Improving vaccine attitudes and misinformation resistance through gamification: a pilot study in Kenya and Uganda

¹Melbourne Centre for Climate Change, University of Melbourne, Melbourne, Australia, ²Kenya Medical Research Institute, Nairobi, Kenya, ³Department of Chemistry, University of Nairobi, Nairobi, Kenya, ⁴UNICEF, New York, USA, ⁵Department of Health Policy Planning and Management, Makerere University School of Public Health, Kampala, Uganda, ⁶Population Health and Environment Department, Amref Health Africa, Nairobi, Kenya, ⁷Irimi, Lyon, France, ⁸Department of Communication Studies & Global Health Communication Center, Indiana University School of Liberal Arts at IUPUI, Indianapolis, USA, ⁹Wendy Cook Design, Melbourne, Australia, ¹⁰Sabin Vaccine Institute, Washington DC, USA

^&Corresponding author
John Cook, Melbourne Centre for Climate Change, University of Melbourne, Melbourne, Australia

Introduction    

Misinformation about vaccines can negatively impact vaccine acceptance [1], perceived threat of vaccine-preventable diseases [2], and vaccination uptake [3]. COVID-19 misinformation is more prevalent and more likely to be believed in low-middle income countries [4], where it has a disproportionate impact on marginalized populations [5]. A survey of Kenyan youth found social media misinformation and misperceptions about adverse effects were determinants of intention to vaccinate [6]. A population-based study in Uganda found a high prevalence of COVID-19 misinformation, with 10% of the respondents believing that the COVID-19 vaccine does not work [7].

The negative impact of vaccine misinformation necessitates the development of interventions that neutralize its influence. This article aims to: 1) document an intervention, the Cranky Uncle Vaccine game, that combines inoculation theory and gamification to increase vaccine acceptance and discernment between vaccine facts and misinformation, and 2) report the results of pilot studies in both Kenya and Uganda, which sought to establish its effectiveness in increasing vaccine acceptance, intent to get vaccinated, and discernment between vaccine facts and fallacies.

Misinformation about vaccines can negatively impact vaccine acceptance [1], perceived threat of vaccine-preventable diseases [2], and vaccination uptake [3]. COVID-19 misinformation is more prevalent and more likely to be believed in low-middle income countries [4], where it has a disproportionate impact on marginalized populations [5]. A survey of Kenyan youth found social media misinformation and misperceptions about adverse effects were determinants of intention to vaccinate [6]. A population-based study in Uganda found a high prevalence of COVID-19 misinformation, with 10% of the respondents believing that the COVID-19 vaccine does not work [7]. The negative impact of vaccine misinformation necessitates the development of interventions that neutralize its influence. This article aims to: 1) document an intervention, the Cranky Uncle Vaccine game, that combines inoculation theory and gamification to increase vaccine acceptance and discernment between vaccine facts and misinformation, and 2) report the results of pilot studies in both Kenya and Uganda, which sought to establish its effectiveness in increasing vaccine acceptance, intent to get vaccinated, and discernment between vaccine facts and fallacies.

Inoculation theory As misinformation is hard to dislodge once it has taken hold [8], increasing attention has been paid towards pre-emptive strategies to neutralize misinformation before exposure. Inoculation theory offers a psychological framework for pre-emptive interventions, applying the metaphor of vaccination to knowledge [9]. Inoculation involves exposing people to a weakened form of misinformation in order to build up their cognitive immunity to real-world misinformation.

Inoculating messages consist of two elements: a warning of the threat of being misled, and counter-arguments explaining how the misinformation misleads. The threat warning puts the recipient on guard that they are vulnerable to being misled, motivating them to protect themselves from being deceived by false arguments [10]. The counter-arguments provide specific information about the false arguments in order to help recipients identify misleading content. Inoculation has been tested on misinformation across a range of topics, including climate change [11,12] and health [13].

A common inoculation approach, described as fact-based, exposes how the misinformation is false through factual explanations [14]. A limitation of the fact-based approach is that it can only address specific pieces of misinformation [15], and can be canceled out by misinformation [11,16]. Consequently, interest is growing in more general approaches such as technique-based or logic-based inoculation, which involves explaining the logical fallacies or rhetorical techniques in misinformation [14,17]. While both fact-based and technique-based approaches are effective in correcting specific misinformation, the technique-based approach can generalize across topics [12]. While fact-based communication may be canceled out, technique-based approaches are more robust to being presented alongside misinformation [18].

