescale level.We use a representative sample (n = 1000) for the Austrian population to explore its potential threatening effect on people's food and media literacy in the context of evaluating food advertising critically to identify misleading information here.Our study also follows previous suggestions from the field of information literacy (Gross & Latham, 2012) to investigate flawed selfassessments among different sub-populations and, more precisely, for the population at large -here for the Austrian population -thereby moving beyond a student sample.The present study emphasizes the urgency of dealing with misleading information distributed by food ads and, simultaneously, provides important implications for public health campaigns and policy makers to counteract the DKE's effect.The central aim of this study is thus to answer the following research question: Does the DKE exist in the context of misleading food advertisements and is it relevant to health communication due to affecting users' self-perceptions of (a) being able to report the actual health score of food portrayed in ads (food literacy) and (b) being able to detect advertising strategies used to deceive them (media literacy)?
Innovative aspects of our study are that both health-related food and media literacy are analyzed simultaneously as both are needed to detect misleading advertising.Moreover, our study analyzes several potential consequences related to the DKE.

Previous Research on the DKE
Since Kruger and Dunning's (1999) publication, the DKE has been discovered in various fields, for example, in sports coaching (Sullivan, Ragogna, & Dithurbide, 2019), in the area of egalitarianism (West & Eaton, 2019), wine knowledge (Aqueveque, 2018), or information literacy (Gross & Latham, 2012).While individuals having highest knowledge underestimated their abilities, individuals having lowest knowledge in a domain overestimated their abilities.Especially in the field of information literacy, this effect seems to be highly problematic since acquiring information literacy skills will become especially important in a society where more and more information is available via different types of media (Gross & Latham, 2012).Motta, Callaghan, and Sylvester. (2018) highlighted the DKE's consequences regarding vaccine policies.Individuals with the least knowledge about autism and its potential link to vaccines, regarded themselves as even more knowledgeable than medical experts and were also less convinced by mandatory implementations of vaccination policies.The findings on the DKE of Motta et al. (2018) in the health area suggest potential negative effects on people's health, highlighting the urgency of dealing with the DKE.Recent research (Choi, Northup, & Reid, 2021;Sivakumar et al., 2016) indicates the importance of distinguishing between actual and subjective expertise in the area of health literacy as well.Compared to those showing lower health literacy (measured as actual knowledge), more highly health literate people were more likely to understand nutrition information.Being more health conscious (subjective expertise) did not guarantee higher actual knowledge about food ads advertising unhealthy food (Choi et al., 2021).As recommended (Choi et al., 2021), we also extend these findings using a different claim ("Protein and Lower Carb") and food item (chocolate cereal bar) in the misleading ad stimulus applied in this study.Similarly, Sivakumar et al. (2016) also found that individuals with lower levels of health literacy rated their Medicare knowledge as sufficient, when, in fact, their actual Medicare knowledge was low.Neither study (Choi et al., 2021;Sivakumar et al., 2016) analyzed the DKE explicitly, nor did they analyze food and media literacy.Thus, transferring the DKE to health communication, we hypothesize: 1 Our focus is on media and food literacy.Information and health literacy are related concepts.To avoid confusion, we distinguish these two concepts as well.
Media literacy also shares similarities with information literacy, though some differences remain.The focus of information literacy is on the access to texts/information and barriers/enablers in this regard, whereas media literacy research focuses more on the critical analysis of information (Livingstone, Van Couvering, & Thumim, 2008), therefore we focus on media literacy.A related concept to media literacy and food literacy is health literacy.Health literacy has been defined as: "The degree to which individuals can obtain, process, understand, and communicate about healthrelated information needed to make informed health decisions" (Berkman, Davis, & McCormack, 2010, p. 16).In this vein, media literacy skills are also especially important when acquiring information related to health and for rejecting misinformation in this regard (Austin et al., 2021).Food literacy is more specific and more closely related to the evaluation of the information provided in the food advertisement than health literacy, which is why we use food literacy.
H1: Individuals who are least knowledgeable about (a) the actual nutrition score of food products and (b) ad strategies used in food ads, will overestimate their (a) food and (b) media literacy.
H2: Individuals who are most knowledgeable about (a) the actual nutrition score of food products and (b) ad strategies used in food ads, will underestimate both their (a) food and (b) media literacy.

