The study sample was composed of residents of North Carolina, USA and neighboring states who were in social networks of students in the North Carolina State University Fisheries, Wildlife and Conservation Biology Program (FWCB), excluding immediate family members and other students (N = 340). North Carolina is a coastal state where shark bites regularly occur, and access to a wide network of family and friends of students at a large public university enabled us to develop a geographically and demographically diverse sample. This sampling strategy resulted in a small number of respondents from neighboring states (n = 4; Georgia and South Carolina).

Videos included in the two treatments (positive or pro-shark and negative or anti-shark framing) were identified using keyword searches on YouTube. Keywords were “shark,” “shark week,” and “Discovery Channel.” Official Shark Week footage and videos with high view counts were prioritized (Figure 1). Results from the keyword searches were refined based on clarity of the positive or negative message framing and total view count (minimum of 3,000 views). Additionally, we excluded videos containing celebrities, compilation videos, and videos over 4 min long. Positive-framed videos showed non-aggressive human-shark interactions (e.g., being pet by divers) and/or the portrayal of sharks through a scientific lens (e.g., the process of giving birth). Negative-framed videos showed sharks behaving violently or aggressively toward humans (e.g., shark attacks). This initial screening process identified 16 videos for potential inclusion in the two video treatments.

This initial list of 16 videos was refined to 10 videos based on the results of a systematic classification process completed by 73 undergraduate students within the FWCB program at NC State University. The 73 students watched each of the 16 videos and responded to four prompts asking how the video portrayed sharks on −3 to +3 value scales of “worthless to valuable,” “unpleasant to pleasant,” “harmful to beneficial,” and “bad to good.” We aggregated scores, calculated summary statistics and 95% confidence intervals. The final 10 videos (5 most positive and 5 most negative, Table 1) were based on the positive/negative ranking which emerged from this classification scheme and selected to maximize or minimize average rankings and remove any overlap in CIs between positive and negative videos. The chosen 5 most positive videos had an average attitude score of 1.92 (SD = 0.38), a combined run time of 11:32 min, and a median view count of 316,334 views. The chosen 5 most negative videos had an average attitude score of −2.04 (SD = 0.32), a combined run time of 14:10 min, and a median view count of 4,650,532 views. One video (“Sharks Love to be Petted, They're Like Dogs”) included in the final positive treatment was from National Geographic's SharkFest, not from Discovery Shark Week; however, it was returned via the above keyword search criteria and may reflect the relative scarcity of positively framed Shark Week content on YouTube that fulfilled our search criteria. Treatment playlists were shuffled and viewed in a random order. Respondents were randomly assigned to each treatment based on their association with the FWCB member who asked them to participate (Lawson et al., 2019; Casola et al., 2020; Beall et al., 2022).

The final pre-test and post-test questionnaires were broken into three parts. First respondents answered a set of Likert-type questions measuring their fear and perceptions of intentionality associated with shark bites. Fear consisted of two measures, fright, and danger. Fright was assessed by asking “On a scale of 1-10, how frightened are you of sharks? With 1 being not frightened at all and 10 being extremely frightened.” Danger was assessed by asking “On a scale of 1-10, how dangerous do you think sharks are? With 1 being not dangerous at all and 10 being extremely dangerous.” Intentionality was assessed by asking “On a scale of 1-10, how intentional do you think most shark bites are? With 1 being completely accidental and 10 being completely intentional.” Second respondents were asked to indicate “Who do you think is most to blame when sharks bite people? Choose one.” Options were sharks, swimmers, no one, the government, and other. Third, respondents were asked to what extent they agreed or disagreed with various ways a local government could respond when a shark bite occurs. Response options are listed in Table 5. All scales were adapted from Pepin-Neff and Wynter (2018b). All respondents answered all three sections in both their pre-test questionnaire and post-test questionnaire.

The post-test questionnaire contained additional questions related to past experience with sharks. Past experiences included seeing a shark in the wild, seeing a shark at an aquarium or catching a shark while fishing. Additionally, we asked respondents if they had swam in the ocean, fished in the ocean, or surfed in the ocean within the last year. Lastly, in the post-test questionnaire respondents reported sociodemographic attributes including gender, age, education, residency, and income.

Survey administration followed four main steps. First participants were asked to review the Informed Consent for Participation in Research. Consenting respondents were then asked to complete the pre-treatment questionnaire. Following completion, respondents were asked to watch the YouTube videos contained within their assigned treatment. Videos were displayed in a random order using YouTube's shuffle option. Participants then completed the post-treatment questionnaire. Both the pre-treatment and post-treatment questionnaires were administered using Qualtrics between March and April 2021. Completed responses were cleaned by removing straight-lined responses and those with abnormally fast completion times.

