Insights into elections: An ensemble bot detection coverage framework applied to the 2018 U.S. midterm elections

Area of Science:

• Computational Social Science
• Network Science
• Artificial Intelligence

Background:

• Social bots are increasingly prevalent in online social networks (OSNs), raising concerns about their impact on information dissemination.
• OSNs are a primary information source for many individuals, making the detection of automated social bots critical.
• Current social bot detection methods are diverse and vary in effectiveness, potentially limiting research scope.

Purpose of the Study:

• To address the limitations of single-source social bot detection.
• To introduce and test an ensemble bot detection coverage framework using multiple detection sources.
• To assess social bot activity on Twitter during the 2018 U.S. Midterm Election.

Main Methods:

• Developed an ensemble framework combining multiple social bot detection sources.
• Applied the framework to a corpus of Twitter data related to the 2018 U.S. Midterm Election.
• Utilized three distinct bot detection sources for analysis.

Main Results:

• Minimal overlap was observed between bot accounts detected by different sources within the same dataset.
• The ensemble approach demonstrated the ability to detect a wider array of social bots compared to single methods.
• Significant diversity in bot detection outcomes highlights the limitations of relying on a single detection tool.

Conclusions:

• Social bot research necessitates the integration of multiple detection sources to capture the full spectrum of bot activity.
• The evolving complexity of social bots requires continuous development of improved and novel detection methodologies.
• An ensemble approach is vital for comprehensive and accurate social bot detection in online social networks.