Entertainment-education media can be an effective strategy for influencing health behaviors [1-3]. However, entertainment-education media face the same challenges as other traditional persuasion methods [4]. Persuasive health messages often fail to achieve the desired effect [5], and in some cases, may arouse the motivation to reject a message, a phenomenon known as reactance [6].

The theory of reactance comprises 4 elements [6]: (1) freedom, which individuals possess insofar as they are aware of it and can enact it; (2) threat to freedom, which involves any pressure on the individual, making it more difficult to enact that freedom; (3) reactance, which is the motivation to re-establish the freedom if that freedom is eliminated or threatened with elimination; and (4) direct restoration, which involves the freedom of the individual to perform a forbidden act. Research in this field has led to the development of several strategies to reduce reactance to health messages pertaining to littering [7], use of e-cigarettes [8], use of alcohol [9], and eating behaviors [10], among other health-related messages [11-16].

We are particularly interested in the effect of social authority on reactance to persuasive health messages. Our starting assumption is that an agent (human or otherwise) that delivers a health message possesses some influence or social authority [17]. For example, persons who have high social authority, such as experts or doctors, are often recruited to promote health messages [18-20]. However, research has shown that individuals may perceive health messages from experts as coercive, threatening, or having an ulterior motive [21], which could provoke reactance and negate the impact of the intervention [22,23].

In a web-based entertainment-education video setting, there is limited, high-quality experimental evidence on the relationship between reactance and the perceived social authority of a message agent. Using a randomized controlled trial (RCT), we will evaluate the effect of social authority status on reactance to a short animated video on the intake of added sugars. The sugar message will be narrated by either (1) a preadolescent daughter, who has low authority status relative to the other narrators, (2) the daughter’s mother, who has equivalent social authority to the target audience, or (3) the family physician, who is an expert with high social authority. Results from this study will facilitate the development of videos for reducing reactance and improve the persuasiveness of health messages in web-based settings.

This study aims to achieve the following objectives:

Our null hypothesis is that the social authority of the child, who has low perceived social authority, the mother, who has equivalent perceived social authority, or the family physician, who has high perceived social authority, will have no effect on reactance to a video about reducing sugar intake.

This study consists of a parallel group RCT. Participants will be randomized to 1 of 5 arms: either the same sugar-intervention video narrated by a preadolescent daughter (arm 1: low social authority), the daughter’s mother (arm 2: equivalent social authority), or a family physician (arm 3: high social authority), or a content placebo video with a health message about tanning and sunscreen (arm 4: no sugar message), or a placebo video about earthquakes (arm 5: no sugar or health message). We will randomize the participants in a 1:1:1:1:1 ratio to the trial arms. Participants will watch 1 video once from start to finish.

Nested in each of the five trial arms is a list experiment. For each list experiment, participants will be randomized at a 1:1 ratio to a control or treatment group. The control group will receive a list of 5 items about behavioral intent (unrelated to sugar consumption). The treatment group will receive the same 5 items and a sensitive item about behavioral intent to reduce sugar intake. We will use the list experiment to reduce social desirability bias, as participants may already be primed to answer favorably to questions about sugar consumption.

At the end of the study, participants assigned to the content placebo (arm 4) or placebo (arm 5) will be given a choice to watch the video intervention. Their choices will be recorded. Participants who choose to watch will be randomized at a 1:1:1 ratio to the sugar video narrated by the daughter, mother, or physician. The complete trial flowchart is presented in Figure 1.

The study setting will be on the internet. We will use the web-based recruitment platform Prolific [24] to enroll study participants. We host and deploy our study on a web-based platform called Gorilla [25], a cloud platform that provides versatile tools for web-based, experimental, and behavioral research [26].

Eligibility criteria included being between the ages of 18 and 59 years (male, female, or other), being able to speak English, and being a resident in the United Kingdom; not being eligible according to any of the inclusion criteria was the exclusion criteria. We will not exclude participants on an existing health condition (eg, diabetes) because Prolific does not collect health information from its users.

The participants will undergo a process of informed consent. The consent form, which will be hosted on the Prolific platform, explains the purpose of the study, the risks and benefits of the research, and how a participant can contact the researcher or the human subjects review board at Heidelberg University. By clicking a link, participants will consent to participate in the study and will be redirected to the Gorilla platform. The Gorilla landing page contains additional information about the platform. Participants can exercise their freedom to not participate at any point during the study.

