Spreading paths in partially observed social networks

Area of Science:

• Network Science
• Epidemiology
• Computational Social Science

Background:

• Understanding information, behavior, and pathogen spread in social networks is crucial for predicting epidemic size and designing interventions.
• Inferring spread paths faces challenges due to difficulties in observing dynamic processes and partial network observation.

Purpose of the Study:

• To investigate how well partially observed shortest paths in social networks emulate unobserved stochastic spreading paths.
• To analyze the impact of partial network observation on the estimated length of spread paths.

Main Methods:

• Simulations were conducted on static, structurally realistic social networks.
• Three types of paths were compared: stochastic spread path, shortest path in a fully observed network, and shortest path in a partially observed network.

Main Results:

• Partial network observation can inflate the length of the shortest path.
• The stochastic nature of spreading processes often causes the actual spread path to deviate from the shortest path.
• Counterintuitively, partial observation did not always lead to an inflated estimate of the spread path length; it could even make it appear shorter.

Conclusions:

• Partially observed shortest paths may not accurately reflect the true length of spreading processes in social networks.
• The interplay between network structure, partial observation, and stochasticity complicates the accurate inference of spread dynamics.