One concern about technique-based inoculation is that while they reduce vulnerability to misinformation, they may also reduce trust in facts by fostering general distrust [19,20]. Consequently, a growing theme in misinformation research is the concept of discernment, which focuses on people's ability to accurately distinguish facts from misinformation, as opposed to solely focusing on reducing belief in or vulnerability to misinformation [21]. This study focuses on the FLICC framework, summarizing five rhetorical techniques used in science misinformation: fake experts, logical fallacies, impossible expectations, cherry picking, and conspiracy theories, a framework that has been expanded into a more comprehensive taxonomy of rhetorical techniques [22]. Building public awareness of the large number of misleading techniques presents an education challenge, requiring innovative, creative approaches that engage and educate the public. Most inoculation interventions adopt a passive form of communication, where recipients passively read the inoculating message. However, a more interactive approach known as active inoculation has recipients actively generate misleading arguments to build their resistance to misinformation [23]. Another way to engage audiences is through digital games.

Using games to counter misinformation

Games offer a tool for overcoming the challenges of misinformation by packaging educational material in an engaging, interactive format [24]. Gaming interventions are more effective than other types of interventions in reducing cognitive biases, with some evidence that game interventions convey benefits across topics [25].

A number of games have focused on addressing the problem of misinformation. In the Bad News game, players adopt the role of a fake news producer, learning the manipulative techniques that attract and mislead social media followers [26]. Follow-up games using the same template include Go Viral, focusing on COVID-19 misinformation [27] and Harmony Square, addressing misinformation that undermines democratic elections [1].

Cranky Uncle Vaccine is a browser and smartphone misinformation game that synthesizes active inoculation research with the literature on humor-based communication. It was developed in collaboration with UNICEF, Sabin Vaccine Institute (Sabin), and Irimi Company. The Cranky Uncle Vaccine game focuses on misinformation about vaccines, adapted from the general digital game Cranky Uncle, which addressed misinformation across a range of scientific topics [28]. While the inoculation in Cranky Uncle is purely logic-based, the focus on vaccination in Cranky Uncle Vaccine necessitated the inclusion of fact-based inoculation, communicating facts about the safety and efficacy of vaccines. A literature review identified ten of the most common fallacies found in vaccine misinformation globally [29] (see section S1 for full details). A health worker character was added to the game to deliver factual information (Figure 1 A,B) while Cranky Uncle (Figure 1 C,D) explained fallacy techniques.

An explicit threat warning is provided in the onboarding process at the start of the game, which informs the player that Cranky Uncle uses tricks to cast doubt on facts, and that understanding Cranky Uncle's tricks is the key to being less likely to be fooled. Refutational preemptions are provided throughout the explanations of each fallacy, which feature two-sided presentations of vaccine facts and Cranky Uncle's rhetorical tricks in response. As Cranky Uncle explains how his techniques are designed to mislead and distort or ignore the facts, they are presented as weak counterarguments against the facts (Figure 1).

To produce regional versions of the game with localized content, co-design workshops were conducted in Uganda, Kenya, and Rwanda with parents and child caregivers, young people, and community health workers [29]. Workshop feedback was incorporated, resulting in the development of localized character designs and the simplification of the game's language. As well as Cranky Uncle and the health worker character explaining vaccine facts and fallacies (Figure 1 A,D), a set of supporting characters consisting of an older woman, a younger man, a younger woman, and a man with a disability were used in quiz questions (Figure 1 E,G).

A pilot study was conducted in Rwanda, testing French and Kinyarwandan translations of the East-African version [30]. The game showed the greatest effects among vaccine-hesitant players - 55% of French and 71% of Kinyarwandan players switched to being likely to get vaccinated after completing the game. In addition, an English version of the game was co-designed and pilot-tested in Ghana [31]. This pilot showed similar results to Rwanda, with more than half of vaccine-hesitant players (53%) switched to being likely to get vaccinated at the end of the game. In this paper, we document a pilot study of the English version of Cranky Uncle Vaccine in Uganda and Kenya, to test the effectiveness of Cranky Uncle Vaccine in improving vaccine attitudes and the ability to discern facts from fallacies in those two countries.

Methods     

Study design: a pilot study was conducted to determine change in vaccine attitudes and misinformation discernment after playing the Cranky Uncle Vaccine game.