The DKE's Potential Consequences
Our study extends previous research by also exploring the DKE's potential consequences on important advertising evaluation variables like skepticism toward advertising, ad evaluation, and purchase intention as well as individual variables like involvement with a healthy diet, opinion leadership, and social media usage.

Advertising Evaluation Variables and Involvement with a Healthy Diet
In this study, we refer to skepticism toward advertising in general as "the tendency toward disbelief of advertising claims" (Obermiller & Spangenberg, 1998, p. 160).2Actual knowledge about nutrition might be linked to consumers' skepticism toward ads.Consumers scoring high on nutrition knowledge were more skeptical toward health claims.Those consumers who showed higher levels of skepticism toward health claims were also less likely to rely on health claims when deciding which products to buy (Tan & Tan, 2007).By contrast, people who indicated to have less knowledge about nutrition were less skeptical of health claims (Mitra, Hastak, Ringold, & Levy, 2019).However, people with low levels of skepticism might not evaluate ads more favorably (Koinig, Diehl, & Mueller, 2018).We also want to investigate, if people with lowest knowledge about nutrition and persuasive ad strategies evaluate/like the misleading ad differently compared to individuals with highest knowledge.People with lowest knowledge might also have a higher purchase intention toward the advertised product.Choi et al. (2021) showed that individuals with lower health literacy were more likely to purchase unhealthy food advertised with a nutrient-content claim for lower fat compared to individuals with higher health literacy.Another relevant construct is involvement.Highly involved individuals ascribe special importance to certain objects (Zaichkowsky, 1985), based on personal needs, values and interests.3+4 Compared to those individuals with lowest involvement with a healthy diet, individuals showing highest involvement with a healthy diet might also show higher knowledge about the actual nutrition score of products and ad strategies used in food ads, given that higher involvement with products also contributes to higher product knowledge (Liang, 2012).Thus, individuals who ascribe special relevance to a healthy diet (show higher involvement with a healthy diet) might have higher nutrition knowledge, compared to less involved individuals, who are less interested in a healthy diet.Given these initial considerations, we aim to answer the following questions to uncover potential consequences of the DKE: RQ1: Do those individuals who are least knowledgeable about the actual nutrition score of products and ad strategies used in food ads have (a) a lower skepticism toward advertising, (b) a lower involvement with a healthy diet, (c) a more favorable ad evaluation (ad liking), and (d) a higher purchase intention for the advertised product than those who are most knowledgeable?

Opinion Leadership and Social Media Usage
People have the ability to influence other people's attitudes and behavioral intentions (Ho et al., 2020;Jung, Shim, Jin, & Khang, 2016;Onofrei, Filieri, & Kennedy, 2022).Following Rodgers and Cartano (1962), there are individuals "who exert an unequal amount of influence on the decision of others" whom they call "opinion leaders" (p.435).Insights from Social Cognitive Theory (Bandura, 1999(Bandura, , 2001) ) suggest a bidirectional relationship between personal, environmental and behavioral factors determining each other.Human learning is influenced, either consciously or unconsciously, by observing the behavior of other people, and this is also true for electronic mass media (Bandura, 1999).People imitate the behavior of opinion leaders (Kim & Dearing, 2016).Also, in the health sector, followers value opinion leaders' views on health issues.Opinion leaders are gaining importance, especially given today's health-related information overload caused by a plethora of communication channels (Kim & Dearing, 2016).As individuals often place high trust in them and their knowledge, opinion leaders can either protect others from disinformation or aggravate disinformation effects (Dubois, Minaeian, Paquet-Labelle, & Beaudry, 2020).Consequently, there might be a serious problem if those people are opinion leaders who know least about the actual healthiness of advertised foods (regarding e.g., sugar/fat content) and deceptive advertising practices in food advertising.They may share their incorrect knowledge with the people placing high trust in them and their expertise in a misleading manner, thus reinforcing the DKE.Moreover, if low-knowledge individuals are also active social media users, they may spread their flawed knowledge further on social media, as active social media users tend to share or post contents on social networking sites (Li, 2016), causing concerns about the quality of health information (Zhao & Zhang, 2017).This would mean that less knowledgeable individuals could pose a particularly high risk to other individuals, as there might be the danger that they reinforce the DKE.Thus, we also want to answer the following questions: RQ2: Are those individuals who are least knowledgeable about the actual nutrition score of products and ad strategies used in food ads more likely to be opinion leaders than individuals with the highest knowledge?RQ3: Are those individuals who are least knowledgeable about the actual nutrition score of products and ad strategies more active or passive social media users than individuals with the highest knowledge?

