Algorithmic bias amplifies opinion fragmentation and polarization: A bounded confidence model

Area of Science:

• Computational Social Science
• Sociophysics
• Information Science

Background:

• Online media platforms optimize information flow based on popularity and proximity, not content relevance.
• This optimization strategy can lead to algorithmic bias, potentially increasing societal fragmentation and polarization.

Purpose of the Study:

• To investigate the impact of algorithmic bias on opinion dynamics.
• To modify a bounded confidence model to incorporate preferential interaction with similar peers.

Main Methods:

• Modification of the continuous opinion dynamics model of bounded confidence.
• Introduction of an enhanced probability for selecting discussion partners with similar opinions, mimicking online media behavior.

Main Results:

• Observed increased opinion fragmentation, even in conditions where the original model predicted consensus.
• Documented increased polarization of opinions within the simulated population.
• Noted a significant slowing down of convergence to an asymptotic state, indicating system instability.

Conclusions:

• Algorithmic bias in information dissemination exacerbates societal fragmentation and polarization.
• The modified opinion dynamics model demonstrates how preferential interaction can destabilize societal debate.
• Initial population fragmentation further amplifies the observed negative effects.