Cracking the 'Meta' Code: Advanced Machine Learning-Based Sentiment Analysis of Water Fluoridation Debates on Facebook and Instagram

Cracking the 'Meta' Code: Advanced Machine Learning-Based Sentiment Analysis of Water Fluoridation Debates on Facebook and Instagram | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Cracking the 'Meta' Code: Advanced Machine Learning-Based Sentiment Analysis of Water Fluoridation Debates on Facebook and Instagram Nilesh Torwane, Ratilal Lalloo, Diep H. Ha, Loc G. Do This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6266081/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Social media platforms like Facebook and Instagram are pivotal in shaping public opinion on health interventions, including Community Water Fluoridation (CWF). Despite its recognition as a safe and effective public health measure, CWF remains a polarising topic, with misinformation on these platforms contributing to public mistrust. This study collected 109,117 Facebook and Instagram posts from 2014 to 2023 to examine public sentiment surrounding CWF. The analysis revealed a mix of opinions, with 42.1% positive, 39.1% negative, and 18.8% neutral sentiments. Trends highlighted a surge in negative sentiment during 2017–2019, likely influenced by misinformation and significant public events, while positive sentiment has gradually regained ground in recent years. Key themes included health benefits, safety concerns, and government trust, with positive discussions emphasising CWF’s role in public health and negative discussions focusing on risks and chemical exposure. The study used advanced sentiment analysis models to highlight the importance of monitoring public discourse and addressing misinformation to build trust and support for evidence-based health policies like CWF. These findings provide digital data-driven insights for public health communication strategies to enhance community understanding and acceptance of vital health interventions. Community Water Fluoridation Social Media Sentiment Analysis Public Health Communication Misinformation on Social Media Public Opinion Trends Health Policy and Public Perception. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction In today’s digital age, social media platforms like Facebook and Instagram, both part of the Meta family, play a significant role in shaping public opinion on social, political, scientific, and health-related issues [ 1 ]. These platforms are major sources of information, sharing news updates and spreading misinformation [ 2 ]. This makes them highly influential in forming public perceptions [ 1 , 2 ]. With a vast reach—Facebook alone has over 2.8 billion monthly active users—these platforms engage millions of people in health discussions, which increases their impact on public health [ 3 ]. The broad reach and high interaction levels make Facebook and Instagram powerful in influencing people’s beliefs and behaviours about health. One primary advantage of analysing sentiment on these platforms is the large amount and variety of user-generated content, which offers valuable data to understand public opinion across different groups and locations [ 4 ]. Facebook and Instagram encourage communities where people can share their experiences and views on health, even on controversial topics [ 5 ]. For example, studies have shown that misinformation about vaccines in Facebook groups has contributed to vaccine hesitancy and mistrust, showing the platform’s influence on health behaviours [ 6 ]. Similarly, parents discussing teething pain management often recommend unverified remedies, like amber necklaces, for relief, even though there is limited scientific support [ 7 ]. On Instagram, where visual content is popular, people share opinions through images and videos. This adds another layer to public opinion analysis but also spreads misinformation on topics like drug use, diet, and vaccines [ 8 ]. The popularity of unproven health products in online communities shows how these platforms shape health perceptions, often without scientific backing. This can create risks from unchecked information sharing [ 7 – 9 ]. A similar pattern is seen with Community Water Fluoridation (CWF), a well-established public health intervention that reduces dental cavities by adjusting fluoride in water supplies. Although CWF is supported by major health organisations like the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), and the Australian National Health and Medical Research Council (NHMRC), it remains a polarising issue, with opposition fuelled by misinformation on social media [ 10 – 14 ]. As with teething remedies, social media amplifies questionable studies and unverified claims, making it harder to maintain public support for evidence-based health practices like CWF [ 10 , 11 , 13 ]. Although widely recognised as safe and effective, CWF faces challenges due to the spread of misinformation, which undermines public trust in this vital intervention. Therefore, analysing sentiment on Facebook and Instagram offers a valuable way to track how public opinions on CWF form and evolve. Unlike traditional surveys, which capture only a single point, social media provides real-time data, monitoring public attitudes continuously as they react to new information or misinformation [ 9 , 14 ]. Facebook’s recommendation system, which suggests content based on user interactions, offers insights into how CWF-related information spreads in specific communities. Similarly, Instagram’s use of hashtags allows for tracking discussions and sentiment trends [ 10 , 15 ]. These tools are essential for identifying recurring misinformation trends and understanding how they affect public health communication. This study aimed to conduct a detailed sentiment analysis of Facebook and Instagram (Meta platforms) to understand public views on CWF. By exploring what drives these opinions, the research aims to provide findings that can guide public health communication strategies and inform policymakers. Ultimately, this study contributes to a deeper understanding of how social media affects public health perceptions, highlighting the importance of evidence-based policies and supporting better health outcomes related to CWF. Material and Methods Scope of the Study This sentiment analysis investigation used posts from Meta-based platforms like Facebook and Instagram to understand public sentiment regarding CWF. Our research spanned January 2014 to December 2023, using various parameters such as textual sentiment analysis, keyword analysis, topic modelling, and model performance testing. These 10 years were selected to capture a broad spectrum of discussions and shifts in public opinion over time, ensuring a comprehensive view of sentiment trends. The approach is aligned with the methodologies recommended in previous research on sentiment analysis in health communication and misinformation [ 16 – 18 ]. Facebook and Instagram were merged for this analysis because both platforms operate under the Meta company, sharing similar algorithms, data collection methodologies, and user interaction capabilities. As Meta-owned platforms, Facebook and Instagram allow users to seamlessly share posts, stories, and media across both networks. This cross-platform sharing provides unified access to public discussions and interactions, allowing for a more comprehensive dataset that reflects user behaviour across these integrated platforms. Merging the data enables capturing broader interactions and opinions on CWF while ensuring consistency in data collection, processing, and analysis [ 18 , 19 ]. Given their shared infrastructure, combining data from both platforms helps improve the validity and scope of the analysis by providing a fuller picture of sentiment across these widely used platforms. Data Confidentiality and Ethical Considerations The sentiment analysis used only publicly available data, and no direct human participants were involved in the research. Exemption from ethical review was granted by the Institute’s Human Ethics Committee (Project Number: 2022/HE002248). Data confidentiality was ensured by encrypting all datasets with secure passwords to prevent unauthorised access. Additionally, all personally identifiable information (PII), such as user identities, was anonymised through unique identifiers before analysis. This anonymisation process was automated to safeguard privacy, and the research team had exclusive access to the data. To further protect privacy, only essential data necessary for the sentiment analysis were collected, and any irrelevant PII was removed or obscured. These measures ensured adherence to data privacy standards throughout the study. Sentiment Analysis Stages The sentiment analysis process consisted of several stages: 1. Data Collection Publicly available posts from Facebook and Instagram were collected using Meta's Graph Application Programming Interface (API). Data was gathered from January 2014 to December 2023, capturing a wide range of discussions related to CWF. Data collected included text from posts and metadata such as creation time, likes, and shares. To ensure consistency in analysis, non-English posts were excluded, as the sentiment analysis tools used were optimised for English-language processing. The collected data were saved in Extensible Markup Language (XML) format for easier processing and analysis. 2. Data Source and Search Strategy Data were sourced through Meta’s Graph API, which allowed access to posts and comments adhering to specific search criteria. This provided the metadata needed to capture the range of discussions surrounding CWF. The search terms included keywords such as "#WaterFluoridation," "#CommunityWaterFluoridation," "#FluorideInWater," and similar variations, ensuring that discussions focused on the specific topic of CWF. 3. Data Processing The collected data underwent a thorough cleansing process. Python libraries removed extraneous elements such as URLs, user mentions, punctuation, emojis, and stopwords. This standardisation ensured that the data were ready for sentiment analysis. Since Facebook and Instagram posts are often informal and rich in context, refining the text for optimal processing by sentiment analysis models was essential. 4. Sentiment Analysis Framework The 'SentimentIntensityAnalyzer' from the Natural Language Toolkit (NLTK) was used for the analysis. This tool is well-suited for analysing the short text typical of social media posts and is reliable for differentiating between positive, negative, and neutral sentiments. The choice of NLTK was based on its ease of implementation, effectiveness with short, informal texts, and its proven reliability in similar studies on social media sentiment [ 9 , 10 ]. The NLTK framework enabled tasks such as tokenising text, tagging parts of speech, and recognising named entities, forming a comprehensive pipeline for sentiment analysis. The sentiment scores generated ranged from − 1 (highly negative) to 1 (highly positive), with neutral sentiments scored at 0 (Supplementary file 1). This allowed us to quantify the emotional tone of each post and aggregate these scores to determine overall sentiment trends related to CWF [ 11 , 8 ]. The ‘SentimentIntensityAnalyzer’ from NLTK was chosen due to its efficiency in handling social media text and high accuracy in sentiment scoring. Compared to other models like TextBlob and Stanford NLP (Natural Language Processing), it offers a more granular understanding of word context within sentences, essential for analysing the often informal, context-rich language used in Facebook and Instagram posts [ 20 , 21 ]. NLTK also offers highly validated user documentation and widespread use, making it an ideal tool for our sentiment analysis framework. Model Performance Testing To ensure the reliability of the sentiment analysis, model performance was tested against a manually annotated subset of the data [ 17 ]. This validation step helped refine the model's accuracy, ensuring it captured posts' sentiment about CWF. The final sentiment scores were then aggregated to identify the dominant public sentiments regarding CWF on Facebook and Instagram during the study period [ 17 – 21 ]. Factor-Based Data Analysis for Meta Platforms (Facebook and Instagram) Engagement Metrics