Study setting and population: the pilot study was conducted in Uganda and Kenya, located in East Africa. The Ugandan pilot ran from February 2023 to May 2023, while the Kenyan pilot was conducted from March 2023 to May 2023. In Uganda, participants were recruited using a multi-pronged approach. Partner organizations that participated in the co-design workshop, including Makerere University (targeting undergraduate and master's students), Busoga Health Forum (targeting health professionals), and +256 Youth Platform (targeting rural-based young people), facilitated recruitment through convenience sampling. The recruitment process focused within closed groups on the organizations' social media platforms, including WhatsApp and Facebook. Closed groups were used to ensure recruited and enrolled participants were primarily from and within the target country. Representatives from the partner organizations also actively recruited participants through in-person contact and one-to-one messaging on WhatsApp and Facebook. A Snowball sampling technique, where existing study participants were encouraged to share the link with their contacts, was used to reach additional participants, particularly, non-health sector participants and individuals from older age groups. For the Kenyan pilot, participants were recruited via snowball sampling through various personal contacts (e.g., telephone calls, social media, email, in-person). KEMRI also contacted previous study participants who had consented to future communications. Convenience sampling was used to select Trans-Nzoia and Nairobi counties as the enrollment sites. The University of Nairobi Vaccination Implementation team is well established within the rural community of Trans-Nzoia, Kenya, and encourages uptake of routine childhood immunization by working closely with a large cohort of parents and child caregivers, a key target audience for Cranky Uncle Vaccine. KEMRI and AMREF were well established in Nairobi, which provided access to an urban population.

Data collection: for both country pilots, when players first opened Cranky Uncle Vaccine, they filled out a research quiz measuring demographics, vaccine attitudes, and ability to detect fallacies in vaccine misinformation. After completing the survey, players proceeded to play the game. If the player completed all ten explanations of misinformation fallacies (and had consented to fill out the pre-game research survey), they were then prompted to fill out the same research survey post-game (minus the demographic items).

Variables: we recorded the median time spent playing the game and the median time spent on the overall study, including surveys and gameplay. In the pre-quiz research survey, participants were asked demographic questions covering age, gender, region, and education level. In the moderation analysis, participants were divided into two groups by age (younger and older groups) and education (lower and higher education). To measure vaccine attitudes, participants answered three questions on a four-point scale, measuring whether the participant was personally for or against vaccination, agreement with the statement “I feel that it is important that I get vaccinated,” and likelihood to get recommended vaccinations for themselves or their children [32]. To measure discernment between facts and fallacies, participants were shown two factual statements and six misleading statements, each representing a different fallacy, and asked to rate how true or false they thought they were on a four-point scale. The fallacies were natural is best, false cause, evil intent, conspiracy theory, pick and choose, and personal attack. After completing the game, participants answered two open-ended questions in the post-game survey on what stood out in the game and what changes they would like made.

Data analysis: paired t-tests were conducted to measure changes between pre-game and post-game measures of vaccine attitudes, fact reliability, and fallacy reliability. ROC (Receiver Operator Characteristic) analysis was used as an alternative measure of discernment, incorporating the fact and fallacy reliability data. Paired samples t-test was used to test whether AUC (a single value summarising discernment) significantly increased after playing the game. Linear regression was conducted to explore the moderating effect of age, gender, and education level on the game's effectiveness in improving vaccine attitudes, fact perceptions, and fallacy perceptions. Content analysis of the two feedback questions was conducted by first using ChatGPT to identify themes in the responses, then categorise each response into one of the themes.

Ethical consideration: the Ugandan pilot was approved by the Makerere University School of Public Health Research Ethics Committee (SPH-2022-317) and the Ugandan National Council for Science and Technology (HS2519ES). The Kenyan pilot was approved by the KEMRI Scientific and Ethics Review Unit (SERU No. 4556).

Results     

Participants and descriptive data. In Uganda, 1058 participants were recruited. For the Kenyan pilot, 1031 participants were recruited. The median time spent playing the game was 22.1 minutes, while the median time spent on the overall study, including surveys and gameplay, was 26.9 minutes. Participants' demographic details covered age (M = 27.8, SD = 7.8), gender (47.8% female, 49.9% male), and education level (median education level “Some/all college”). Dividing the dataset into two groups by age yielded a younger group (18 to 25 years, M = 22.4 years, N = 1052) and older group (26+, M = 33.3 years, N = 1037). Participants were divided into two groups by education: lower education (“Some/all college/diploma level” or lower, N=1141) and higher education (“Some/all university/degree level” or higher, N=948). Participants were also split into two groups by education: lower education (“Some/all college/diploma level” or lower, N=1141) and higher education (“Some/all university/degree level” or higher, N=948).