Method
To investigate the hypotheses and research questions, an online survey was conducted with a representative sample for the Austrian population.The survey was conducted with the support of the Gallup International Association (https://www.gallup.at/de/home/).

Applied Stimulus
For the present study, a fictitious ad was developed, based on which participants were invited to answer questions (stimulus available upon request).This ad -its design and the depicted product -was based on an existing food ad that has previously been criticized by the consumer rights organization Foodwatch (Foodwatch, n.d.).Modifying real-world food ads for research purposes has also been practiced previously to enhance external validity (e.g., Choi, Paek, & Whitehill King, 2012;Choi & Reid, 2015).The stimulus depicted a chocolate cereal bar containing a lot of sugar and fat that is misleadingly advertised as being healthy by using the textual claims "Protein" and "Lower Carb" and the visual element of a slim blue silhouette in the background.

Procedure, Sample and Measurement
A qualitative pretest with n = 6 participants (66.6% female; age: M = 42.00years, SD = 11.92)ensured the realistic design of the fictitious stimulus and the questionnaire's comprehensibility.We followed a collaborative and iterative procedure for translation based on Douglas and Craig (2007) as most of the scales used were translated from English into German.To counteract common method bias we followed suggestions by MacKenzie and Podsakoff (2012).
For the main study, participants were recruited via the Gallup's ISO certified online panel (Gallup Institut, 2022, p. 3).In total, n = 1000 people from the Austrian population, which is a representative sample, participated in the study (51.1% female, 49,7% male, 0.2% diverse; age: M = 47.98 years, SD = 17.25).Table A1 shows an overview of participants' sociodemographic characteristics.Participants received credit points as compensation.
All measures and items used in the questionnaire (including reliability estimates) are shown in Table A2.Unless otherwise indicated, all scales are seven-point Likert scales with the endpoints "strongly disagree" to "strongly agree."At the beginning of the questionnaire, demographic data were collected, participants were thanked for their participation and informed about the aim of the study, the procedure, and the importance of their participation, and they were assured of the anonymous data collection.They were told that the purpose of the survey is to evaluate the effects of advertisements for food products.Following previous research (e.g., McMahon, Stoll, & Linthicum, 2020;Sullivan et al., 2019;West & Eaton, 2019), this study applied self-assessment measures (perceived food/media literacy) and an objective knowledge test to detect the DKE and to uncover whether people's actual knowledge about food products and applied advertising strategies corresponded to their self-views.Following the self-assessment measures, participants were presented with the fictitious advertising stimulus which served as a basis for the knowledge test (announced as quiz to participants).Table A3 presents the objective knowledge test.Participants were thanked again for their participation and the survey ended with a debriefing.Data were analyzed with IBM SPSS Version 28.Confirmatory factor analyses were performed using IBM SPSS AMOS Version 28.