Analyses To assess public interaction with CWF on Meta platforms (Facebook and Instagram), we implemented algorithms to analyse engagement metrics, including the time posts were created and the number of likes, comments, and shares they received. These metrics allowed us to determine the proportion of users discussing CWF over time, highlighting trends in public interest. Additionally, we analysed fluctuations in engagement levels to assess whether public interaction with CWF-related topics increased or decreased throughout the study period. The rise and fall of engagement metrics such as likes, comments, and shares were further examined to provide insights into patterns of public attention on CWF-related discussions [ 18 ]. Textual Sentiment Analysis Textual sentiment analysis included examining sentiment trends, word sentiment analysis, and word co-occurrence network analysis. The goal was to understand the overall sentiment toward CWF on Facebook and Instagram posts. Sentiment Trends and Patterns : We monitored general sentiment trends and changes in patterns to understand the consensus among Facebook and Instagram users regarding CWF. By tracking these trends, we identified fluctuations in public sentiment in response to critical events, such as news articles, policy changes, or public health campaigns. This analysis helped us determine whether public opinion was predominantly positive, negative, or neutral at different times. By studying how these sentiments evolved, we gained insights into changing public perceptions about CWF [ 17 – 19 ]. Word Sentiment Analysis : Word sentiment analysis identified posts' most frequently used positive and negative terms. After removing stopwords, we utilised a lexicon of pre-assigned sentiment scores for words and analysed the sentiment scores for terms in each post. The formula for normalised frequency was applied to investigate the most common sentiment-laden words used in posts each year: This analysis revealed the most used sentiment-driving terms in posts related to CWF on both platforms [ 22 ]. Word Co-occurrence Network Analysis We established a network of co-occurring terms by examining the most frequently used keywords in posts related to CWF. This network analysis provided insights into how different topics and key terms were connected, allowing us to identify the relationships between discussions surrounding CWF and other relevant issues. This method helped deepen our understanding of the core themes in public conversations about CWF on Meta platforms [ 10 , 11 ]. Evaluation of Model Performance The sentiment analysis of posts related to CWF on Facebook and Instagram was carried out using established machine learning models, including Logistic Regression, Decision Tree, Naive Bayes, and Random Forest. These models are well-regarded for their efficacy in text classification tasks. To evaluate the performance of these models, we used standard machine learning metrics such as accuracy, precision, recall, F1 score, and Area Under the Curve (AUC) [ 21 , 22 ]. Accuracy : Proportion of correctly classified instances out of the total cases. Precision : Proportion of true positive predictions among all positive predictions, indicating the accuracy of positive predictions. Recall : Proportion of true positive predictions among all actual positives, reflecting the model's ability to capture relevant instances. F1 Score : The harmonic mean of precision and recall, balancing both concerns. AUC : Represents the model's ability to distinguish between different classes, with a higher AUC indicating better model performance. These metrics allowed us to assess how well the models performed in classifying the sentiment expressed in Facebook and Instagram posts about CWF. Using multiple models ensured we could identify the most effective approach for analysing sentiment data from these platforms. Results Engagement Metrics Analyses A total of 109,117 posts related to community water fluoridation (CWF) discussions were collected from Meta platforms (Facebook and Instagram) from 2014 to 2023. Facebook contributed 73,938 (67.8%) posts, and Instagram provided 35,179 (32.2%). After data processing to remove duplicates, URL links, images, and other irrelevant content, 63,806 (58.5%) posts from Facebook and Instagram were retained for analysis. The post-count analysis showed fluctuations in public engagement over time (Fig. 1 A). From 2014 to 2017, post volume rose steadily, indicating growing interest in CWF, with peak engagement occurring between 2017 and 2019. Post activity declined gradually from 2020 onward, with moderate engagement levels of 3,000 to 5,000 posts from 2020 to 2023, suggesting a gradual decrease in public interest. Engagement patterns revealed that the LIKE reaction was most common, peaking in 2022, with a rise in SHARE reactions, indicating users preferred likes and shares over other forms of interaction (Fig. 1 B). Regarding post content, 42,789 (67.1%) posts contained sentences ranging from 100 to 300 characters, while 3,276 (5.1%) posts exceeded 500 characters, reflecting the social media preference for concise, easily digestible content. (Supplementary file 1). Textual Sentiment Analyses Sentiment Trends and Patterns: Of the posts regarding CWF, 42.1% conveyed a positive sentiment, 39.1% were negative, and 18.8% were neutral, indicating a mix of public support and opposition. Sentiment trends fluctuated notably over time. From 2014 to 2016, public opinion remained stable, with positive sentiments consistently outnumbering negative ones. In 2016, among 6,220 posts, 49.1% were positive, while 28.9% were negative, indicating a generally favourable stance on community water fluoridation (CWF). Negative sentiment spiked in 2017, peaking in 2018 when negative posts significantly outpaced positive ones due to adverse events (Fig. 2 ). By 2019, sentiment had further shifted, with 53.2% of posts negative and 33.6% positive out of 7,376 posts, reflecting rising controversy and misinformation. From 2020 to 2021, both positive and negative sentiments declined, though negative sentiment remained dominant, accounting for 47.0% and 49.6% of posts, respectively. Event markers suggest that discussions during this period were reignited by policy debates or misinformation. By 2022–2023, sentiment balance improved, with positive perceptions increasing. In 2023, among 4,404 posts, 48.0% were positive, while 33.5% were negative, signalling a more favourable outlook. However, engagement remained lower than peak years, suggesting waning public interest in the topic. Overall, public discussions on CWF have been mostly positive, although negative sentiments were slightly more prominent during 2017–2019. However, the intensity of negative reactions has recently declined while positive sentiments have gradually increased. Word Sentiment Analysis The analysis of favourable terms in posts about CWF on Meta platforms suggests social acceptance of fluoridation (Fig. 3 A). The most common term, "health," reflects the public's association of fluoridation with health benefits. Other frequently used terms include "public," "strengthen," and "advocate," indicating a focus on CWF’s broader societal health advantages. Words like "safe," "research," and "evidence" highlight scientific backing and safety, while terms such as "protection," "prevent," and "proven" convey its effectiveness as a preventive health measure. Additional terms like "community," "education," and "support" emphasise the importance of community benefits and educational advocacy. In contrast, negative terms used in posts expressing unfavourable views illustrate different concerns (Fig. 3 B). The term "toxic" is the most common, indicating fears of harmful effects from fluoridation. Other key terms such as "criticise," "waste," and "fluorosis" reflect worries about health risks, possible mismanagement, and side effects like dental fluorosis. Terms like "chemical," "harm," and "poison" reveal safety concerns related to chemical exposure, while "adverse," "unsafe," and "detrimental" express fears of adverse health impacts. Additionally, words like "controversy," "lobby," and "misinformation" suggest public distrust in fluoridation practices and information. Overall, positive discussions focus on health, safety, and community benefits, while negative discussions focus on health risks, chemical safety concerns, and policy criticisms. Word Co-occurrence Network Analysis : The network graph displays prominent terms and their interconnections in CWF discussions on Meta platforms. Node size and colour intensity indicate term frequency and centrality, with darker red nodes representing more frequent and strongly connected terms. The analysis reveals several main clusters in the discussion themes (Fig. 4 ). Central terms, including teeth, drinking, health, and poison, are highly interconnected, suggesting that concerns about health risks related to drinking water and dental health dominate the conversation. Filter and toxic are also closely linked with these core terms, underscoring public apprehension about contamination and the need for water filtration. Other prominent terms like chemical, government, cancer, and public connect with these central themes, highlighting distrust in government actions on fluoridation and fears of severe health effects, such as cancer. Terms like research and Harvard appear, indicating that academic studies from respected institutions are referenced in discussions. On the periphery, terms such as cavities, levels, community, and thyroid reflect additional concerns, particularly around dosage levels, thyroid impacts, and community health outcomes. In summary, the network graph shows that CWF discussions focus on health risks, drinking water concerns, and chemical safety, with frequent mentions of filtration and toxicity. Government trust and research are significant topics, reflecting public scepticism and ongoing debate. Model Performance Evaluation Table 1 presents the performance evaluation of various sentiment analysis models. The Logistic Regression model achieved the highest accuracy at 92.1% and an AUC of 0.9, making it the top performer for sentiment classification on Meta platforms. The Random Forest model followed with an accuracy of 88.4% and an AUC of 0.9, demonstrating strong reliability. The Decision Tree model showed moderate performance, with an accuracy of 87.% and an AUC of 0.8. The Naive Bayes model had the lowest performance, with an accuracy of 74.8% and an AUC of 0.78, suggesting it may be less suitable for this dataset. Table 1 Model Performance and Evaluation Details: This table presents the performance metrics of various machine learning models used for sentiment analysis of CWF-related posts. Logistic Regression achieved the highest accuracy (92.31%) and AUC value (0.95), making it the most effective model for this study. Other models, including Random Forest, Decision Tree, and Naive Bayes, demonstrated varying performance levels, highlighting the importance of model selection in text classification tasks. Model Accuracy (%) Precision (%) Recall (%) F1 Score (%) AUC Value Logistic Regression 92.31 92.35 92.31 92.32 0.95 Decision Tree 87.02 87.06 87.02 87.03 0.87 Naive Bayes 74.85 76.63 74.85 72.61 0.78 Random Forest 88.46 88.52 88.46 88.47 0.90 Discussion This study explores the trends in community water fluoridation (CWF) discussions on Meta platforms, examining how various factors, including scientific discussions, information dissemination, and policy developments shape public sentiment. Our findings underscore the influence of social media in shaping public health perceptions and highlight the importance of data-driven communication strategies to ensure accurate information dissemination. These observations align with broader patterns in public discussions on preventive health measures, such as vaccinations, where sentiment is constantly affected by misinformation exposure, media framing, and policy interventions [ 1 , 2 , 4 , 10 ]. We found that public opinion on CWF remains highly polarised, with 42.1% positive, 39.1% negative, and 18.8% neutral sentiments, suggesting that the topic remains a point of contention. While strong public support for fluoridation exists, safety concerns, distrust in