Main effects

Figure 2 shows the pre-game and post-game vaccine attitudes from both pilots. To control for Type I error rate due to multiple comparisons, a Bonferroni correction was applied, adjusting the alpha level to 0.0045. For the Uganda participants, paired t-tests showed significant improvement from pre-game to post-game in general vaccine attitude (t(1057) = -14.1, p < .001, d = 0.43), vaccine importance (t(1057) = -9.5, p < .001, d = 0.29), and intent to get vaccinated (t(1057) = -8.4, p < 0.001, d = 0.26). Similarly for the Kenyan participants, significant improvement was observed in general vaccine attitude (t(1030) = -6.5, p < 0.001, d = 0.20), vaccine importance (t(1030) = -3.5, p < 0.001, d = 0.11), and intent to get vaccinated (t(1030) = -4.1, p < .001, d = 0.13). Across both countries, among the 165 participants who expressed vaccine hesitancy in the pre-game survey (somewhat or very unlikely to get vaccinations), 58% switched to being somewhat or very likely to get vaccinations in the post-game survey.

Figure 3 shows the pre-game and post-game perceived reliability of vaccine facts and fallacies for both countries. In paired t-tests, Ugandan participants showed improvement in perceived reliability for both fact #1 (t(1057) = -8.0, p < 0.001, d = 0.25) and fact #2 (t(1057) = -4.4, p < 0.001, d = 0.14). While Kenyan participants showed significant increase in reliability of fact #1 (t(1030) = -5.3, p < 0.001, d = 0.16), they showed no significant change in fact #2 (t(1030) = 0.9, p = 0.38, d = .03). Ugandan participants showed a significant decrease in their reliability ratings for the natural is best fallacy (t(1057) = 6.7, p < 0.001, d = .21), false cause (t(1057) = 8.5, p < 0.001, d = 0.26), evil intent (t(1057) = 7.3, p < 0.001, d = 0.23), pick and choose (t(1057) = 7.3, p < .001, d = 0.17), and personal attack (t(1057) = 6.3, p < 0.001, d = 0.19). The only exception was conspiracy theory, which showed no significant change (t(1057) = 1.6, p = 0.104, d = 0.04). Kenyan participants also showed a significant decrease in their reliability ratings for the natural is best fallacy (t(1030) = 5.8, p < 0.001, d = 0.18), false cause (t(1030) = 6.9, p < 0.001, d = 0.21), evil intent (t(1030) = 6.7, p < 0.001, d = 0.21). However, conspiracy theory showed no significant change (t(1030) = -0.6, p = 0.53, d = 0.02), while personal attack (t(1030) = 2.7, p = 0.008, d = 0.1) and pick and choose (t(1030) = 2.8, p = 0.005, d = 0.09) showed marginal improvement given the Bonferroni-adjusted alpha value of p = 0.0045.

ROC analysis offers an alternative method to test whether Cranky Uncle Vaccine improves discernment between factual and fallacious statements [33]. For Uganda, a paired samples t-test showed an increase from pre-game AUC values (M = 0.71, SD = 0.32) to post-game AUC values (M = 0.76, SD = 0.33) with the increase in AUC being statistically significant, t(1057) = 6.52, p < .001, d = 0.154, 95% CI [0.035, 0.067], BF10 = 0.707. Similarly, for Kenya, a paired samples t-test showed an increase from pre-game AUC values (M = 0.80, SD = 0.27) to post-game AUC values (M = 0.82, SD = 0.28), noting that pre-game AUC values in Kenya started higher than those in Uganda. Correspondingly, the increase in AUC in Kenya was statistically significant but with a smaller effect size, t(1030) = 2.11, p= 0.035, d= 0.063, 95% CI [0.001, 0.033], BF10 = 0.319.