Results
The objective knowledge test included questions measuring people's actual food and media knowledge.Therefore also the selfassessment measures (food literacy and media literacy scales) were combined to form one variable.To correctly evaluate a food ad, both food and media literacy are needed simultaneously.A sum score was calculated for the 19 food and media literacy scale items (13 items measuring subjective food literacy and 6 items measuring subjective media literacy).Given that "7" was the highest possible score participants could indicate on the food and media literacy scales, in total, participants could reach 133 (19 multiplied by 7) points displaying their own perceived food and media literacy.Respondents showed a mean self-assessment score of 88.62 (SD = 19.40)for food and media literacy out of 133 points.In the objective knowledge test, their scores ranged from 4 to 18 (M = 10.36,SD = 2.43), out of a total of 21 achievable points.
In our research we followed a similar approach for data analysis to detect the DKE as researchers have used previously.Based on the participants' performance in the objective test, they are grouped into quartiles (e.g., McMahon et al., 2020;Sullivan et al., 2019;West & Eaton, 2019).In our study, we conducted a quartiles analysis with SPSS 28, which resulted in the groups as reported in Table 1.
Paired sample t-tests were performed in the groups to check for the difference in mean scores between actual ability (food and media knowledge) and self-assessment (food and media literacy).Actual ability and self-assessment were measured using different scales.Hence, Z scores were calculated to interpret the results.Our results revealed that the "lowest-knowledge" group showed significantly higher self-assessment scores (food and media literacy) (M z-score =-.20, SD = 1.09) than actual ability (food and media knowledge) (M z-score =-1.32,SD = 0.41) (t (231) =-14.89,p < .001,d=-.98), thus supporting H1(a) and H1(b) 5 .Also, in the "low-knowledge" group, participants' self-assessment scores (M z-score =-.14, SD = 1.02) were significantly higher than their actual ability scores (M z-score =-.56, SD=.00) (t (138) =-4.81, p < .001,d=-.41).In the "highknowledge" group no significant difference emerged between selfassessment (M z-score =.01, SD=.98) and actual ability (M z-score =.04, 5 Findings are based on the combined scores of food and media literacy (self-assessment) as well as food and media knowledge (actual ability).