governmental health policies, and misinformation narratives continue to spark opposition. These findings are consistent with previous studies demonstrating that health misinformation can significantly shift public sentiment, often more than evidence-based interventions alone [ 6 , 15 , 18 ]. The notable spike in negative sentiment between 2017 and 2019 aligns with significant policy changes and misinformation amplification, similar to vaccine hesitancy trends observed during high-profile anti-vaccine campaigns [ 10 , 11 , 14 , 17 ]. The Calgary fluoridation cessation case is a key example of how digital misinformation can directly influence policy decisions, reinforcing the need for early intervention and proactive health communication strategies [ 13 , 23 , 24 ]. Beyond the sentiment divide, neutral discussions present a critical opportunity for engagement. Research indicates that neutral perceptions are particularly susceptible to influence from emerging information, whether fact-based or misinformation-driven [ 14 , 25 , 26 ]. In the case of CWF, neutral discussions likely reflect a lack of access to clear, science-backed information or public uncertainty due to conflicting messages from different sources [ 10 , 14 , 27 ]. This underscores the importance of engaging neutral audiences with accessible, transparent, and community-centred health messaging before they become deeply embedded in negative narratives [ 28 ]. Additionally, the recent Cochrane systematic review on water fluoridation (2024) highlights that while CWF remains effective in reducing dental caries, the magnitude of its effect may be lower than earlier estimates, partly due to additional fluoride sources now widely available. Importantly, this review underscores the equity debate, noting that communities with less access to alternative fluoride interventions may benefit substantially from CWF. This resonates with our findings, where neutral sentiments—representing a potential ‘swing’ group—could be influenced by targeted messaging about protective and equitable aspects of fluoridation policies. Policymakers should thus consider the efficacy, fairness, and reach of CWF initiatives when designing public health strategies. The presence of emotionally charged language in negative discussions, mainly terms like "toxic," "harmful," and "chemical," indicates a broader public scepticism toward chemically mediated health interventions [ 5 , 11 , 29 ]. Like the anti-vaccine debate, these discussions are often fuelled by fear-based narratives rather than scientific evidence [ 16 , 17 , 19 , 22 ]. Prior research has shown that misinformation spreads faster than factual information, particularly when sensational language or emotionally compelling arguments are used [ 1 , 22 ]. The amplification of misinformation in CWF discussions suggests that the absence of early counter-narratives escalates scepticism, making real-time monitoring and misinformation debunking critical components of health communication strategies [ 10 , 11 , 14 ]. The role of machine learning in tracking sentiment and misinformation is increasingly evident in public health decision-making [ 29 , 30 ]. Our model evaluation demonstrated that Logistic Regression performed best (92.3% accuracy, AUC 0.9), followed by Random Forest (88.4% accuracy, AUC 0.9), while Naïve Bayes struggled to classify sentiment accurately (74.8% accuracy, AUC 0.7). The high performance of Logistic Regression and ensemble-based models like Random Forest indicates their potential for real-time monitoring of public sentiment trends, enabling health agencies to track misinformation patterns and adjust communication strategies accordingly [ 31 ]. These models can help detect shifts in public debates, allowing policymakers to intervene before misinformation narratives escalate [ 30 – 32 ]. However, future applications should explore deep-learning approaches, such as transformer-based models, to improve sentiment classification accuracy to interpret the sarcastic language used in social media texts [ 33 ]. Given the persistence of misinformation and public scepticism, addressing CWF-related discussions requires a multifaceted approach prioritising trust, engagement, and evidence accessibility [34,35]. Early detection of misinformation trends through real-time sentiment analysis and automated misinformation detection models can help counter false claims before they gain traction [ 2 , 8 , 14 ]. Additionally, proactive engagement strategies, such as fact-checking initiatives and targeted messaging that frames CWF in terms of community benefits rather than technical discussions, can improve public trust and reduce resistance [ 5 , 9 , 12 ]. Strengthening two-way communication through interactive discussions, Q&A sessions, and collaborations with trusted healthcare professionals could further enhance credibility and encourage community participation in fluoridation policies [ 4 , 7 , 16 ]. This study also has some limitations that warrant consideration. Temporal and demographic biases may have influenced sentiment trends, as changes in social media user demographics and regional events could have shaped online discussions. While we accounted for these biases by contextualising sentiment shifts with significant policy debates, future studies should incorporate demographic and geographic metadata to improve analytical precision. Additionally, this analysis was limited to English-language posts, which may exclude significant discussions in other languages. Since misinformation narratives vary across cultures and linguistic groups expanding sentiment analysis to multilingual datasets would provide a broader perspective on CWF discourse [ 19 ]. Moreover, while automated sentiment analysis is efficient, it may not fully capture human emotions and contextual details. Combining manual validation with AI-driven sentiment tools in future studies could enhance classification accuracy. Furthermore, while this study focused on Facebook and Instagram, different platforms attract distinct user demographics and engagement patterns. Expanding research to LinkedIn, TikTok, and YouTube could help uncover additional sentiment trends and misinformation patterns, providing a more holistic view of public opinion on CWF [ 20 , 36 ]. Another limitation is that we did not analyse the credibility of users posting CWF-related content—whether they were general users, key influencers, bots, or institutional accounts. Understanding the role of key opinion leaders and misinformation networks could explain how narratives spread and influence public perceptions of CWF. Overall, this study underscores the critical role of social media in shaping public sentiment and influencing health policy decisions regarding CWF. While digital platforms facilitate health advocacy and knowledge sharing, they also serve as hubs for misinformation, intensifying public scepticism and policy resistance. Addressing these challenges requires real-time monitoring, adaptive public health communication, and community-driven engagement. Integrating advanced sentiment analysis models into public health surveillance systems could improve early detection of misinformation trends, ensuring that public health messaging remains accurate, timely, and impactful. By adapting to the increasing digital health ecosystem, policymakers can strengthen evidence-based decision-making and reinforce public trust in fluoridation and other preventive health measures. Declarations Acknowledgements: The authors acknowledge the support provided by The University of Queensland. Funding Declaration : NT is supported by a University of Queensland’s Earmarked PhD scholarship provided to LD’s MRFF grant # 2024439. Author Contributions: N.T., L.D., R.L., and D.H. contributed to the concept and design of the study and supervised the research. N.T. conducted the data analysis, drafted the original manuscript, and managed data curation and visualisation. L.D., R.L., and D.H. provided validation, resources, and critical manuscript revisions. Project administration was managed by N.T., with funding acquisition supported by L.D., R.L., and D.H. All authors approved the final article and agreed to the submission. Conflict of Interest Statement: The authors declare no conflicts of interest related to this study. Data Availability Statement: The datasets generated and/or analysed during the current study are available from the corresponding author upon reasonable request. Ethics Approval Statement: The study used publicly available data from 'X' and did not involve human participants. Ethics approval was obtained from The University of Queensland’s Human Ethics Committee (Approval Number: 2022/HE002248). Patient Consent Statement: Not applicable, as this study did not involve direct interaction with patients. Permission to Reproduce Material: Not applicable, as no previously published materials requiring permissions were used. Clinical Trial Registration: Not applicable. References Vosoughi S, Roy D, Aral S. The spread of true and false news online. Science. 2018;359(6380):1146–51. Cinelli M, Quattrociocchi W, Galeazzi A, et al. The COVID-19 social media infodemic. Sci Rep. 2020;10(1):16598. Statista. (2021). Number of monthly active Facebook users worldwide. Available at: [ https://www.statista.com](https://www.statista.com) Merchant RM, Lurie N. Social media and emergency preparedness in response to novel coronavirus. JAMA. 2020;323(20):2011–2. Gabarron E, Wynn R. Use of social media for sexual health promotion: A scoping review. Global Health Action. 2016;9(1):32193. Larson HJ, et al. The state of vaccine confidence 2016: global insights through a 67-country survey. EBioMedicine. 2016;12:295–301. George DR, Rovniak LS, Kraschnewski JL. Dangers and opportunities for social media in medicine. Clin Obstet Gynecol. 2013;56(3):453–62. Roozenbeek J, van der Linden S. Fake news game confers psychological resistance against online misinformation. Palgrave Commun. 2020;6:65. Keim-Malpass J, Mitchell EM, Sun E, Kennedy C. Using Twitter to understand public perceptions regarding the #HPV vaccine: Opportunities for public health nurses to engage in social marketing. Public Health Nurs. 2017;34(4):316–23. Lotto M, Menezes TS, Zakir Hussain I, Tsao S, Butt ZA, Morita PP, et al. Characterization of false or misleading fluoride content on Instagram: infodemiology study. J Med Internet Res. 2022;24(5):e37519. [FREE Full text] [doi: 10.2196/37519] [Medline: 35588055]. Mertz A, Allukian M. Community water fluoridation on the Internet and social media. J Mass Dent Soc. 2014 Summer;63(2):32–6. PMID: 25230407. Whelton HP, Spencer AJ, Do LG, Rugg-Gunn AJ. Fluoride revolution and dental caries: evolution of policies for global use. J Dent Res. 2019;98(8):837–46. [doi: 10.1177/0022034519843495] [Medline: 31282846]. O'Neill B, Kapoor T, McLaren L. Politics, Science, and Termination: A Case Study of Water Fluoridation Policy in Calgary in 2011. Rev Policy Res. 2018;00:1. https://doi.org/10.1111/ropr.12318 . Oh O, Agrawal M, Rao HR. Community water fluoridation and online discourse: How misinformation spreads on social media. J Public Health Dent. 2019;79(2):101–7. Jones SP, et al. A social media study of anti-vaccination commentary and information sharing: The reach and influence of anti-vaccination opinions. J Community Health. 2019;44(6):1020–32. Black DR, Dietz JE, Stirratt AA, Coster DC. Do social media have a place in public health emergency response? J Emerg Manag. 2015;13(3):217–26. Chou WS, Oh A, Klein WMP. Addressing Health-Related Misinformation on Social Media. JAMA. 2018;320(23):2417–2418. 10.1001/jama.2018.16865 . PMID: 30428002. Oh HJ, Kim CH, Jeon JG. Public Sense of Water Fluoridation as Reflected on Twitter 2009–2017. J Dent Res. 2020;99(1):11–17. 10.1177/0022034519885610 . Epub 2019 Nov 4. PMID: 31682777. Cavazos-Rehg PA, Krauss M, Fisher SL, Salyer P, Grucza RA, Bierut LJ. Twitter chatter about marijuana. J Adolesc Health. 2015;56(2):139–45. Eysenbach G. Infodemiology and infoveillance tracking online health information and cyberbehavior for public health. Am J Prev Med. 2011;40(5 Suppl 2):S154–8. Umunyana C, Tuyizere G, Mbarushimana A. 2024. Students’ Sentiment Analysis Using Natural Language Toolkit in Machine Learning for Module Evaluation. European Journal of Engineering and Technology Research . 