Detecting bias in algorithms used to disseminate information in social networks and mitigating it using...

Area of Science:

• Social Network Analysis
• Information Propagation Dynamics
• Computational Social Science

Background:

• Social connections facilitate communication, information spread, and disease transmission.
• Identifying key individuals (influencers) is crucial for effective campaigns and epidemic control.
• Existing influence maximization algorithms aim to identify these influencers but may have unintended consequences.

Purpose of the Study:

• To evaluate the information equity of current influence maximization algorithms.
• To develop a novel algorithm that optimizes both influence spread and information equity.
• To demonstrate that maximizing information spread does not necessitate compromising information equality.

Main Methods:

• Extensive computer simulations on synthetic and 10 real-world social networks.
• Analysis of information dissemination patterns using state-of-the-art influence maximization methods.
• Development and testing of a multiobjective algorithm balancing influence and equity.

Main Results:

• Current influence maximization methods create significant information gaps.
• Selected influencers often do not disseminate information equitably, potentially increasing societal inequality.
• The proposed multiobjective algorithm effectively reduces information gaps with minimal trade-off in spread.

Conclusions:

• Influence maximization algorithms can inadvertently exacerbate societal inequalities.
• A balanced approach is needed to ensure equitable information dissemination.
• It is feasible to achieve widespread information reach without sacrificing information equality.