710
R. Scheiber et al.SD=.21) (t (300) =.66, p = .51,d=.04).In contrast, the "highestknowledge" group's self-assessment scores (M z-score =.20, SD=.91) were significantly lower than their actual ability (M z-score = 1.13,SD=.50) (t (327) = 16.50, p < .001,d=.91), thus also supporting H2 (a) and H2(b) 5 .Figure 1 shows a summary of the results, which confirm the existence of the DKE for food and media literacy.To support our findings, we also analyzed the DKE separately for food literacy/food knowledge and media literacy/media knowledge.We could confirm the DKE separately for food literacy/food knowledge and media literacy/media knowledge, further supporting H1(a+b) and H2(a+b). 6 Next, one-way ANOVAs were conducted to address RQs 1-3.For the ANOVAs, we defined 1) the quartiles7 as independent variable and 2) the variables: skepticism toward advertising, involvement with a healthy diet, ad evaluation (ad liking), purchase intention, opinion leadership subscales connectivity/ persuasiveness/mavenness, and active and passive media usage as dependent variables.Comparisons were made among all four knowledge groups.Table 2 presents the mean values and standard deviations for the variables across groups.Post-hoc comparisons were consulted and the LSD test was interpreted for all variables.Table A4 depicts the post-hoc test results and significant/non-significant differences related to all groups.Table A5 shows the results of a correlation analysis between all variables.
To answer RQ1, focusing on ad (effectiveness)-related variables and involvement, the results of the ANOVAs indicated that there are significant differences between the groups for skepticism toward advertising, F(3, 995) = 4.46, p=.004; purchase intention, F(3, 995) = 3.99, p=.008; and involvement with a healthy diet, Welch's8 F(3, 448.82) = 7.89, p < .001.In the mean comparison tests, no significant differences emerged between the groups regarding ad evaluation, F(3, 995)=.72,p=.539.Post-hoc comparisons (LSD test) showed that the "lowest-knowledge" group scored significantly lower on skepticism toward advertising (M = 4.47, SD = 1.42) compared to the "highest-knowledge" group (M = 4.89, SD = 1.32, p < .001)indicating that the "lowestknowledge" group is less skeptical (RQ1a).Furthermore, the "lowest-knowledge" group scored significantly lower on involvement with a healthy diet (M = 4.88, SD = 1.54) compared to the "highest-knowledge" group (M = 5.43, SD = 1.19, p < .001),indicating that the "lowest-knowledge" group has a lower involvement (RQ1b).The "lowestknowledge" group scored significantly higher on purchase For self-assessment respondents could reach a maximum score of 133 points (highest possible sum score of the food and media literacy scales).21 points could be reached in total for actual ability (scores on the objective knowledge test on food and media knowledge).For both, self-assessment and actual ability, higher scores indicated more of the construct.Respondents reaching zero to eight points were grouped into "lowest knowledge" (n = 232), nine points into "low knowledge" (n = 139), ten to eleven points into "high knowledge" (n = 301), and twelve and more points into "highest knowledge" (n = 328). 6 To provide further support for our findings, we also analyzed the DKE separately for food literacy (self-assessment)/food knowledge (actual ability) and media literacy (self-assessment)/media knowledge (actual ability).We were able to confirm the DKE separately for food literacy/food knowledge and media literacy/media knowledge, further supporting H1(a+b) and H2(a+b).In the objective knowledge test (measuring actual abilities), twelve questions tested 1) food knowledge and nine questions measured 2) media knowledge.Hence, for 1) food knowledge, participants could reach a maximum of twelve points and for 2) media knowledge nine points.We again generated quartiles and grouped participants into quartiles based on their scores reached for each of the two variables (food/media knowledge) separately in SPSS 28 (see Table A6 for an overview of the groups).Paired samples' ttests revealed a similar picture.The "lowest-knowledge" group showed significantly higher self-assessment scores (food literacy) (M z-score =-.03, SD=1.08) than actual ability (food knowledge) (M z-score =-1.39,SD=.35) (t (180) =-16.83, p<.001, d=-1.25),thus providing support to H1(a).The same picture could be found for media literacy.The "lowest-knowledge" group's self-assessment scores (media literacy) (M z- score =-.26, SD=.96) were significantly higher than their actual ability (media knowledge) (M z-score =-1.71,SD=.34)(t (104) =-14.50, p<.001, d=-1.41),supporting H1(b).In contrast, yet as expected, the "highestknowledge" group's self-assessment scores (food literacy) (M z- score =.08, SD=.92) were significantly lower than their actual ability (food knowledge) (M z-score =1.02, SD=.57) (t (397) =17.05, p<.001, d=.86), thus also supporting H2(a).Their self-assessment scores (media literacy) (M z-score =.19, SD=.97) were also significantly lower than their actual ability (media knowledge) (M z-score =.95, SD=.58) (t (423) =14.46, p<.001, d=.70), again lending support to H2(b). Figure A1 shows a summary of the results.intention (M = 3.38, SD = 1.75) compared to the "highestknowledge" group (M = 2.89, SD = 1.58, p < .001),suggesting that the "lowest-knowledge" group is more likely to purchase the product (RQ1d).For ad evaluation, no significant differences emerged between the "lowest-knowledge" group and the "highest-knowledge" group (RQ1c).
Moreover, we wanted to analyze if the members of the "lowest-knowledge" group are opinion leaders (RQ2).The ANOVA's results showed significant differences between the four knowledge groups for the subscale persuasiveness, F(3, 995) = 3.74, p = .011,but no significant differences for the subscales connectivity, F(3, 995) = 1.66, p = .17,and mavenness, F (3, 995) = 2.21, p = .09.Post-hoc comparisons (LSD test) showed that for persuasiveness, the mean scores for the "lowest-knowledge" group were significantly lower (M = 4.38, SD = 1.15) compared to the "highest-knowledge" group (M = 4.66, SD = 1.12, p = .005).For mavenness, the "lowest-knowledge" group's scores were also significantly lower (M = 3.93, SD = 1.37) compared to the "highest-knowledge" group (M = 4.19, SD = 1.24, p = .023).For connectivity, no significant differences emerged between all four groups.Interestingly, those who show the lowest knowledge are not more likely to be opinion leaders.It seems that those who are highly knowledgeable are more likely to persuade others and be consulted by others for information.Thus, individuals who are least knowledgeable about nutrition and ad strategies in food ads are not opinion leaders (RQ2).
Finally, when exploring RQ3 relating to media usage, in the mean comparison tests, no significant differences emerged between the groups regarding active media usage, F(3, 995) = 1.83, p = .140and passive media usage, F(3, 995)=.73,p = .533.Post-hoc comparisons (LSD test) revealed that the "lowest-knowledge" group scored significantly higher on active media usage (M = 2.86, SD = 1.59) compared to the "highest-knowledge" group (M = 2.60, SD = 1.47, p = .045),suggesting a more active use of social media in the "lowest-knowledge" group compared to the "highest-knowledge" group.For passive media usage, no significant differences emerged between the "lowest-knowledge" group and the "highestknowledge" group (RQ3).