9, 1 (Mar. 2024), 72–75. https://doi.org/10.24018/ejeng.2024.9.1.3006 Boon-Itt S, Skunkan Y. Public perception of the COVID-19 pandemic on Twitter: sentiment analysis and topic modelling Study. JMIR Public Health Surveill. 2020;6(4):e21978. [FREE Full text] [doi: 10.2196/21978] [Medline: 33108310]. McLaren L, Patterson S, Thawer S, et al. Measuring the short-term impact of fluoridation cessation on dental caries in grade 2 children using tooth surface indices. Community Dent Oral Epidemiol. 2016;44(3):274–82. 10.1111/cdoe.12216 . City of Calgary. 2021 municipal election results. City of Calgary website. Published 2021. Accessed August 10. 2024. Available from: https://www.calgary.ca/election/results/2021-results.html McLaren L, Singhal S. Does cessation of community water fluoridation lead to an increase in tooth decay? A systematic review of published studies. J Epidemiol Community Health. 2016;70(9):934–40. 10.1136/jech-2015-206402 . Taddi VV, Kohli RK, Puri P. Perception, use of social media, and its impact on the mental health of Indian adolescents: A qualitative study. World J Clin Pediatr. 2024;13(3):97501. 10.5409/wjcp.v13.i3.97501 . PMID: 39350908; PMCID: PMC11438920. Armfield JM, Akers HF. Risk perception and water fluoridation support and opposition in Australia. J Public Health Dent. 2010 Winter;70(1):58–66. 10.1111/j.1752-7325.2009.00144.x . PMID: 19694932. Iheozor-Ejiofor Z, Walsh T, Lewis SR, Riley P, Boyers D, Clarkson JE, Worthington HV, Glenny A-M, O'Malley L. Water fluoridation for the prevention of dental caries. Cochrane Database of Systematic Reviews 2024, Issue 10. Art. No.: CD010856. 10.1002/14651858.CD010856.pub3 . Accessed 03 March 2025. James G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning: with Applications in R. New York: Springer; 2013. Sebastiani F. Machine learning in automated text categorization. ACM Comput Surv. 2002;34(1):1–47. Breiman L. Random forests. Mach Learn. 2001;45(1):5–32. Quinlan JR. Induction of Decision Trees. Mach Learn. 1986;1(1):81–106. Khan S, Qasim I, Khan W, Aurangzeb K, Khan JA, Anwar MS. (2025). A novel transformer attention-based approach for sarcasm detection. Expert Systems , 42(1), e13686. https://doi.org/10.1111/exsy.1368634 . Mackert M, Bouchacourt L, Lazard A, Wilcox GB, Kemp D, Kahlor LA, George C, Stewart B, Wolfe J. Social media conversations about community water fluoridation: formative research to guide health communication. J Public Health Dent. 2021;81(2):162–166. doi: 10.1111/jphd.12404. Epub 2020 Oct 15. PMID: 33058200. Sabine Meier C, Stock A, Krämer. The contribution of health discussion groups with students to campus health promotion, Health Promotion International , Volume 22, Issue 1, March 2007, Pages 28–36. https://doi.org/10.1093/heapro/dal041 Dredze M. How Social Media Will Change Public Health. IEEE Intell Syst. 2012;27(4):81–4. 10.1109/MIS.2012.74.ss . Additional Declarations No competing interests reported. Supplementary Files Supplementaryfiles.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6266081","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":433546052,"identity":"eaee6e94-8e18-4ec3-9f5f-61d6e72c472e","order_by":0,"name":"Nilesh Torwane","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYDADfjDJRpxixgYQKdlAshaDA8RqMTiee/zBxz2H5Y1vJG9g+FB2mIF/RgIBLWfeJTbOeHbYcNuNtALGGecOM0jcIKTlRo5hM8+B24zbbuQYMPO2HWZgIFaL/eYZQC1/gVrkidWSuEECqIURqMWAkBbJM28MZ8448D95xplnBQd7zqXzGJ55gF8L3/Ecgw8fDqTZ9rcnb3zwo8xaTu44AVsYGBAKwFHDQ0g9qhYiVI+CUTAKRsFIBAApxUxKk1p31QAAAABJRU5ErkJggg==","orcid":"","institution":"The University of Queensland","correspondingAuthor":true,"prefix":"","firstName":"Nilesh","middleName":"","lastName":"Torwane","suffix":""},{"id":433546053,"identity":"5f2b9172-17df-4ea3-ad4d-1648ed1b7e12","order_by":1,"name":"Ratilal Lalloo","email":"","orcid":"","institution":"The University of Queensland","correspondingAuthor":false,"prefix":"","firstName":"Ratilal","middleName":"","lastName":"Lalloo","suffix":""},{"id":433546054,"identity":"5bde95f4-aa0d-4217-b835-175384d086d1","order_by":2,"name":"Diep H. Ha","email":"","orcid":"","institution":"The University of Queensland","correspondingAuthor":false,"prefix":"","firstName":"Diep","middleName":"H.","lastName":"Ha","suffix":""},{"id":433546055,"identity":"0b6f6848-0e4f-4123-9461-422bb6044d49","order_by":3,"name":"Loc G. Do","email":"","orcid":"","institution":"The University of Queensland","correspondingAuthor":false,"prefix":"","firstName":"Loc","middleName":"G.","lastName":"Do","suffix":""}],"badges":[],"createdAt":"2025-03-20 04:53:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6266081/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6266081/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":79237140,"identity":"539fe271-5d35-4847-b627-4a08627174fc","added_by":"auto","created_at":"2025-03-26 04:26:57","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":425920,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA- \u003c/strong\u003eYearly Distribution of CWF-Related Posts. This bar chart displays the yearly volume of CWF-related posts on Meta platforms (Facebook and Instagram) from 2014 to 2023. The data show a steady increase in post volume from 2014, peaking in 2017 (8,693 posts), followed by a gradual decline from 2021 onward. The lowest engagement was recorded in 2021 (3,115 posts), with a slight increase in activity in 2023 (4,404 posts).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB\u003c/strong\u003e- Engagement Activities for CWF-Related Posts. This line graph illustrates engagement activities (likes, shares, and other reactions) related to CWF posts on Meta platforms (Facebook and Instagram) from 2014 to 2023. The data show a steady rise in engagement over time, with \"like\" and \"share\" activities being the most dominant interactions. A notable spike in engagement is observed in 2022-2023, particularly in likes, indicating increased public interaction with CWF-related content during this period.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6266081/v1/746602ca73132c57b06eb9bc.png"},{"id":79237335,"identity":"a546f35d-0b3e-4811-ad91-a779cb1cc777","added_by":"auto","created_at":"2025-03-26 04:34:57","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":263845,"visible":true,"origin":"","legend":"\u003cp\u003eSentiment Trends in CWF-Related Posts Over Time. This chart illustrates the yearly distribution of positive and negative sentiments in community water fluoridation (CWF)-related posts from 2014 to 2023. The bars represent the number of positive and negative posts per year, while the dashed lines indicate the percentage trends of each sentiment. The data show a significant increase in negative sentiment from 2017 to 2019, aligning with key events marked as \"Event Impact.\" After 2019, both positive and negative sentiments declined, with negative sentiment leading slightly during 2020-2021. However, in 2022-2023, positive sentiment rebounded, suggesting a shift toward a more balanced public perception of CWF.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6266081/v1/604284c36165cc85d0b6b1d6.png"},{"id":79237141,"identity":"1ba611d1-ec54-4b6a-a0b9-65fd20e7f4d7","added_by":"auto","created_at":"2025-03-26 04:26:57","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1027360,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA\u003c/strong\u003e- Top Positive (Favourable) Words in CWF-Related Posts: This radial plot visualises the most frequently used positive terms in community water fluoridation (CWF)-related posts on Meta platforms. Prominent terms like \"health,\" \"public,\" and \"strengthen\" emphasise the perceived health benefits, societal advantages, and advocacy for CWF.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eB\u003c/strong\u003e - Top Negative (Unfavourable) Words in CWF-Related Posts: This radial plot highlights the most frequently used negative terms in CWF-related posts on Meta platforms. Words such as \"toxic,\" \"harm,\" and \"chemical\" underscore safety concerns and scepticism surrounding CWF, reflecting public apprehensions about its risks and effects.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6266081/v1/6cfeea479a6048b2791e2980.png"},{"id":79237336,"identity":"e3e17fa8-50b9-46d8-8777-3a23509902e1","added_by":"auto","created_at":"2025-03-26 04:34:57","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1218919,"visible":true,"origin":"","legend":"\u003cp\u003eWord Co-Occurrence Network Analysis of CWF Discussions on Meta Platforms: This network graph shows the relationships between frequently used terms in CWF-related posts. Nodes represent key terms, with their size and color intensity indicating frequency and connectivity. Central terms like \"health,\" \"drinking,\" and \"poison\" dominate the discussion, highlighting concerns about dental health, water safety, and chemical risks. Peripheral nodes, such as \"thyroid\" and \"community,\" represent additional discussion themes, showcasing the multifaceted nature of public discourse on CWF.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6266081/v1/566bf0af6f3099b2f472a796.jpg"},{"id":79238451,"identity":"1403d6c3-d61f-40ee-acbd-dc360cb0a8f8","added_by":"auto","created_at":"2025-03-26 04:59:02","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3382286,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6266081/v1/c5cad109-6ce1-47fd-998e-e7b0752c9984.pdf"},{"id":79237154,"identity":"85fe904e-2095-4c6f-bc15-3cba69792f69","added_by":"auto","created_at":"2025-03-26 04:26:57","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":90909,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfiles.docx","url":"https://assets-eu.researchsquare.com/files/rs-6266081/v1/0e0afd7cfa256462a4fe0d8f.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Cracking the 'Meta' Code: Advanced Machine Learning-Based Sentiment Analysis of Water Fluoridation Debates on Facebook and Instagram","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIn today\u0026rsquo;s digital age, social media platforms like Facebook and Instagram, both part of the Meta family, play a significant role in shaping public opinion on social, political, scientific, and health-related issues [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. These platforms are major sources of information, sharing news updates and spreading misinformation [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. This makes them highly influential in forming public perceptions [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. With a vast reach\u0026mdash;Facebook alone has over 2.8\u0026nbsp;billion monthly active users\u0026mdash;these platforms engage millions of people in health discussions, which increases their impact on public health [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The broad reach and high interaction levels make Facebook and Instagram powerful in influencing people\u0026rsquo;s beliefs and behaviours about health.\u003c/p\u003e \u003cp\u003eOne primary advantage of analysing sentiment on these platforms is the large amount and variety of user-generated content, which offers valuable data to understand public opinion across different groups and locations [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Facebook and Instagram encourage communities where people can share their experiences and views on health, even on controversial topics [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. For example, studies have shown that misinformation about vaccines in Facebook groups has contributed to vaccine hesitancy and mistrust, showing the platform\u0026rsquo;s influence on health behaviours [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Similarly, parents discussing teething pain management often recommend unverified remedies, like amber necklaces, for relief, even though there is limited scientific support [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. On Instagram, where visual content is popular, people share opinions through images and videos. This adds another layer to public opinion analysis but also spreads misinformation on topics like drug use, diet, and vaccines [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe popularity of unproven health products in online communities shows how these platforms shape health perceptions, often without scientific backing. This can create risks from unchecked information sharing [\u003cspan additionalcitationids=\"CR8\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. A similar pattern is seen with Community Water Fluoridation (CWF), a well-established public health intervention that reduces dental cavities by adjusting fluoride in water supplies. Although CWF is supported by major health organisations like the World Health Organization (WHO), the Centers for Disease Control and Prevention (CDC), and the Australian National Health and Medical Research Council (NHMRC), it remains a polarising issue, with opposition fuelled by misinformation on social media [\u003cspan additionalcitationids=\"CR11 CR12 CR13\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. As with teething remedies, social media amplifies questionable studies and unverified claims, making it harder to maintain public support for evidence-based health practices like CWF [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Although widely recognised as safe and effective, CWF faces challenges due to the spread of misinformation, which undermines public trust in this vital intervention.