An approach to sociotechnical transparency of social media algorithms using agent-based modelling

Area of Science:

• Social Sciences
• Computer Science
• Information Science

Background:

• Social media recommendation algorithms significantly influence information dissemination and user interaction.
• The opacity of these algorithms poses challenges for understanding their societal impact and for effective regulation.
• Existing transparency approaches often fail to account for the complex interplay between technical systems and their social context.

Purpose of the Study:

• To introduce and define the concept of sociotechnical transparency for social media algorithms.
• To present a novel agent-based modeling approach for achieving sociotechnical transparency.
• To explore the alignment between algorithm priorities, platform claims, and public preferences.

Main Methods:

• Development of a novel agent-based model (ABM) to simulate social media recommendation algorithms.
• Implementation of a multi-objective recommendation algorithm within the ABM.
• Empirical validation of the ABM using data from X (formerly Twitter).

Main Results:

• Agent-based modeling provides a viable method for achieving sociotechnical transparency in recommendation systems.
• The model demonstrates how recommendation algorithms prioritize different content curation signals for specific topics.
• Insights gained can illuminate discrepancies between platform stated goals and actual algorithmic behavior.

Conclusions:

• Sociotechnical transparency is crucial for understanding and governing social media algorithms.
• Agent-based modeling offers a powerful tool for auditing and analyzing algorithmic decision-making processes.
• Further research can assess the alignment of algorithmic priorities with societal values and user expectations.