Discussion and Implications
This study set out to explore whether food and media literacy might be affected by the Dunning-Kruger effect (Kruger & Dunning, 1999), when evaluating potentially misleading food advertisements.The study's results show the presence of the DKE in the area of food and media literacy.Those individuals who were least knowledgeable about the potentially misleading advertising strategies used to market an unhealthy product as healthy and the actual nutrition score of an advertised cereal bar were most likely to overestimate their own food and media literacy.In contrast, those individuals who were most knowledgeable underestimated their food and media literacy.These results are in line with previous research that has detected this effect across a variety of domains (e.g., Aqueveque, 2018;Gross & Latham, 2012;Motta et al., 2018;Sullivan et al., 2019;West & Eaton, 2019), thereby extending previous As an important theoretical contribution, we also support the previously stated concern (Choi et al., 2021) that there is a mismatch between subjective and objective expertise when evaluating misleading food ads -in the present study on a large-scale level and by investigating reactions to a different product (chocolate cereal bar) and a different nutrient-content claim (Protein/Lower Carb).Our findings also provide important theoretical contributions in light of the "health halo effect" (Choi, Yoo, Hyun Baek, Reid, & Macias, 2013;Choi et al., 2021), where recipients maintain cognitive consistency of the positive aspects emphasized in health and nutrition claims and are consequently more likely to overlook the unhealthy components in the advertised product.Our study suggests that the health halo effect might be less present among individuals who have highest foodand media-related knowledge, since they are less likely to buy the advertised product compared to those with lowest knowledge.
In line with the study by Heiss, Naderer, and Matthes (2021), our study suggests that people face difficulties in identifying healthwashed and potentially misleading ads.We outline an important additional aspect that deserves further consideration.Participants who were least knowledgeable about misleading food ads overestimated their competences in relation to food and media literacy and were actually not aware of their inability.According to Kruger and Dunning (1999), those people who lack knowledge "suffer a dual burden."Adding to their problem of drawing incorrect conclusions about their own abilities, their limited knowledge in a domain prevents them from realizing their own incompetence (Kruger & Dunning, 1999, p. 1132).The latter aspect bears important implications especially for health communication efforts.Boosting people's knowledge about misleading food ads and the DKE is crucial.Campaigns should create awareness not only of deceptive ads but also of the DKE's existence.People should be made aware of the phenomenon and get informed that their selfassessment might be flawed to make them subsequently more attentive and receptive to validated knowledge communicated by, for example, health authorities.Such interventions, to help people with lowest knowledge levels, are crucial, since we were able to show that consumers with the lowest abilities but with high confidence in their abilities are most vulnerable regarding potentially misleading food ads, as they are less skeptical and more willing to buy the product without realizing the potentially negative effects of the product.However, health authorities should use complementary strategies in addition to increasing knowledge (e.g., develop tools assisting individuals in detecting trustworthy information sources) (Austin, Austin, Borah, Domgaard, & McPherson, 2023), given that high knowledge often does not automatically translate into desired nutrition behavior (Larsen, 2015;O'Brien & Davies, 2007). 9 Additionally, food has recently been described as "a protagonist of the posting activity on social networks" (Ventura, 9 Based on our study's findings, the "highest-knowledge" group scored significantly lower on purchase intention (M=2.89,SD=1.58) compared to the "lowest-knowledge" group (M=3.38,SD=1.75,p<.001).This lower purchase intention suggests that the "highest knowledge group" is less likely to purchase an unhealthy advertised product thus benefitting their health.Thus, higher knowledge might help to protect against purchasing unhealthy food products.However, of course, in the present study we did not assess actual behavior, but behavioral intentions.Most of the measures we used are self-assessments (with the exception of objective knowledge), which are, of course, essential for determining the DKE, as the focus is on flawed self-assessments.We measured behavioral intentions, yet future research and in particular longitudinal studies could also investigate actual behavior.
The Dunning-Kruger Effect in Health Communication Cavaliere, & Iannò, 2021, p. 674).Our study's results showed that especially low-knowledge individuals are more likely to be active users of social networking sites.This finding may indicate that they could spread their incorrect knowledge more easily via social media channels.These results deliver cause for concern because they suggest that the DKE may be reinforced by the least knowledgeable group, but the results regarding opinion leaders paint a different, more optimistic picture.
Interestingly, our results suggest that it is not individuals with the lowest knowledge that are more likely to be opinion leaders but those with the highest knowledge.People who are most knowledgeable about food ingredients and misleading advertising tactics are most likely to persuade and be consulted by others for information.This finding aligns with Social Cognitive Theory (Bandura, 1999(Bandura, , 2001)), suggesting that people use opinion leaders and people from their surroundings to acquire information.Our result is promising as it suggests that high-knowledge individuals can pass on their correct knowledge to others and thus positively influence other people's knowledge.If low-knowledge individuals are reluctant to change their self-views, it may be opinion leaders who have the power to persuade them to change their selfperceptions and abilities and assist them to use media more effectively.Public health campaigns, which should use social media channels, could then recruit opinion leaders who create awareness for the DKE's existence among low-knowledge individuals and simultaneously provide them with correct knowledge about misleading food advertising tactics, with positive health effects.Social media influencers and campaigns appear to be especially promising as low-knowledge individuals use them a lot.As low-knowledge individuals also have a lower involvement toward a healthy diet, these campaigns could try to increase their involvement and thus interest in nutrition-related health information.Targeted marketing efforts might provide an efficient means to select opinion leaders who enjoy trust among different age groups.
Our results also lead to important implications for policy makers seeking to enforce stricter governmental regulations.Our study results highlight the importance of enforcing stricter laws on advertising unhealthy food items, since consumers' self-assessment to detect misleading food ads are flawed and inaccurate.We thus suggest, related to regulatory implications, to make the Nutri-Score nutrition label mandatory in Austria and other countries, as it enables consumers to more easily identify the actual nutritional quality of food products (Egnell et al., 2020).Also, the IARC has called for the Nutri-Score's wide-spread use in Europe and beyond (IARC, 2021).In Austria, companies that would be willing to use the Nutri-Score on their products voluntarily are, by law, not even allowed to do so (Foodwatch, 2020).Our study's results emphasize the need that companies use the Nutri-Score as it also provides marketers with advantages.For example, purchase intentions for healthy products can be increased and even if unhealthy products display the Nutri-Score, marketers can show consumers that they want to be transparent and care about their health (De Temmerman, Heeremans, Slabbinck, & Vermeir, 2021).To protect especially individuals with low knowledge about food and media tactics, marketers should urgently consider using the Nutri-Score on their product packaging.