\u003c/p\u003e \u003cp\u003eTherefore, analysing sentiment on Facebook and Instagram offers a valuable way to track how public opinions on CWF form and evolve. Unlike traditional surveys, which capture only a single point, social media provides real-time data, monitoring public attitudes continuously as they react to new information or misinformation [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Facebook\u0026rsquo;s recommendation system, which suggests content based on user interactions, offers insights into how CWF-related information spreads in specific communities. Similarly, Instagram\u0026rsquo;s use of hashtags allows for tracking discussions and sentiment trends [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. These tools are essential for identifying recurring misinformation trends and understanding how they affect public health communication.\u003c/p\u003e \u003cp\u003eThis study aimed to conduct a detailed sentiment analysis of Facebook and Instagram (Meta platforms) to understand public views on CWF. By exploring what drives these opinions, the research aims to provide findings that can guide public health communication strategies and inform policymakers. Ultimately, this study contributes to a deeper understanding of how social media affects public health perceptions, highlighting the importance of evidence-based policies and supporting better health outcomes related to CWF.\u003c/p\u003e"},{"header":"Material and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eScope of the Study\u003c/h2\u003e\n \u003cp\u003eThis sentiment analysis investigation used posts from Meta-based platforms like Facebook and Instagram to understand public sentiment regarding CWF. Our research spanned January 2014 to December 2023, using various parameters such as textual sentiment analysis, keyword analysis, topic modelling, and model performance testing. These 10 years were selected to capture a broad spectrum of discussions and shifts in public opinion over time, ensuring a comprehensive view of sentiment trends. The approach is aligned with the methodologies recommended in previous research on sentiment analysis in health communication and misinformation [\u003cspan class=\"CitationRef\"\u003e16\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e\n \u003cp\u003eFacebook and Instagram were merged for this analysis because both platforms operate under the Meta company, sharing similar algorithms, data collection methodologies, and user interaction capabilities. As Meta-owned platforms, Facebook and Instagram allow users to seamlessly share posts, stories, and media across both networks. This cross-platform sharing provides unified access to public discussions and interactions, allowing for a more comprehensive dataset that reflects user behaviour across these integrated platforms. Merging the data enables capturing broader interactions and opinions on CWF while ensuring consistency in data collection, processing, and analysis [\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e]. Given their shared infrastructure, combining data from both platforms helps improve the validity and scope of the analysis by providing a fuller picture of sentiment across these widely used platforms.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eData Confidentiality and Ethical Considerations\u003c/h3\u003e\n\u003cp\u003eThe sentiment analysis used only publicly available data, and no direct human participants were involved in the research. Exemption from ethical review was granted by the Institute\u0026rsquo;s Human Ethics Committee (Project Number: 2022/HE002248). Data confidentiality was ensured by encrypting all datasets with secure passwords to prevent unauthorised access. Additionally, all personally identifiable information (PII), such as user identities, was anonymised through unique identifiers before analysis. This anonymisation process was automated to safeguard privacy, and the research team had exclusive access to the data. To further protect privacy, only essential data necessary for the sentiment analysis were collected, and any irrelevant PII was removed or obscured. These measures ensured adherence to data privacy standards throughout the study.\u003c/p\u003e\n\u003ch3\u003eSentiment Analysis Stages\u003c/h3\u003e\n\u003cp\u003eThe sentiment analysis process consisted of several stages:\u003c/p\u003e\n\u003ch3\u003e1. Data Collection\u003c/h3\u003e\n\u003cp\u003ePublicly available posts from Facebook and Instagram were collected using Meta\u0026apos;s Graph Application Programming Interface (API). Data was gathered from January 2014 to December 2023, capturing a wide range of discussions related to CWF. Data collected included text from posts and metadata such as creation time, likes, and shares. To ensure consistency in analysis, non-English posts were excluded, as the sentiment analysis tools used were optimised for English-language processing. The collected data were saved in Extensible Markup Language (XML) format for easier processing and analysis.\u003c/p\u003e\n\u003ch3\u003e2. Data Source and Search Strategy\u003c/h3\u003e\n\u003cp\u003eData were sourced through Meta\u0026rsquo;s Graph API, which allowed access to posts and comments adhering to specific search criteria. This provided the metadata needed to capture the range of discussions surrounding CWF. The search terms included keywords such as \u0026quot;#WaterFluoridation,\u0026quot; \u0026quot;#CommunityWaterFluoridation,\u0026quot; \u0026quot;#FluorideInWater,\u0026quot; and similar variations, ensuring that discussions focused on the specific topic of CWF.\u003c/p\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e3. Data Processing\u003c/h2\u003e\n \u003cp\u003eThe collected data underwent a thorough cleansing process. Python libraries removed extraneous elements such as URLs, user mentions, punctuation, emojis, and stopwords. This standardisation ensured that the data were ready for sentiment analysis. Since Facebook and Instagram posts are often informal and rich in context, refining the text for optimal processing by sentiment analysis models was essential.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003e4. Sentiment Analysis Framework\u003c/h3\u003e\n\u003cp\u003eThe \u0026apos;SentimentIntensityAnalyzer\u0026apos; from the Natural Language Toolkit (NLTK) was used for the analysis. This tool is well-suited for analysing the short text typical of social media posts and is reliable for differentiating between positive, negative, and neutral sentiments. The choice of NLTK was based on its ease of implementation, effectiveness with short, informal texts, and its proven reliability in similar studies on social media sentiment [\u003cspan class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe NLTK framework enabled tasks such as tokenising text, tagging parts of speech, and recognising named entities, forming a comprehensive pipeline for sentiment analysis. The sentiment scores generated ranged from \u0026minus;\u0026thinsp;1 (highly negative) to 1 (highly positive), with neutral sentiments scored at 0 (Supplementary file 1). This allowed us to quantify the emotional tone of each post and aggregate these scores to determine overall sentiment trends related to CWF [\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe \u0026lsquo;SentimentIntensityAnalyzer\u0026rsquo; from NLTK was chosen due to its efficiency in handling social media text and high accuracy in sentiment scoring. Compared to other models like TextBlob and Stanford NLP (Natural Language Processing), it offers a more granular understanding of word context within sentences, essential for analysing the often informal, context-rich language used in Facebook and Instagram posts [\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e]. NLTK also offers highly validated user documentation and widespread use, making it an ideal tool for our sentiment analysis framework.\u003c/p\u003e\n\u003ch3\u003eModel Performance Testing\u003c/h3\u003e\n\u003cp\u003eTo ensure the reliability of the sentiment analysis, model performance was tested against a manually annotated subset of the data [\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e]. This validation step helped refine the model\u0026apos;s accuracy, ensuring it captured posts\u0026apos; sentiment about CWF. The final sentiment scores were then aggregated to identify the dominant public sentiments regarding CWF on Facebook and Instagram during the study period [\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eFactor-Based Data Analysis for Meta Platforms (Facebook and Instagram)\u003c/h2\u003e\n \u003cdiv id=\"Sec12\" class=\"Section3\"\u003e\n \u003ch2\u003eEngagement Metrics Analyses\u003c/h2\u003e\n \u003cp\u003eTo assess public interaction with CWF on Meta platforms (Facebook and Instagram), we implemented algorithms to analyse engagement metrics, including the time posts were created and the number of likes, comments, and shares they received. These metrics allowed us to determine the proportion of users discussing CWF over time, highlighting trends in public interest. Additionally, we analysed fluctuations in engagement levels to assess whether public interaction with CWF-related topics increased or decreased throughout the study period. The rise and fall of engagement metrics such as likes, comments, and shares were further examined to provide insights into patterns of public attention on CWF-related discussions [\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eTextual Sentiment Analysis\u003c/h2\u003e\n \u003cp\u003eTextual sentiment analysis included examining sentiment trends, word sentiment analysis, and word co-occurrence network analysis. The goal was to understand the overall sentiment toward CWF on Facebook and Instagram posts.