Participants are expected to finish the trial (watch the video, answer the survey questions, and complete the list experiment) in 10 minutes. The complete schedule of enrollment, interventions, and assessments is shown in Figure 3.

We calculated the sample size needed for pairwise comparisons between the 3 groups using a one-way analysis of variance. The formula used to calculate the sample size is [44]

where κ=1, which is the matching ratio, µ_A and µ_B are the group A and B means, σ_A and σ_B are group A and B standard deviations, α=.05 is the type-1 error, β=.20 is the type-II error, z is the quantile function, and τ=2 is the number of comparisons to be made. For the control and treatment groups, we assumed a mean of µ_A=2.0 and µ_B=2.15, respectively (in other words, we expect, on average, that the control group will agree with 2 out of the 5 items and the treatment group will agree with 2.15% of the 6 items). We selected σ_A=0.85 and σ=1.0; this calculation yields a sample size of n=769 per group. For a 5-way comparison, the sample size is n=3845. To ensure sufficient power and account for attrition, we will select a sample size of n=4000.

We will use the Prolific platform to recruit the study participants. The user must create an account on Prolific and provide their personal information. Participants must agree with Prolific’s data privacy terms and conditions. Prolific will assign each participant a unique, anonymized ID. The study investigators will also create a Prolific account. We will instruct the Prolific platform on how many participants need to be recruited, and it will filter out all participants who do not meet the eligibility criteria. Participant entry into our study will happen on a first-come, first -served basis until the recruitment number (sample size) is reached. We will compensate the participants an equivalent of £1 (US $1.4) for the expected 10-minute completion time.

All trial participants will be assigned a unique, anonymized string ID. The ID will be used on the Gorilla platform and linked to the participants’ responses. Gorilla will store the trial data on its cloud platform hosted on Microsoft Azure in the Republic of Ireland. The Gorilla database is encrypted using industry-standard cryptography. The study investigators own the research data collected using Gorilla and have complete control over it. The study investigators can generate and access the completely anonymized data from the Gorilla platform. The data will be downloaded and stored safely for statistical analysis on a computing system maintained by Heidelberg University in Germany.

Participants, who are completely anonymous to us, will have no identifying information associated with their unique IDs. We will inform participants that if they email the study investigators then their names could be revealed to us. The study investigators will keep this information confidential.

This document is the full protocol. Additional data or documentation can be requested from the corresponding author.

Ethical approval was obtained from the Heidelberg University’s ethics committee (Universität Heidelberg Ethikkommission der Medizinische Fakultät) on March 18, 2020, protocol S-088/2020. All participants will undergo a process of informed consent. The consent form, which will be hosted on the Prolific platform, explains the purpose of the study, the risks and benefits of the research, and how a participant can contact a researcher (and/or the human subjects review board at Heidelberg University). By clicking the link, participants consent to participate in the study and are redirected to the Gorilla platform. The landing page contains additional information about the Gorilla platform. Participants can exercise their freedom to not participate at any point during the study (see Multimedia Appendix 1 for the informed consent form).

The data will be collected and stored on the Gorilla platform. The study investigators own and have complete control of the research data, which can be accessed at any time. For statistical analysis, the data will be downloaded and safely stored in a computing system maintained by the University of Heidelberg.

There is growing evidence that entertainment-education media can be an effective strategy for promoting healthy behaviors [45-48]. However, further research is needed to understand which entertainment-education components can be modified to reduce reactance to health messages [3,49,50]. In this proposed study, we focus on a modifiable component—the perceived social authority of the health messenger—and its effect on reactance to a message about reducing sugar intake.

In recent years, video-based animation has emerged as a potentially powerful entertainment-education strategy for changing behavior [51-53]. This animation format has enabled the creative use of nonhuman and nonadult characters to promote more persuasive health messages [54]. We leveraged this animated format to create 3 culturally neutral characters that narrate a health message about sugar consumption. During the design phase, we assumed that a child narrator would be a more persuasive messenger because she would be perceived as nonthreatening or lacking a vested interest and therefore would be less likely to arouse reactance when compared with an adult narrator. However, we also considered that the child would not be taken seriously or that her lack of expertise would nullify the persuasiveness of the health message. We were unable to find prior research studies to inform our decision to use a child narrator. To this end, we propose an RCT to investigate whether reactance to the sugar message would be reduced if it was narrated by a preadolescent daughter, the daughter’s mother, or a family physician. Our results may show that a child can be a powerful and persuasive health promotion agent, which could inform future choices regarding the design and delivery of health messages.