Limitations and Future Research
Moving beyond our contribution, we want to mention some limitations of our study, which provide directions for future research.In a first step, we could detect the DKE across food and media literacy in the context of misleading food ads, suggesting the importance of considering the DKE also in health communication research.Future research might want to test the relationships between these variables in the context of food and media literacy using structural equation modeling (SEM).This analysis replicated the traditional DKE analysis strategy of MANOVAs on samples split into quartiles.It was not an experiment and used a single message, however, limiting its ability to detect effects.Regression/SEM analysis could uncover associations of overconfidence with outcomes such as social media use.To test the robustness of our findings, experimental studies could be conducted that manipulate certain message characteristics in the food ad (e.g., colors, high vs. low information, ad complexity).Follow-up studies could also test how further individual difference variables, not investigated in this study, might have affected the DKE (e.g., gender, age, health status, product affinity, attitude toward advertising in general).Future research might also conduct qualitative interviews to uncover the DKE using additional research methods or longitudinal studies, as our study is a cross-sectional study.
We analyzed the impact of the DKE on several variables.In this context, additional variables might play a role, e.g., Austin et al. (2021) found that, next to knowledge, variables like people's expectancies of the results when performing healthrelated behaviors, willingness to perform the behaviors and their self-efficacy are important precursors of health-related behaviors. 10Future research is needed to investigate and test these additional variables as well.
We focused on the affective elements of skepticism, but there are also logic-based elements (Austin et al., 2002) and skepticism has also been considered as an outcome of critical thinking (Austin, Muldrow, & Austin, 2016).Future research may want to investigate these important aspects of skepticism.The same applies to our measurement of involvement.We focused on the cognitive component.Given that involvement also has an affective component (Zaichkowsky, 1994), it would be important to investigate if and how affective involvement with a healthy diet is related to actual knowledge about food and advertising strategies.Finally, it would be interesting to extend our research to other countries and product categories.10 As suggested by previous studies (e.g., Larsen, 2015;O'Brien & Davies, 2007), there is a distinction between knowledge/actual ability and willingness to apply the knowledge.Future research might investigate factors that intervene between knowing that something/an action would be healthier and factors such as social mechanisms and societal surroundings (Larsen, 2015), fear of missing out, convenience, or fatigue which can prevent people from taking the healthier route.