Treatment effects were measured through the comparison of pre-treatment and post-treatment survey responses. We began by calculating summary statistics (means and standard deviations) for pre and post treatment scores for fright, danger, and intentionality (Table 2) and agreement scores for the various ways a local government could respond when a shark bite occurs (Table 5). To test hypotheses H1 to H5, we compared mean pre and post test scores using paired t-tests to determine if significant changes occurred after treatment administration. We also used ordinary least squares regression to model the change in score for fright, danger and intentionality as a function of video treatment (binary, reference level = negative treatment), pre-treatment score (continuous; to control for a ceiling effect; Theobald and Freeman, 2014), demographic variables, past experience with sharks, and coastal recreational activities. Demographic attributes included age (continuous), college education (binary; 0 = no college and 1 = college), race (binary; 0 = white and 1 = non-white), sex (binary; 0 = male and 1 = female), political affiliation (continuous; 1 = very conservative, 3 = moderate, 5 = very liberal; Rhemtulla et al., 2012; Harpe, 2015), and residency in a coastal county (binary; 1 = coastal and 0 = inland). Coastal counties were identified by the classification outlined within the NC Coastal Area Management Act (NC Department of Environmental Quality, 1974). Past experiences were either seeing a shark in the wild, seeing a shark at an aquarium, or catching a shark while fishing (all binary). Coastal recreational activities were frequency of swimming in the ocean, fishing in the ocean, or surfing in the ocean within the last year (continuous; frequency). Changes in scores were calculated as the post-treatment value minus the pre-treatment value, with a positive change indicating increase in either fear or perceived intentionality and a negative change indicating a decrease in fear or perceived intentionality. To evaluate effect size, we also calculated standardized beta values for all regression models. To test hypotheses H6 and H7 we used chi-squared tests to measure the two treatment's effects on who respondents blamed for shark bites. Lastly, to test hypotheses H8 to H11, we compared mean pre and post test scores using paired t-tests to determine if significant changes occurred after treatment administration. All analyses were completed using R Version 4.2.1. The NC State University Institutional Review Board (IRB #23605) approved this study.

Pre-test scores for fright (M = 5.74, SD = 2.53) and danger (M = 5.61, SD = 2.34) were above 5.5 suggesting respondents had minor levels of fright and danger before receiving any of the treatments. Perceived intentionality (M = 4.24, SD = 2.35) was below 5.5 suggesting respondents perceived shark bites to be more accidental, rather than intentional, before receiving either treatment.

Hypotheses H1 to H5 were supported by both regression and t-test results, indicating positive YouTube messaging decreased measures of fright, danger, and intentionality associated with shark bites. Respondents who received the positive treatment reported a 24% decrease in fright (pre M = 5.95, post M = 4.55, β = −0.45, p < 0.001), a 27% decrease in danger (pre M = 5.69, post M = 4.17, β = −0.52, p < 0.001), and a 29% decrease in intentionality associated with shark bites (pre M = 4.34, post M = 3.10, β = −0.50; p < 0.001) post treatment (Tables 2, 3). Additionally, results indicated negative YouTube messaging increased measures of fright, danger, and intentionality associated with shark bites; however, the magnitudes of these changes were less than those elicited by the positive treatment. Respondents who received the negative treatment reported a 6% increase in fright (pre M = 5.53, post M = 5.85, p < 0.01), a 11% increase in danger (pre M = 5.54, post M = 6.17, p < 0.001), and a 24% increase in intentionality associated with shark bites (pre M = 4.13, post M = 5.11; p < 0.001) post treatment (Tables 2, 3). Coefficients for pre-test scores within the fright (β = −0.31, p < 0.001), danger (β = −0.45, p < 0.001), and intentionality (β = −0.38, p < 0.001) models indicate that respondents with very low or very high pre-test scores for all three measures were correlated with larger post treatment changes than respondents who reported moderate pre-test scores, likely a result of the ceiling effect (Theobald and Freeman, 2014; Tables 2, 3).

Regression results also indicate that in addition to treatment, demographic/behavior attributes also predict the pre-post changes observed. Political affiliation was significant in all three models, indicating larger decreases in fright (β = −0.17, p < 0.05), danger (β = −0.22, p < 0.01), and intentionality (β = −0.16, p < 0.10) for liberals compared to their conservative counterparts (Table 3). In the Change in Danger model, sex, and frequency of swimming in the ocean were also significant. Female respondents reported higher danger scores than male respondents (β = 0.47, p < 0.05), and frequent swimmers had lower danger scores compared to those for rarely or never swim in the ocean (β = −0.29, p < 0.05; Table 3). In the Change in Intentionality model, sex was also significant, indicating female respondents considered shark bites to be more intentional acts than male respondents (β = 0.42, p < 0.05; Table 3).

Results indicate support for H6 but not H7. Chi-square results indicate a significant change in blame pre-post positive treatment [X² (4, N = 172) = 17.91, p = 0.001]. As a result of the positive treatment, respondents tended to shift blame from sharks (15% pre-test, 5% post-test) to swimmers (38% pre-test, 42% post-test) or no one (42% pre-test, 48% post-test). Chi-square results indicate no changes in blame pre-post negative treatment [X² (4, N = 168) = 2.31, p = 0.679; Table 4].