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e\u003cstrong\u003eSentiment Trends and Patterns\u003c/strong\u003e:\u003c/h2\u003e\n \u003cp\u003eWe monitored general sentiment trends and changes in patterns to understand the consensus among Facebook and Instagram users regarding CWF. By tracking these trends, we identified fluctuations in public sentiment in response to critical events, such as news articles, policy changes, or public health campaigns. This analysis helped us determine whether public opinion was predominantly positive, negative, or neutral at different times. By studying how these sentiments evolved, we gained insights into changing public perceptions about CWF [\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e\u003cstrong\u003eWord Sentiment Analysis\u003c/strong\u003e:\u003c/h2\u003e\n \u003cp\u003eWord sentiment analysis identified posts\u0026apos; most frequently used positive and negative terms. After removing stopwords, we utilised a lexicon of pre-assigned sentiment scores for words and analysed the sentiment scores for terms in each post. The formula for normalised frequency was applied to investigate the most common sentiment-laden words used in posts each year:\u003c/p\u003e\n \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" style=\"width: 656px;\"\u003e\u003c/p\u003e\n \u003cp\u003eThis analysis revealed the most used sentiment-driving terms in posts related to CWF on both platforms [\u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eWord Co-occurrence Network Analysis\u003c/h2\u003e\n \u003cp\u003eWe established a network of co-occurring terms by examining the most frequently used keywords in posts related to CWF. This network analysis provided insights into how different topics and key terms were connected, allowing us to identify the relationships between discussions surrounding CWF and other relevant issues. This method helped deepen our understanding of the core themes in public conversations about CWF on Meta platforms [\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eEvaluation of Model Performance\u003c/h2\u003e\n \u003cp\u003eThe sentiment analysis of posts related to CWF on Facebook and Instagram was carried out using established machine learning models, including Logistic Regression, Decision Tree, Naive Bayes, and Random Forest. These models are well-regarded for their efficacy in text classification tasks. To evaluate the performance of these models, we used standard machine learning metrics such as accuracy, precision, recall, F1 score, and Area Under the Curve (AUC) [\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eAccuracy\u003c/strong\u003e: Proportion of correctly classified instances out of the total cases.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003ePrecision\u003c/strong\u003e: Proportion of true positive predictions among all positive predictions, indicating the accuracy of positive predictions.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eRecall\u003c/strong\u003e: Proportion of true positive predictions among all actual positives, reflecting the model\u0026apos;s ability to capture relevant instances.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eF1 Score\u003c/strong\u003e: The harmonic mean of precision and recall, balancing both concerns.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e\u003cstrong\u003eAUC\u003c/strong\u003e: Represents the model\u0026apos;s ability to distinguish between different classes, with a higher AUC indicating better model performance.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cp\u003eThese metrics allowed us to assess how well the models performed in classifying the sentiment expressed in Facebook and Instagram posts about CWF. Using multiple models ensured we could identify the most effective approach for analysing sentiment data from these platforms.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eEngagement Metrics Analyses\u003c/h2\u003e \u003cp\u003eA total of 109,117 posts related to community water fluoridation (CWF) discussions were collected from Meta platforms (Facebook and Instagram) from 2014 to 2023. Facebook contributed 73,938 (67.8%) posts, and Instagram provided 35,179 (32.2%). After data processing to remove duplicates, URL links, images, and other irrelevant content, 63,806 (58.5%) posts from Facebook and Instagram were retained for analysis.\u003c/p\u003e \u003cp\u003eThe post-count analysis showed fluctuations in public engagement over time (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). From 2014 to 2017, post volume rose steadily, indicating growing interest in CWF, with peak engagement occurring between 2017 and 2019. Post activity declined gradually from 2020 onward, with moderate engagement levels of 3,000 to 5,000 posts from 2020 to 2023, suggesting a gradual decrease in public interest.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eEngagement patterns revealed that the LIKE reaction was most common, peaking in 2022, with a rise in SHARE reactions, indicating users preferred likes and shares over other forms of interaction (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Regarding post content, 42,789 (67.1%) posts contained sentences ranging from 100 to 300 characters, while 3,276 (5.1%) posts exceeded 500 characters, reflecting the social media preference for concise, easily digestible content. (Supplementary file 1).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eTextual Sentiment Analyses\u003c/h2\u003e \u003cdiv id=\"Sec21\" class=\"Section3\"\u003e \u003ch2\u003eSentiment Trends and Patterns:\u003c/h2\u003e \u003cp\u003eOf the posts regarding CWF, 42.1% conveyed a positive sentiment, 39.1% were negative, and 18.8% were neutral, indicating a mix of public support and opposition.\u003c/p\u003e \u003cp\u003eSentiment trends fluctuated notably over time. From 2014 to 2016, public opinion remained stable, with positive sentiments consistently outnumbering negative ones. In 2016, among 6,220 posts, 49.1% were positive, while 28.9% were negative, indicating a generally favourable stance on community water fluoridation (CWF).\u003c/p\u003e \u003cp\u003eNegative sentiment spiked in 2017, peaking in 2018 when negative posts significantly outpaced positive ones due to adverse events (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e). By 2019, sentiment had further shifted, with 53.2% of posts negative and 33.6% positive out of 7,376 posts, reflecting rising controversy and misinformation.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFrom 2020 to 2021, both positive and negative sentiments declined, though negative sentiment remained dominant, accounting for 47.0% and 49.6% of posts, respectively. Event markers suggest that discussions during this period were reignited by policy debates or misinformation. By 2022\u0026ndash;2023, sentiment balance improved, with positive perceptions increasing. In 2023, among 4,404 posts, 48.0% were positive, while 33.5% were negative, signalling a more favourable outlook. However, engagement remained lower than peak years, suggesting waning public interest in the topic.\u003c/p\u003e \u003cp\u003eOverall, public discussions on CWF have been mostly positive, although negative sentiments were slightly more prominent during 2017\u0026ndash;2019. However, the intensity of negative reactions has recently declined while positive sentiments have gradually increased.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eWord Sentiment Analysis\u003c/h2\u003e \u003cp\u003eThe analysis of favourable terms in posts about CWF on Meta platforms suggests social acceptance of fluoridation (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). The most common term, \"health,\" reflects the public's association of fluoridation with health benefits. Other frequently used terms include \"public,\" \"strengthen,\" and \"advocate,\" indicating a focus on CWF\u0026rsquo;s broader societal health advantages. Words like \"safe,\" \"research,\" and \"evidence\" highlight scientific backing and safety, while terms such as \"protection,\" \"prevent,\" and \"proven\" convey its effectiveness as a preventive health measure. Additional terms like \"community,\" \"education,\" and \"support\" emphasise the importance of community benefits and educational advocacy.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn contrast, negative terms used in posts expressing unfavourable views illustrate different concerns (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). The term \"toxic\" is the most common, indicating fears of harmful effects from fluoridation. Other key terms such as \"criticise,\" \"waste,\" and \"fluorosis\" reflect worries about health risks, possible mismanagement, and side effects like dental fluorosis. Terms like \"chemical,\" \"harm,\" and \"poison\" reveal safety concerns related to chemical exposure, while \"adverse,\" \"unsafe,\" and \"detrimental\" express fears of adverse health impacts. Additionally, words like \"controversy,\" \"lobby,\" and \"misinformation\" suggest public distrust in fluoridation practices and information.\u003c/p\u003e \u003cp\u003eOverall, positive discussions focus on health, safety, and community benefits, while negative discussions focus on health risks, chemical safety concerns, and policy criticisms.\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003eWord Co-occurrence Network Analysis\u003c/span\u003e:\u003c/h2\u003e \u003cp\u003eThe network graph displays prominent terms and their interconnections in CWF discussions on Meta platforms. Node size and colour intensity indicate term frequency and centrality, with darker red nodes representing more frequent and strongly connected terms. The analysis reveals several main clusters in the discussion themes (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eCentral terms, including teeth, drinking, health, and poison, are highly interconnected, suggesting that concerns about health risks related to drinking water and dental health dominate the conversation. Filter and toxic are also closely linked with these core terms, underscoring public apprehension about contamination and the need for water filtration.\u003c/p\u003e \u003cp\u003eOther prominent terms like chemical, government, cancer, and public connect with these central themes, highlighting distrust in government actions on fluoridation and fears of severe health effects, such as cancer. Terms like research and Harvard appear, indicating that academic studies from respected institutions are referenced in discussions.\u003c/p\u003e \u003cp\u003eOn the periphery, terms such as cavities, levels, community, and thyroid reflect additional concerns, particularly around dosage levels, thyroid impacts, and community health outcomes.\u003c/p\u003e \u003cp\u003eIn summary, the network graph shows that CWF discussions focus on health risks, drinking water concerns, and chemical safety, with frequent mentions of filtration and toxicity. Government trust and research are significant topics, reflecting public scepticism and ongoing debate.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eModel Performance Evaluation\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the performance evaluation of various sentiment analysis models. The Logistic Regression model achieved the highest accuracy at 92.1% and an AUC of 0.9, making it the top performer for sentiment classification on Meta platforms. The Random Forest model followed with an accuracy of 88.4% and an AUC of 0.9, demonstrating strong reliability. The Decision Tree model showed moderate performance, with an accuracy of 87.% and an AUC of 0.8. The Naive Bayes model had the lowest performance, with an accuracy of 74.8% and an AUC of 0.78, suggesting it may be less suitable for this dataset.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eModel Performance and Evaluation Details: This table presents the performance metrics of various machine learning models used for sentiment analysis of CWF-related posts. Logistic Regression achieved the highest accuracy (92.31%) and AUC value (0.95), making it the most effective model for this study. Other models, including Random Forest, Decision Tree, and Naive Bayes, demonstrated varying performance levels, highlighting the importance of model selection in text classification tasks.