Tweeting for Health Using Real-time Mining and Artificial Intelligence–Based Analytics: Design and Development of a Big Data Ecosystem for Detecting and Analyzing Misinformation on Twitter
Bibliographic record
Abstract
BACKGROUND: Digital misinformation, primarily on social media, has led to harmful and costly beliefs in the general population. Notably, these beliefs have resulted in public health crises to the detriment of governments worldwide and their citizens. However, public health officials need access to a comprehensive system capable of mining and analyzing large volumes of social media data in real time. OBJECTIVE: This study aimed to design and develop a big data pipeline and ecosystem (UbiLab Misinformation Analysis System [U-MAS]) to identify and analyze false or misleading information disseminated via social media on a certain topic or set of related topics. METHODS: U-MAS is a platform-independent ecosystem developed in Python that leverages the Twitter V2 application programming interface and the Elastic Stack. The U-MAS expert system has 5 major components: data extraction framework, latent Dirichlet allocation (LDA) topic model, sentiment analyzer, misinformation classification model, and Elastic Cloud deployment (indexing of data and visualizations). The data extraction framework queries the data through the Twitter V2 application programming interface, with queries identified by public health experts. The LDA topic model, sentiment analyzer, and misinformation classification model are independently trained using a small, expert-validated subset of the extracted data. These models are then incorporated into U-MAS to analyze and classify the remaining data. Finally, the analyzed data are loaded into an index in the Elastic Cloud deployment and can then be presented on dashboards with advanced visualizations and analytics pertinent to infodemiology and infoveillance analysis. RESULTS: U-MAS performed efficiently and accurately. Independent investigators have successfully used the system to extract significant insights into a fluoride-related health misinformation use case (2016 to 2021). The system is currently used for a vaccine hesitancy use case (2007 to 2022) and a heat wave-related illnesses use case (2011 to 2022). Each component in the system for the fluoride misinformation use case performed as expected. The data extraction framework handles large amounts of data within short periods. The LDA topic models achieved relatively high coherence values (0.54), and the predicted topics were accurate and befitting to the data. The sentiment analyzer performed at a correlation coefficient of 0.72 but could be improved in further iterations. The misinformation classifier attained a satisfactory correlation coefficient of 0.82 against expert-validated data. Moreover, the output dashboard and analytics hosted on the Elastic Cloud deployment are intuitive for researchers without a technical background and comprehensive in their visualization and analytics capabilities. In fact, the investigators of the fluoride misinformation use case have successfully used the system to extract interesting and important insights into public health, which have been published separately. CONCLUSIONS: The novel U-MAS pipeline has the potential to detect and analyze misleading information related to a particular topic or set of related topics.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.034 | 0.007 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".