Baseline preferences indicated an initial preference for non-lethal shark management strategies. The top four preferred strategies were non-lethal and were education (M = 4.66; Scale: 1 = Strongly Disagree to 5 = Strongly Agree), research (M = 4.29), technology (M = 4.00), and leave the shark alone (M = 3.61). Shark nets was the most preferred lethal option, ranking 5th overall, behind four non-lethal methods, with a pre-test mean approval score of 2.65. The bottom three strategies based on pre-test baseline scores were do nothing (M = 2.62), drum lines (M = 1.90), and hunt the shark that bit someone (M = 1.85).

Results indicate support for H8 and mixed support for H9, H10, and H11. Positive YouTube messaging decreased support for all three lethal post bite response measures. Support for shark nets decreased by 11% (pre M = 2.73, post M = 2.44, p < 0.001), support for hunting the shark that bit someone decreased by 11% (pre M = 1.93, post M = 1.71, p < 0.001), and support for baited drum lines decreased by 10% (pre M = 1.94, post M = 1.75, p < 0.01; Table 5). Negative YouTube messaging increased support for two of the three lethal measures, increasing support for hunting the shark the bit someone by 10% (pre M = 1.77, post M = 1.95, p < 0.001), and support for baited drum lines by 13% (pre M = 1.87, post M = 2.11, p < 0.01; Table 5). For the five non-lethal post bite response strategies, positive YouTube messaging increased support for three measures, increasing support for education by 3% (pre M = 4.61, post M = 4.73, p < 0.01), research by 4% (pre M = 4.24, post M = 4.40, p < 0.01), and leave the shark alone by 7% (pre M = 3.51, post M = 3.77, p < 0.001; Table 5). Negative YouTube messaging decreased support for two non-lethal measures, decreasing support for leave the shark alone by 8% (pre M = 3.72, post M = 3.42, p < 0.001) and do nothing by 8% (pre M = 2.66, post M = 2.45, p < 0.01; Table 5).

This study provides insights about shark-related social media that may be leveraged to promote shark conservation. Results support other studies showing that positive message frames about wildlife are consistently more impactful than negative frames and may help reduce fear, correct misguided perceptions, and promote pro-conservation attitudes and behaviors (Casola et al., 2020; Beall et al., 2022). Thus, popular shark-related media (e.g., Shark Week, SharkFest) would generate greater benefits for shark conservation by presenting positively framed messages about sharks. These findings highlight the importance of responsible media design by content creators and journalists. We realize these suggestions may be difficult to implement since negatively framed messaging has driven one of the most watched, most lucrative, and longest running cable television programming events in history (i.e., Discovery's Shark Week; Fetters, 2012) with related content on social media (e.g., Facebook and Instagram) engaging 48 million users in 2018 (Feldman, 2019). However, Discovery Channel has acknowledged some of the issues with Shark Week content under its former CEO (de Moraes, 2015), and other popular nature shows such as Planet Earth and Blue Planet have demonstrated how the drama and excitement that draws in viewers can be generated without negative framing.

Future studies should build upon this work by addressing several limitations of this study. Although respondents were shown multiple videos, this study is representative of a single virtual exposure to sharks. Additional research is needed to evaluate the long-term impacts of exposure to popular shark content specials such as Shark Week. Such evaluations are especially important because this type of content is created within the context of a larger set of programming, not as single standalone videos. Additionally, previous studies suggest single exposures are unlikely to produce long-term effects without future exposures to similar content (Bode, 2016). Future studies may include “filler tasks” which simulate the realistic scenario of combining different types of information presented in sequence (e.g., commercial breaks interspersed within a television show). Future studies could also strive for a simple random sample to reduce the potential for a pre-existing knowledge bias. Our sample was limited in geographic scope and may have included respondents overly aware of environmental issues (due to their network ties to college students in a wildlife-related major, and proximity to and high frequency of recreating in the ocean) compared to a random sample of the general public, potentially dampening treatment effects observed (Theobald and Freeman, 2014; Bode, 2016). Coastal residents are physically closer to the ocean and their behaviors and policy preferences are more likely to directly impact coastal ecosystems. Additionally, major coastal cities in the US lean politically liberal, which this and previous studies have indicated is likely to further increase the efficacy of positively framed content (Tausanovitch and Warshaw, 2014). Future studies should explore these possible geographic effects and how they may be leveraged to help reduce fear, correct misguided perceptions, and promote pro-conservation attitudes and behaviors. Measures of key outcome variables could also be expanded. Future research should consider the impacts to explaining post shark bite management options in more detail (e.g., efficacy at reducing future incidents, cost, impacts on sharks). This study used two basic measures to gauge fear among respondents. Future studies should build upon scales used in parallel contexts, such as the “Snake Anxiety Questionnaire,” to better understand the psychometric components contributing to overall fear (Klieger, 1987; Zsido et al., 2018) or explore implicit attitudes and their connection with self-declared attitudes. Similarly, future studies might consider impacts of a broader suite of messaging on various types of conservation policy support for sharks, extending beyond management preferences to include behaviors such as donations and volunteerism (Drymon and Scyphers, 2017; Bargnesi et al., 2020). These may include other forms of social media content or other public facing communiques such as educational signage, museum/aquarium content, or coastal tourism campaigns. Additional experimental research will continue to reveal the various ways that positively and negatively framed communication can influence public perceptions of and support for controversial species such as sharks.