Identifying Misinformation About Unproven Cancer Treatments on Social Media Using User-Friendly Linguistic...

Area of Science:

• Health communication
• Computational linguistics
• Social media analysis

Background:

• Health misinformation on social media poses risks, especially for cancer patients.
• Existing advice to avoid misinformation requires advanced research skills.
• This study aims to simplify misinformation identification for general users.

Purpose of the Study:

• Identify user-friendly characteristics of misinformation for social media flagging.
• Develop practical methods for users to detect cancer misinformation online.
• Enhance public health literacy in the digital age.

Main Methods:

• Conducted a structured literature review on algorithmic misinformation detection.
• Collected and manually labeled social media data related to cancer misinformation.
• Employed natural language processing and predictive modeling to analyze linguistic features.

Main Results:

• Extracted user-friendly linguistic characteristics from 88 research papers.
• Achieved 73% accuracy in predicting misinformation by comparing posts with healthcare systems.
• Identified tentative language, location, URLs, and hashtags as negative predictors, while numbers and absolute language were positive predictors.

Conclusions:

• Developed practical recommendations for users to identify misinformation.
• Advised caution with social media posts containing strong assurances or unreferenced numbers.
• Highlighted the need for future studies to assess recommendation efficacy among information users.