714
R. Scheiber et al.For each correct answer participants could reach one point while zero points were given for incorrect answers.A total score of twenty-one points could be achieved.For actual food knowledge: The use of these six ingredients (questions 1-6) was based on the actual product information of an existing cereal bar that has been criticized by Foodwatch (n.d.).Correct answers (questions 7-12) were based on findings provided by research (Schulze-Lohmann, 2012), the German Association for Nutrition (Deutsche Gesellschaft für Ernährung, n.d.) and on the critique by Foodwatch which classified certain amounts of the product's ingredients as either high or low.For media knowledge (actual ability): A team of four experts from the media and health communication field defined which response options transport healthiness and which ones don't.Correct answers were also based on the critique published by Foodwatch's food experts.Prior research followed a similar procedure: Lyons, Montgomery, Guess, Nyhan, and Reifler (2021) also consulted experts (e.g., fact-checking organization) to ensure the correctness of answers in the knowledge test.In our research, we also used scientific findings and the expertise of Foodwatch and experts from the media and health communication field to ensure the correctness and reliability of our test and answers.Our objective knowledge test was based on an existing (misleading) food ad.Similarly, prior research (e.g., by Lyons et al., 2021) also used real-world examples (news articles) as basis to measure people's actual ability to detect false news content.In line with most prior studies on the DKE (Aqueveque, 2018;McMahon et al., 2020;Sullivan et al., 2019), we used multiple-choice and yes/no response options in our objective knowledge test.Here, research has also started to use different scales to measure actual ability, which might be an alternative for future studies as well (see Lyons et al., 2021).
722 R. Scheiber et al.For self-assessment respondents could reach a maximum sum score of 1) 91 points for food literacy − 13 items measuring food literacy -highest response option "7" 2) 42 points for media literacy − 6 items measuring media literacy -highest response option "7."For actual ability, a maximum score of 1) 12 points could be reached for food knowledge 2) 9 points for media knowledge in total (scores on the objective knowledge test).For both, self-assessment and ability, higher scores indicated more of the construct.
The Dunning-Kruger Effect in Health Communication

Figure 1 .
Figure 1.Differences between actual ability (food & media knowledge) and self-assessment (food & media literacy) per quartile.

Table 1 .
Means and standard deviations for each group's actual ability (=food and media knowledge) and self-assessment (=food and media literacy)

Table 2 .
Means and standard deviations across groups findings and identifying the effect's presence among food and media literacy in a health communication context.

Table A2 .
Applied measures -items with means, standard deviations; reliability estimates: confirmatory factor analyses results, cronbach alpha

Table A3 .
Objective knowledge test with correct answers (marked in bold) you can see different graphical elements and texts.Some of these elements in this ad aim at conveying healthiness.Please click on all those elements from the list below, where you think that they try to convey healthiness.13.Chocolate combined with cereals as a high-quality snack Twelve questions (1-12) measured food knowledge (actual ability), nine questions (13-21) measured media knowledge (actual ability).

Table A4 .
Results of the Post-hoc tests: for all four groups

Table A5 .
Descriptive statistics and correlations for study variables

Table A6 .
Number of participants per group, grouping of participants based on scores on objective test, means and standard deviations for each group's: actual ability (=food knowledge) and self-assessment (=food literacy); actual ability (media knowledge) and self-assessment (=media literacy)