Divisible codes were introduced by H.N. Ward in 1981. A divisible code is a linear code over a finite field whose codewords all have weights divisible by some integer Δ > 1, where Δ is called a divisor of the code. A binary linear code is said to be of (divisibility) level e if e is the greatest integer such that 2

With the growing prevalence of health misinformation online, there is an urgent need for tools that can reliably assist the public in evaluating the quality of health information. This study investigates the performance of GPT-3.5-Turbo, a representative and widely used large language model (LLM), in rating the quality of health news and providing explanatory justification for the rating assessment. We evaluated GPT-3.5-Turbo’s performance on 3222 health news articles from an expert-annotated dataset compiled by HealthNewsReview.org, which assesses the quality of health news across nine criteria. GPT-3.5-Turbo was prompted with standardized queries tailored to each criterion. We measured its rating performance using 95% confidence intervals for precision, recall, and F1 scores in binary classification (satisfactory/not satisfactory). Additionally, linguistic complexity, readability, and the quality of GPT-3.5-Turbo’s explainability were assessed through both quantitative linguistic analysis and qualitative evaluation of consistency and contextual relevance. GPT-3.5-Turbo’s rating performance varied across criteria, with the highest accuracy for the Cost criterion (F1 = 0.824) but lower accuracy for Benefit, Conflict, and Quality criteria (F1 < 0.5), underperforming compared to traditional supervised machine learning models. However, its explanations were clear, with readability suited to late high school or early college levels and scored highly for consistency (average score: 2.90/3) and contextual relevance (average score: 2.73/3). These findings highlight GPT-3.5-Turbo’s strength in providing understandable and contextually relevant explanations, despite that its rating accuracy is limited. While GPT-3.5-Turbo’s rating accuracy requires improvement, its strength in offering comprehensible and contextually relevant explanations presents a valuable opportunity to enhance public understanding of health news quality. Leveraging LLMs as complementary tools for health literacy initiatives could help mitigate misinformation by facilitating non-expert audiences to interpret and assess health information.

Background: Machine learning techniques have been shown to be efficient in identifying health misinformation, but the results may not be trusted unless they can be justified in a way that is understandable. Objective: This study aimed to provide a new criteria-based system to assess and justify health news quality. Using a subset of an existing set of criteria, this study compared the feasibility of 2 alternative methods for adding interpretability. Both methods used classification and highlighting to visualize sentence-level evidence. Methods: A total of 3 out of 10 well-established criteria were chosen for experimentation, namely whether the health news discussed the costs of the intervention (the cost criterion), explained or quantified the harms of the intervention (the harm criterion), and identified the conflicts of interest (the conflict criterion). The first step of the experiment was to automate the evaluation of the 3 criteria by developing a sentence-level classifier. We tested Logistic Regression, Naive Bayes, Support Vector Machine, and Random Forest algorithms. Next, we compared the 2 visualization approaches. For the first approach, we calculated word feature weights, which explained how classification models distill keywords that contribute to the prediction; then, using the local interpretable model-agnostic explanation framework, we selected keywords associated with the classified criterion at the document level; and finally, the system selected and highlighted sentences with keywords. For the second approach, we extracted sentences that provided evidence to support the evaluation result from 100 health news articles; based on these results, we trained a typology classification model at the sentence level; and then, the system highlighted a positive sentence instance for the result justification. The number of sentences to highlight was determined by a preset threshold empirically determined using the average accuracy. Results: The automatic evaluation of health news on the cost, harm, and conflict criteria achieved average area under the curve scores of 0.88, 0.76, and 0.73, respectively, after 50 repetitions of 10-fold cross-validation. We found that both approaches could successfully visualize the interpretation of the system but that the performance of the 2 approaches varied by criterion and highlighting the accuracy decreased as the number of highlighted sentences increased. When the threshold accuracy was ≥75%, this resulted in a visualization with a variable length ranging from 1 to 6 sentences. Conclusions: We provided 2 approaches to interpret criteria-based health news evaluation models tested on 3 criteria. This method incorporated rule-based and statistical machine learning approaches. The results suggested that one might visually interpret an automatic criterion-based health news quality evaluation successfully using either approach; however, larger differences may arise when multiple quality-related criteria are considered. This study can increase public trust in computerized health information evaluation.