In a systematic review, Shen and Han [4] concluded that there is a lack of experimental methods to evaluate the effectiveness of entertainment-education media. They call for “controlled experiments to uncover the cognitive and/or affective factors that mediate entertainment-education’s effects” [4]. Our protocol responds to this call by leveraging experimental methods and carefully considering several factors that may mediate the role of social authority on reactance to the sugar reduction message, which we discuss below.

First, our study will use an RCT design to randomize participants to 1 of 3 sugar-intervention videos, a content placebo video, or a placebo video. The 3 intervention videos are exactly the same, except that the sugar message is narrated by the daughter, the daughter’s mother, or the family physician. The RCT should ensure that the enrollment stage does not introduce systematic differences between trial arms. For example, participants may have pre-existing health conditions, such as diabetes, which could affect their responses to the survey questions. However, random assignment will take care of this potential source of bias by distributing it uniformly across trial arms. Thus, holding all else equal, and because of randomization, differences in reactance toward the sugar message should be because of the experimental manipulation of the narrator’s social authority.

Second, the content placebo and placebo videos are an innovative feature of our study and will enable us to isolate the health awareness effect and content effect of the intervention video. The content effect can be quantified as the difference in mean state reactance (MSR) between the intervention arm and the content placebo arm. As both videos promote a health message, and because we will randomize, any significant difference in MSR should be because of the sugar reduction content of the intervention video. We can calculate the health awareness effect as the difference in MSR between the content placebo arm and placebo arm. As the content placebo video promotes a health message and the placebo video does not, the difference in MSR should be because of the health awareness generated by the sunscreen message. Thus, we can decompose the total intervention effect—the difference in MSR between the intervention and placebo arms—as the sum of health awareness and content effects. We are not aware of any previous entertainment-education study that has used an experimental approach to partition the effect of an intervention video in this way.

We do not believe that the differences between the placebo videos (which we did not create) and the sugar-intervention videos (which we created) will confound our results. Such differences may relate to the animation style, background shapes or colors, and target audience, among other design decisions. As mentioned, the main interest of our study is the difference in state reactance toward the sugar reduction message following random assignment to the 3 social authority levels. On the basis of the theoretical model described earlier, it is only the content of the message or the characteristics of the messenger that can threaten an individual’s freedom and arouse reactance. As the placebo video of earthquakes does not promote a health message, we expect it to arouse a very small (or even null) level of state reactance (the type of animation style of the placebo videos, for example, cannot realistically threaten an individual’s freedom). Therefore, the placebo video will provide a baseline measure of state reactance, which will enable us to quantify the content and health awareness effects. Further, we were very careful to select a content placebo video in which the messenger had culturally neutral or agnostic characteristics. In this case, the narrator of the sunscreen message is not seen, and it is not possible to determine his social authority. A possible exception is that the placebo videos are narrated by male voices and the sugar-intervention videos by female voices. However, it is unlikely that this design difference will be sufficiently large to bias our results significantly.

Third, we will conduct a list experiment in each of the 5 arms. This is the second experimental method that we leverage in our study design. We use a list experiment to reduce social desirability bias in participants’ responses to the behavioral intent questions about reducing sugar intake.It is likely that participants will already be primed to give socially acceptable responses to questions about their health and sugar consumption. The indirect questions (ie, how many statements do you agree with) provide protection to participants if they want to reject the sugar message without revealing this intention. To the best of our knowledge, few (if any) studies have used a list experiment approach to evaluate the effectiveness of an entertainment-education video to improve a given health outcome.

Finally, we will also use an experimental approach to measure participants’ engagement with the sugar videos. We will do this by piggy-backing this objective on the ethical requirement to give participants not randomized to the intervention the opportunity of watching the intervention video at the end of the study. Participants will be informed that they will not be compensated for their additional time, thus enabling us to estimate the proportion of participants who will voluntarily watch the sugar video and how long they watch the video. These findings will help us determine participants’ willingness to watch an entertainment-education video, especially when this willingness must be balanced against a time cost, as is generally the case in a real-world scenario.

We expect that our study will make important contributions to entertainment-education literature. The lessons learned can help us improve the design of entertainment-education videos that facilitate disseminating persuasive health messages to a global audience at a rapid scale.