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAccuracy (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePrecision (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRecall (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eF1 Score (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUC Value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLogistic Regression\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e92.31\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e92.35\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e92.31\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e92.32\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDecision Tree\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e87.02\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e87.06\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e87.02\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e87.03\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.87\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eNaive Bayes\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e74.85\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e76.63\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e74.85\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e72.61\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.78\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRandom Forest\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e88.46\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e88.52\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e88.46\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e88.47\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e\u003cb\u003e0.90\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study explores the trends in community water fluoridation (CWF) discussions on Meta platforms, examining how various factors, including scientific discussions, information dissemination, and policy developments shape public sentiment. Our findings underscore the influence of social media in shaping public health perceptions and highlight the importance of data-driven communication strategies to ensure accurate information dissemination. These observations align with broader patterns in public discussions on preventive health measures, such as vaccinations, where sentiment is constantly affected by misinformation exposure, media framing, and policy interventions [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWe found that public opinion on CWF remains highly polarised, with 42.1% positive, 39.1% negative, and 18.8% neutral sentiments, suggesting that the topic remains a point of contention. While strong public support for fluoridation exists, safety concerns, distrust in governmental health policies, and misinformation narratives continue to spark opposition. These findings are consistent with previous studies demonstrating that health misinformation can significantly shift public sentiment, often more than evidence-based interventions alone [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. The notable spike in negative sentiment between 2017 and 2019 aligns with significant policy changes and misinformation amplification, similar to vaccine hesitancy trends observed during high-profile anti-vaccine campaigns [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. The Calgary fluoridation cessation case is a key example of how digital misinformation can directly influence policy decisions, reinforcing the need for early intervention and proactive health communication strategies [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eBeyond the sentiment divide, neutral discussions present a critical opportunity for engagement. Research indicates that neutral perceptions are particularly susceptible to influence from emerging information, whether fact-based or misinformation-driven [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. In the case of CWF, neutral discussions likely reflect a lack of access to clear, science-backed information or public uncertainty due to conflicting messages from different sources [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. This underscores the importance of engaging neutral audiences with accessible, transparent, and community-centred health messaging before they become deeply embedded in negative narratives [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Additionally, the recent Cochrane systematic review on water fluoridation (2024) highlights that while CWF remains effective in reducing dental caries, the magnitude of its effect may be lower than earlier estimates, partly due to additional fluoride sources now widely available. Importantly, this review underscores the equity debate, noting that communities with less access to alternative fluoride interventions may benefit substantially from CWF. This resonates with our findings, where neutral sentiments\u0026mdash;representing a potential \u0026lsquo;swing\u0026rsquo; group\u0026mdash;could be influenced by targeted messaging about protective and equitable aspects of fluoridation policies. Policymakers should thus consider the efficacy, fairness, and reach of CWF initiatives when designing public health strategies.\u003c/p\u003e \u003cp\u003eThe presence of emotionally charged language in negative discussions, mainly terms like \"toxic,\" \"harmful,\" and \"chemical,\" indicates a broader public scepticism toward chemically mediated health interventions [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Like the anti-vaccine debate, these discussions are often fuelled by fear-based narratives rather than scientific evidence [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Prior research has shown that misinformation spreads faster than factual information, particularly when sensational language or emotionally compelling arguments are used [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The amplification of misinformation in CWF discussions suggests that the absence of early counter-narratives escalates scepticism, making real-time monitoring and misinformation debunking critical components of health communication strategies [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe role of machine learning in tracking sentiment and misinformation is increasingly evident in public health decision-making [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Our model evaluation demonstrated that Logistic Regression performed best (92.3% accuracy, AUC 0.9), followed by Random Forest (88.4% accuracy, AUC 0.9), while Na\u0026iuml;ve Bayes struggled to classify sentiment accurately (74.8% accuracy, AUC 0.7). The high performance of Logistic Regression and ensemble-based models like Random Forest indicates their potential for real-time monitoring of public sentiment trends, enabling health agencies to track misinformation patterns and adjust communication strategies accordingly [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. These models can help detect shifts in public debates, allowing policymakers to intervene before misinformation narratives escalate [\u003cspan additionalcitationids=\"CR31\" citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. However, future applications should explore deep-learning approaches, such as transformer-based models, to improve sentiment classification accuracy to interpret the sarcastic language used in social media texts [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eGiven the persistence of misinformation and public scepticism, addressing CWF-related discussions requires a multifaceted approach prioritising trust, engagement, and evidence accessibility [34,35]. Early detection of misinformation trends through real-time sentiment analysis and automated misinformation detection models can help counter false claims before they gain traction [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Additionally, proactive engagement strategies, such as fact-checking initiatives and targeted messaging that frames CWF in terms of community benefits rather than technical discussions, can improve public trust and reduce resistance [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Strengthening two-way communication through interactive discussions, Q\u0026amp;A sessions, and collaborations with trusted healthcare professionals could further enhance credibility and encourage community participation in fluoridation policies [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study also has some limitations that warrant consideration. Temporal and demographic biases may have influenced sentiment trends, as changes in social media user demographics and regional events could have shaped online discussions. While we accounted for these biases by contextualising sentiment shifts with significant policy debates, future studies should incorporate demographic and geographic metadata to improve analytical precision. Additionally, this analysis was limited to English-language posts, which may exclude significant discussions in other languages. Since misinformation narratives vary across cultures and linguistic groups expanding sentiment analysis to multilingual datasets would provide a broader perspective on CWF discourse [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Moreover, while automated sentiment analysis is efficient, it may not fully capture human emotions and contextual details. Combining manual validation with AI-driven sentiment tools in future studies could enhance classification accuracy.\u003c/p\u003e \u003cp\u003eFurthermore, while this study focused on Facebook and Instagram, different platforms attract distinct user demographics and engagement patterns. Expanding research to LinkedIn, TikTok, and YouTube could help uncover additional sentiment trends and misinformation patterns, providing a more holistic view of public opinion on CWF [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. Another limitation is that we did not analyse the credibility of users posting CWF-related content\u0026mdash;whether they were general users, key influencers, bots, or institutional accounts. Understanding the role of key opinion leaders and misinformation networks could explain how narratives spread and influence public perceptions of CWF.\u003c/p\u003e \u003cp\u003eOverall, this study underscores the critical role of social media in shaping public sentiment and influencing health policy decisions regarding CWF. While digital platforms facilitate health advocacy and knowledge sharing, they also serve as hubs for misinformation, intensifying public scepticism and policy resistance. Addressing these challenges requires real-time monitoring, adaptive public health communication, and community-driven engagement. Integrating advanced sentiment analysis models into public health surveillance systems could improve early detection of misinformation trends, ensuring that public health messaging remains accurate, timely, and impactful. By adapting to the increasing digital health ecosystem, policymakers can strengthen evidence-based decision-making and reinforce public trust in fluoridation and other preventive health measures.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u0026nbsp;\u003c/strong\u003eThe authors acknowledge the support provided by The University of Queensland.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Declaration\u003c/strong\u003e: NT is supported by a University of Queensland’s Earmarked\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePhD scholarship provided to LD’s MRFF grant # 2024439.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u0026nbsp;\u003c/strong\u003eN.T., L.D., R.L., and D.H. contributed to the concept and design of the study and supervised the research. N.T. conducted the data analysis, drafted the original manuscript, and managed data curation and visualisation. L.D., R.L., and D.H. provided validation, resources, and critical manuscript revisions. Project administration was managed by N.T., with funding acquisition supported by L.D., R.L., and D.H. All authors approved the final article and agreed to the submission.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest Statement:\u0026nbsp;\u003c/strong\u003eThe authors declare no conflicts of interest related to this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement:\u0026nbsp;\u003c/strong\u003eThe datasets generated and/or analysed during the current study are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval Statement:\u003c/strong\u003e The study used publicly available data from 'X' and did not involve human participants. Ethics approval was obtained from The University of Queensland’s Human Ethics Committee (Approval Number: 2022/HE002248).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePatient Consent Statement:\u003c/strong\u003e Not applicable, as this study did not involve direct interaction with patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePermission to Reproduce Material:\u003c/strong\u003e Not applicable, as no previously published materials requiring permissions were used.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical Trial Registration:\u003c/strong\u003e Not applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eVosoughi S, Roy D, Aral S. The spread of true and false news online. Science. 2018;359(6380):1146\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCinelli M, Quattrociocchi W, Galeazzi A, et al. The COVID-19 social media infodemic. Sci Rep. 2020;10(1):16598.