Moderating effect of demographics

To explore the moderating effect of demographics (age, gender, and education level) on the game's effectiveness, three composite variables were calculated representing fact perception (average of the two fact items), fallacy perception (average of the six fallacy items), and vaccine attitude (average of the three vaccine items). Linear regression was conducted to determine whether age, gender, or education level predicted change in the three composite values from pre-game to post-game. There was a collective significant effect between gender, age, and education on the change in fact perceptions (F(3, 2085) = 6.406, p < 0.001, R2 = 0.009) with age (t = 2.206, p = 0.027) and education (t = -3.783, p < 0.001) being significant predictors. Similarly, there was a collective significant effect between gender, age, and education on the change in fallacy perceptions (F(3, 2085) = 3.879, p = 0.009, R2 = 0.006), with age (t = -2.718, p = 0.007) being the only significant predictor. Lastly, there was a collective significant effect between gender, age, and education on the change in vaccine attitudes (F(3, 2085) = 4.668, p = 0.003, R2 = 0.007), with the only significant predictor being education (t = -3.392, p = 0.001).

Figure 4 visualizes how demographic variables influenced the effectiveness of the game across different demographic groups. For average vaccine attitude (Figure 4 A), lower education levels started at a lower vaccine attitude than higher education levels, with both groups ending at similar levels. On fallacy perceptions, the game was more effective with older participants compared to younger participants (Figure 4 B). Younger participants began with higher agreement with fallacy statements, but older people showed a greater reduction in fallacy belief after the game. Fact perceptions show similar patterns to vaccine attitudes across different education levels, with lower education participants showing lower factual perceptions pre-game, but both groups show similar levels of factual perceptions post-game (Figure 4 C). Both younger and older participants started at similar fact perceptions, but older people showed a greater increase in fact belief after the game (Figure 4 D).

Content analysis of game feedback

In the post-game survey, players were asked what stood out in the game and what changes they would like made. Both sets of answers were content-analyzed using ChatGPT. ChatGPT is equal or superior to humans in some annotation tasks [34] and offers multi-lingual annotation [35], an important feature given current efforts to translate Cranky Uncle Vaccine into Kinyarwanda, Swahili, French, Urdu, and Roman Urdu. The highlight themes are listed in Table 1, showing that the importance or benefits of vaccines were the dominant theme. The requested change themes are listed in Table 2, with the most prevalent answer being no change. Players urged that the game continue to be maintained and made available to the public.

Discussion     

In this study, we pilot-tested the effectiveness of Cranky Uncle Vaccine in improving vaccine attitudes and discernment between vaccine facts and misinformation in Uganda and Kenya, finding positive impacts in both countries. The game was more effective in the Ugandan pilot, which had an older and less educated participant sample relative to the Kenyan participants. As the game was more effective with older participants and less educated participants, the difference in results may be due to differences in the demographics of the pilot participants. Overall, our results show that digital games are an effective tool for countering the negative influence of misinformation, a finding consistent with past studies [20,27]. An important finding was that while agreement with vaccine fallacies decreased, agreement with vaccine facts also increased. This indicates that the game improved players' ability to discern between facts and fallacies. Possible mechanisms by which Cranky Uncle Vaccine was able to raise factual perceptions include the combination of fact-based and technique-based inoculation, and the feedback that players receive every time they fill out a quiz question.

Moderation analysis showed that the game was most effective among participants with lower levels of education as well as older participants. This has implications for which community groups should be targeted for this style of intervention. An unexpected result was older players showing greater improvement relative to younger players (Figure 4 B,D), especially given the qualitative feedback indicating that the game resonated more with younger players. The result is particularly confounding as older players began the game with more accurate perceptions of facts and fallacies compared to younger people. The qualitative feedback obtained in the post-game survey was overwhelmingly positive. Players appreciated the novel combination of interactivity, cartoon characters, and educational content having fun while boosting their vaccine literacy and critical thinking. They enjoyed the cranky uncle character and his facial expressions, demonstrating how humor can make an educational game more entertaining [36]. Another recurring comment was the challenging nature of the critical thinking game, encouraging feedback given that greater cognitive effort leads to longer-lasting effects [37].

A limitation of our study is that, as the pilot studies recruited convenience samples, the participant samples weren't representative of the countries' populations, skewing younger with higher levels of education, particularly in the Kenya population. Given the high levels of education among the participants, and the lower-educated participants showed greater improvement in vaccine knowledge and attitudes, future research should look to recruit participants who are more representative of public education levels. Another limitation is that the study lacked a control group or alternative treatments. Future research could compare a control group to different inoculation techniques and delivery platforms/formats such as videos, chatbot games, or interactive voice response (IVR) games. The research survey also failed to measure perceived threat and counterarguments, both important concepts in inoculation research [10,38].

