Using impression data to improve models of online social influence

Area of Science:

• Social media dynamics
• Computational social science
• Network analysis

Background:

• Influence is crucial on social media, yet current research is limited by data focusing on expressions (likes, comments) over impressions (views).
• Existing studies lack ground truth measures for true influence.
• Understanding social media influence requires a comprehensive approach beyond observable engagement.

Purpose of the Study:

• To develop and validate novel models of social media influence.
• To investigate the relative importance of impressions versus expressions in driving influence.
• To provide a more accurate method for identifying influential social media accounts.

Main Methods:

• Implemented a social media simulation on an original web-based micro-blogging platform.
• Developed three distinct influence models incorporating both expressions and impressions data.
• Utilized simulated user data to test and compare model performance.

Main Results:

• Impressions were found to be significantly more important drivers of influence than expressions.
• The proposed models accurately identified the most influential accounts within the simulation.
• Analysis revealed the emergence of a few highly influential accounts and the formation of opinion echo chambers.

Conclusions:

• Impressions are a critical, often overlooked, factor in social media influence.
• Accurate influence modeling requires integrating both exposure and engagement data.
• This approach enhances understanding of key social media phenomena like opinion polarization.