Interdisciplinary Approach to Identify and Characterize COVID-19 Misinformation on Twitter: Mixed Methods Study

Area of Science:

• Computational social science
• Health informatics
• Digital communication

Background:

• Studying COVID-19 misinformation on Twitter faces methodological challenges.
• Computational methods struggle with context, while qualitative methods are labor-intensive.
• An interdisciplinary approach is needed to address these limitations.

Purpose of the Study:

• To identify and characterize tweets containing COVID-19 misinformation.
• To explore the formats and discursive strategies of misinformation.
• To evaluate the effectiveness of computational and qualitative methods.

Main Methods:

• Tweets from the Philippines (Jan-Mar 2020) were collected and analyzed using biterm topic modeling.
• Key informant interviews informed keyword identification for manual coding of subcorpora.
• Natural language processing (NLP) was used for initial identification, followed by manual review.

Main Results:

• Biterm topic modeling identified key themes related to COVID-19.
• Manual coding revealed misinformation formats like misleading content and conspiracy theories.
• NLP misidentified a significant portion of tweets as misinformation, particularly those in Filipino.

Conclusions:

• An interdisciplinary approach combining computational and qualitative methods is effective for identifying COVID-19 misinformation.
• Manual coding and human expertise are crucial for accurate analysis, especially with multilingual content.
• NLP models require refinement to handle language nuances in social media data.