What remains is to test digital misinformation games in a broader set of circumstances. While Cranky Uncle Vaccine has been shown to work in several African countries across three languages, is the game effective in different cultural contexts where different vaccine misconceptions and fallacies may be more prevalent? Does the game's approach still work when applied in different formats, such as chatbots, print, and audio versions? Other studies have shown equivalent results comparing a smartphone game vs. print version [39]. An important practical question is how the game might be incorporated into larger immunization programs, such as training for community health workers, equipping them to better respond to vaccine misinformation from the public.

Conclusion     

In summary, the Cranky Uncle Vaccine game has been shown to be an effective tool in improving vaccine attitudes and misinformation resistance in pilots conducted in Uganda and Kenya. This adds to the growing body of literature finding that digital games are a valuable and scalable tool in countering misinformation. It also builds on research exploring misinformation interventions in the Global South, an important contribution given that over 80% of misinformation intervention research was conducted in Global North countries. These results point to the potential of Cranky Uncle Vaccine both as a public game played on smartphones or browsers, and as part of other educational programs, such as in school curriculum or training of community health workers. This study lays a foundation for future research and deployment of the game in a range of countries, languages, and other formats, to facilitate at-scale programs among diverse populations, building vaccine acceptance and resistance against misinformation.

Competing interests     

The auhtors declare no competing inetrests.

Authors' contributions     

John Cook, Surangani Abeyesekera, Chelsey Lepage, Angus Thomson, Stacey Lizbeth Knobler, and Kathryn Lee Hopkins contributed to research design and Wendy Cook contributed to the intervention content design and image creation. Doris Njomo, Caroline Aura, Benson Wamalwa, Jacquellyn Nambi Ssanyu, Lydia Kabwijamu, Michael Ofire and Peter Waiswa contributed to pilot study management, recruitment of study participants, and data collection. John Cook conducted the literature review and data analysis. Chelsey Lepage, Angus Thomson, Stacey Lizbeth Knobler, Kathryn Lee Hopkins contributed to the interpretation of results. John Cook led the first draft of the manuscript, and Doris Njomo, Caroline Aura, Benson Wamalwa, Surangani Abeyesekera, Jacquellyn Nambi Ssanyu, Lydia Kabwijamu, Michael Ofire, Chelsey Lepage, Angus Thomson, Peter Waiswa, Wendy Cook, Stacey Lizbeth Knobler, and Kathryn Lee Hopkins contributed to reviewing and further revisions, writing of the manuscript. All authors provided final approval of the paper prior to submission.

Acknowledgments     

This paper has been published with the permission of the Director General of KEMRI. We thank KEMRI, Amref Health Africa, Clinton Health Access Initiative, and the Ministry of Health in Kenya for their support and collaboration in recruiting participants in Kenya. We also extend our gratitude to Makerere University, the Busoga Health Forum, and the +256 Youth Platform in Uganda for their support and collaboration during participant recruitment.

Tables and figures     

Table 1: themes identified in response to the question "What stood out the most to you in the game?", percentage of participants from Uganda and Kenya in each theme, and examples

Table 2: themes identified in response to the question "What changes would you like made to the game?", percentage of participants from Uganda and Kenya in each theme, and examples

Figure 1: screens showing an explanation of a misinformation technique: (A, B) show the health worker character explaining facts about vaccines and vaccine research (fact-based inoculation); (C, D) show Cranky Uncle explaining a misinformation technique he uses to cast doubt on vaccine facts; selected quiz questions using different supporting characters to give examples of fallacies: E) pick and choose with the young man and young woman; F) personal story with the older woman; G) false cause with the professional with disability; H) natural is best

Figure 2: vaccine attitudes before and after playing the game, measuring whether participants are generally for or against vaccines, perceived importance of vaccines, and likelihood of getting recommended vaccinations in A) Uganda and B) Kenya

Figure 3: perceived reliability of vaccine facts and fallacies in A) Uganda and B) Kenya; lighter shades represent pre-game measures; darker shades represent post-game measures

Figure 4: visualization of moderation effects for significant moderators: A) average vaccine attitude for lower and higher educated participants; B) average fallacy perceptions for younger and older participants; C) average fact perceptions for lower and educated participants; D) average fact perceptions for younger and older participants

References