Background: Emerging research has highlighted the potential of virtual reality (VR) as a tool for training health care students and professionals in care skills for individuals with Alzheimer disease and related dementias (ADRD). However, there is limited research on the use of VR to engage the general public in raising awareness about ADRD. Objective: This research aimed to examine the impact of the VR video “A Walk-Through Dementia” on YouTube users by analyzing their posts. Methods: We collected 12,754 comments from the VR video series “A Walk-Through Dementia,” which simulates the everyday challenges faced by individuals with ADRD, providing viewers with an immersive experience of the condition. Topic modeling was conducted to gauge viewer opinions and reactions to the videos. A pretrained Bidirectional Encoder Representations from Transformers (BERT) model was used to transform the YouTube comments into high-dimensional vector embeddings, allowing for systematic identification and detailed analysis of the principal topics and their thematic structures within the dataset. Results: We identified the top 300 most frequent words in the dataset and categorized them into nouns, verbs, and adjectives or adverbs using a part-of-speech tagging model, fine-tuned for accurate tagging tasks. The topic modeling process identified eight8 initial topics based on the most frequent words. After manually reviewing the 8 topics and the content of the comments, we synthesized them into 5 themes. The predominant theme, represented in 2917 comments, centered on users’ personal experiences with the impact of ADRD on patients and caregivers. The remaining themes were categorized into 4 main areas: positive reactions to the VR videos, challenges faced by individuals with ADRD, the role of caregivers, and learning from the VR videos. Conclusions: Using topic modeling, this study demonstrated that VR applications serve as engaging and experiential learning tools, offering the public a deeper understanding of life with ADRD. Future research should explore additional VR applications on social media, as they hold the potential to reach wider audiences and effectively disseminate knowledge about ADRD.

Information voids have been anecdotally marked as a precursor to outbreaks of misinformation during health and climate crises but remain an empirically untested phenomenon. Understanding the public’s information needs and identifying voids is critical if disaster risk communication must preserve lives and livelihoods during climate emergencies. This paper conceptualizes and tests a novel survey tool designed to detect information voids across four key dimensions: information quantity, quality, source, and channel. Following major climate emergencies in 2024, a series of quantitative, online cross-sectional surveys were conducted in four countries, with a total of 897 respondents. The study aimed to assess the reliability and sensitivity of the tool in identifying gaps in information during climate crises. Study 1 in Belgium (n = 202) detected information voids across all dimensions after Storm Boris (  p < .001). Study 2 demonstrated the ability of the tool to detect information voids in other geographical contexts [Germany (n = 197) and Poland (n = 191)] during similar flooding events (both  p < 0.01). Study 3 confirmed the tool’s ability to detect information voids across all dimensions in the context of a different climate emergency: Hurricane Helene in Florida, USA (n = 307) (  p < 0.01). Our findings demonstrate the potential of the survey to generate rapid evidence around information gaps and deliver detailed, actionable insights to improve disaster communication during emergencies in various regions. We discuss implications for addressing misinformation and disinformation during climate emergencies, as well as strategies for enhancing flood (disaster) risk communication and management.

Background: Social media has become a common source for health information, yet the relationship between users’ experiences with this form of digital health information and health self-efficacy remains poorly understood. This study investigates how the reported reliance on and perceived evaluative difficulty of social-media-based health information are associated with self-efficacy levels in a misinformation-prone environment. Methods: This cross-sectional study analyzed data from the 2022 Health Information National Trends Survey (HINTS 6), a nationally representative sample of U. S. adults ( N = 4,352). Binomial logistic regression was used to estimate the likelihood of being “completely confident” in one’s ability to manage health. Key predictors included: reported use of social media to make health decisions, use of such information in clinical discussions, and the perceived difficulty distinguishing between true and false health information. The model controlled for age, sex, race/ethnicity, education, daily social media use, and self-rated health. Results: Approximately 27% of respondents reported complete health self-efficacy. Multivariable analysis showed that individuals who strongly disagreed with using social media information for health decisions were significantly more likely to report high self-efficacy (OR = 1.37, p = 0.041). Similarly, those who strongly disagreed that it was difficult to differentiate between true and false health information were 47% more likely to be completely confident about health management (OR = 1.47, p = 0.009). Using social media information in discussions with providers was not associated with health self-efficacy. Notably, women (OR = 1.57, p < 0.001) and daily social media users (OR = 1.38, p = 0.013) reported higher confidence, while Asian respondents were less likely to report being completely confident compared to their White counterparts (OR = 0.44, p = 0.004). Conclusion: Health self-efficacy is associated with selective non-reliance on social-media-based health information for decision making and perceived evaluative ease. These patterns are consistent with Social Cognitive Theory, which emphasizes the role of perceived competence in shaping confidence beliefs. Although causal direction cannot be established, the findings suggest that public health efforts may benefit from supporting digital health literacy and critical evaluative confidence may be relevant to strengthening confidence in health management within increasingly complex digital information environments.