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStatista. (2021). Number of monthly active Facebook users worldwide. Available at: [\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.statista.com](https://www.statista.com)\u003c/span\u003e\u003cspan address=\"https://www.statista.com](https://www.statista.com)\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMerchant RM, Lurie N. Social media and emergency preparedness in response to novel coronavirus. JAMA. 2020;323(20):2011\u0026ndash;2.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGabarron E, Wynn R. Use of social media for sexual health promotion: A scoping review. Global Health Action. 2016;9(1):32193.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLarson HJ, et al. The state of vaccine confidence 2016: global insights through a 67-country survey. EBioMedicine. 2016;12:295\u0026ndash;301.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeorge DR, Rovniak LS, Kraschnewski JL. Dangers and opportunities for social media in medicine. Clin Obstet Gynecol. 2013;56(3):453\u0026ndash;62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoozenbeek J, van der Linden S. Fake news game confers psychological resistance against online misinformation. Palgrave Commun. 2020;6:65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeim-Malpass J, Mitchell EM, Sun E, Kennedy C. Using Twitter to understand public perceptions regarding the #HPV vaccine: Opportunities for public health nurses to engage in social marketing. Public Health Nurs. 2017;34(4):316\u0026ndash;23.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLotto M, Menezes TS, Zakir Hussain I, Tsao S, Butt ZA, Morita PP, et al. Characterization of false or misleading fluoride content on Instagram: infodemiology study. J Med Internet Res. 2022;24(5):e37519. [FREE Full text] [doi: 10.2196/37519] [Medline: 35588055].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMertz A, Allukian M. Community water fluoridation on the Internet and social media. J Mass Dent Soc. 2014 Summer;63(2):32\u0026ndash;6. PMID: 25230407.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhelton HP, Spencer AJ, Do LG, Rugg-Gunn AJ. Fluoride revolution and dental caries: evolution of policies for global use. J Dent Res. 2019;98(8):837\u0026ndash;46. [doi: 10.1177/0022034519843495] [Medline: 31282846].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO'Neill B, Kapoor T, McLaren L. Politics, Science, and Termination: A Case Study of Water Fluoridation Policy in Calgary in 2011. Rev Policy Res. 2018;00:1. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/ropr.12318\u003c/span\u003e\u003cspan address=\"10.1111/ropr.12318\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOh O, Agrawal M, Rao HR. Community water fluoridation and online discourse: How misinformation spreads on social media. J Public Health Dent. 2019;79(2):101\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJones SP, et al. A social media study of anti-vaccination commentary and information sharing: The reach and influence of anti-vaccination opinions. J Community Health. 2019;44(6):1020\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlack DR, Dietz JE, Stirratt AA, Coster DC. Do social media have a place in public health emergency response? J Emerg Manag. 2015;13(3):217\u0026ndash;26.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChou WS, Oh A, Klein WMP. Addressing Health-Related Misinformation on Social Media. JAMA. 2018;320(23):2417\u0026ndash;2418. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jama.2018.16865\u003c/span\u003e\u003cspan address=\"10.1001/jama.2018.16865\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 30428002.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOh HJ, Kim CH, Jeon JG. Public Sense of Water Fluoridation as Reflected on Twitter 2009\u0026ndash;2017. J Dent Res. 2020;99(1):11\u0026ndash;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/0022034519885610\u003c/span\u003e\u003cspan address=\"10.1177/0022034519885610\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Epub 2019 Nov 4. PMID: 31682777.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCavazos-Rehg PA, Krauss M, Fisher SL, Salyer P, Grucza RA, Bierut LJ. Twitter chatter about marijuana. J Adolesc Health. 2015;56(2):139\u0026ndash;45.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEysenbach G. Infodemiology and infoveillance tracking online health information and cyberbehavior for public health. Am J Prev Med. 2011;40(5 Suppl 2):S154\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUmunyana C, Tuyizere G, Mbarushimana A. 2024. Students\u0026rsquo; Sentiment Analysis Using Natural Language Toolkit in Machine Learning for Module Evaluation. \u003cem\u003eEuropean Journal of Engineering and Technology Research\u003c/em\u003e. 9, 1 (Mar. 2024), 72\u0026ndash;75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.24018/ejeng.2024.9.1.3006\u003c/span\u003e\u003cspan address=\"10.24018/ejeng.2024.9.1.3006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoon-Itt S, Skunkan Y. Public perception of the COVID-19 pandemic on Twitter: sentiment analysis and topic modelling Study. JMIR Public Health Surveill. 2020;6(4):e21978. [FREE Full text] [doi: 10.2196/21978] [Medline: 33108310].\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcLaren L, Patterson S, Thawer S, et al. Measuring the short-term impact of fluoridation cessation on dental caries in grade 2 children using tooth surface indices. Community Dent Oral Epidemiol. 2016;44(3):274\u0026ndash;82. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/cdoe.12216\u003c/span\u003e\u003cspan address=\"10.1111/cdoe.12216\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCity of Calgary. 2021 municipal election results. City of Calgary website. Published 2021. Accessed August 10. 2024. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.calgary.ca/election/results/2021-results.html\u003c/span\u003e\u003cspan address=\"https://www.calgary.ca/election/results/2021-results.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcLaren L, Singhal S. Does cessation of community water fluoridation lead to an increase in tooth decay? A systematic review of published studies. J Epidemiol Community Health. 2016;70(9):934\u0026ndash;40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/jech-2015-206402\u003c/span\u003e\u003cspan address=\"10.1136/jech-2015-206402\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaddi VV, Kohli RK, Puri P. Perception, use of social media, and its impact on the mental health of Indian adolescents: A qualitative study. World J Clin Pediatr. 2024;13(3):97501. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.5409/wjcp.v13.i3.97501\u003c/span\u003e\u003cspan address=\"10.5409/wjcp.v13.i3.97501\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 39350908; PMCID: PMC11438920.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArmfield JM, Akers HF. Risk perception and water fluoridation support and opposition in Australia. J Public Health Dent. 2010 Winter;70(1):58\u0026ndash;66. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.1752-7325.2009.00144.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1752-7325.2009.00144.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PMID: 19694932.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIheozor-Ejiofor Z, Walsh T, Lewis SR, Riley P, Boyers D, Clarkson JE, Worthington HV, Glenny A-M, O'Malley L. Water fluoridation for the prevention of dental caries. Cochrane Database of Systematic Reviews 2024, Issue 10. Art. No.: CD010856. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/14651858.CD010856.pub3\u003c/span\u003e\u003cspan address=\"10.1002/14651858.CD010856.pub3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 03 March 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJames G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning: with Applications in R. New York: Springer; 2013.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSebastiani F. Machine learning in automated text categorization. ACM Comput Surv. 2002;34(1):1\u0026ndash;47.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreiman L. Random forests. Mach Learn. 2001;45(1):5\u0026ndash;32.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQuinlan JR. Induction of Decision Trees. Mach Learn. 1986;1(1):81\u0026ndash;106.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhan S, Qasim I, Khan W, Aurangzeb K, Khan JA, Anwar MS. (2025). A novel transformer attention-based approach for sarcasm detection. \u003cem\u003eExpert Systems\u003c/em\u003e, 42(1), e13686. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/exsy.1368634\u003c/span\u003e\u003cspan address=\"10.1111/exsy.1368634\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Mackert M, Bouchacourt L, Lazard A, Wilcox GB, Kemp D, Kahlor LA, George C, Stewart B, Wolfe J. Social media conversations about community water fluoridation: formative research to guide health communication. J Public Health Dent. 2021;81(2):162\u0026ndash;166. doi: 10.1111/jphd.12404. Epub 2020 Oct 15. PMID: 33058200.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSabine Meier C, Stock A, Kr\u0026auml;mer. The contribution of health discussion groups with students to campus health promotion, \u003cem\u003eHealth Promotion International\u003c/em\u003e, Volume 22, Issue 1, March 2007, Pages 28\u0026ndash;36. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/heapro/dal041\u003c/span\u003e\u003cspan address=\"10.1093/heapro/dal041\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDredze M. How Social Media Will Change Public Health. IEEE Intell Syst. 2012;27(4):81\u0026ndash;4. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/MIS.2012.74.ss\u003c/span\u003e\u003cspan address=\"10.1109/MIS.2012.74.ss\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Community Water Fluoridation, Social Media Sentiment Analysis, Public Health Communication, Misinformation on Social Media, Public Opinion Trends, Health Policy and Public Perception.","lastPublishedDoi":"10.21203/rs.3.rs-6266081/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6266081/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSocial media platforms like Facebook and Instagram are pivotal in shaping public opinion on health interventions, including Community Water Fluoridation (CWF). Despite its recognition as a safe and effective public health measure, CWF remains a polarising topic, with misinformation on these platforms contributing to public mistrust. This study collected 109,117 Facebook and Instagram posts from 2014 to 2023 to examine public sentiment surrounding CWF. The analysis revealed a mix of opinions, with 42.1% positive, 39.1% negative, and 18.8% neutral sentiments. Trends highlighted a surge in negative sentiment during 2017\u0026ndash;2019, likely influenced by misinformation and significant public events, while positive sentiment has gradually regained ground in recent years. Key themes included health benefits, safety concerns, and government trust, with positive discussions emphasising CWF\u0026rsquo;s role in public health and negative discussions focusing on risks and chemical exposure. The study used advanced sentiment analysis models to highlight the importance of monitoring public discourse and addressing misinformation to build trust and support for evidence-based health policies like CWF. These findings provide digital data-driven insights for public health communication strategies to enhance community understanding and acceptance of vital health interventions.\u003c/p\u003e","manuscriptTitle":"Cracking the 'Meta' Code: Advanced Machine Learning-Based Sentiment Analysis of Water Fluoridation Debates on Facebook and Instagram","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-26 04:26:52","doi":"10.21203/rs.3.rs-6266081/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"82782412-f46b-44f9-9086-dff9a3de5f80","owner":[],"postedDate":"March 26th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-03-26T04:26:55+00:00","versionOfRecord":[],"versionCreatedAt":"2025-03-26 04:26:52","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6266081","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6266081","identity":"rs-6266081","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}