Abstract In increasingly personalized and opaque media environments, traditional methods in communication research struggle to capture the dynamics of algorithm-mediated systems. This paper reviews the rise of agent-based approaches as a methodological response to the observability crisis and mechanistic dilemma posed by algorithmic personalization. We mainly focus on two complementary paradigms: agent-based testing (ABT) and agent-based modeling (ABM). While ABT provides empirical measurement of algorithmic behavior, ABM formalizes theoretical assumptions and enables scenario-based exploration. We further highlight their integration into an iterative loop of measurement, modeling, prediction, and validation, and discuss how recent advances in large language models transform agents into adaptive, human-like entities. Finally, we address methodological and ethical challenges, proposing agent-based approaches as a systematic framework to study, explain, and anticipate the societal consequences of algorithm-mediated communication.
Shi et al. (2026) studied this question.