Performance of a Genetic Algorithm for Estimating DeGroot Opinion Diffusion Model Parameters for Health Behavior...

Area of Science:

• Computational Social Science
• Network Science
• Public Health Interventions

Background:

• Social influence is key for public health interventions, but effectiveness assessment requires opinion diffusion modeling.
• Previous work developed a genetic algorithm for the DeGroot model in small networks.
• Practical applications demand assessing algorithm performance under realistic, imperfect conditions.

Purpose of the Study:

• Evaluate a genetic algorithm's performance for opinion diffusion modeling under challenging, real-world conditions.
• Assess the impact of ordinal opinion data, network sampling, and model misspecification on algorithm accuracy.
• Apply findings to understand opinion dynamics for social network interventions, specifically pre-exposure prophylaxis (PrEP) uptake.

Main Methods:

• Simulation study to test algorithm performance with ordinal opinion measurements.
• Investigated algorithm robustness against network sampling and model misspecification.
• Applied simulation insights to analyze opinion diffusion for a social network intervention targeting PrEP uptake.

Main Results:

• The genetic algorithm demonstrates robustness across various non-ideal conditions, including ordinal data and model misspecification.
• Algorithm performance is sensitive to the precision of the ordinal opinion scale.
• Complete network sampling is not essential for the method's effective application.

Conclusions:

• The validated genetic algorithm is a reliable tool for modeling opinion diffusion in complex social networks.
• The findings support the use of this method for projecting the effectiveness of public health interventions.
• Insights gained can inform the design of social network strategies to increase PrEP uptake among key populations.