Retention Induced Biases in a Recommendation System with Heterogeneous Users

Recommendation algorithms are machine learning systems that are all around us and influence social processes, democratic systems and individual cognition. They can lead to recommendation biases when they optimise for user retention in ways that shape cognitive processing. This is why it’s important that algorithm evaluation frameworks achieve a careful balance of engagement with cognitive health, diversity of thought and long‐term user satisfaction. Dr Shichao Ma is an expert in the field of data science who sees limitations with a lot of A/B testing. He believes that A/B tests measured by short-term metrics can be misleading and biased as they don’t account for long‐term user inflow and churn dynamics. Ma is exploring machine learning systems under a critical social science lens, considering the long‐term dynamics of recommender biases. He is applying conceptual insights from social theory and empirical analysis into practical contexts. His research is addressing a critical gap by moving away from single-algorithm optimisation and instead analysing the broader, long‐term social implications of recommendations. Ma is focusing on the interplay between algorithmic changes, user retention dynamics and data biases that impact evaluation accuracy. The goal is to create greater awareness and spark deeper discussions about how recommendation systems evolve in real‐world settings.