Biases in online reputation systems: a survey of the empirical literature

Abstract Reputation systems are essential for creating trust and reducing information asymmetries in online markets. However, they are vulnerable to biases that distort information, erode trust and potentially cause market failure. This paper surveys the empirical literature on five key sources of bias: (i) strategic actions of sellers such as fake reviews and rebate-for-review programs, (ii) reciprocity, (iii) social influence bias, (iv) selection bias and (v) noise. It analyzes the biases through a unified framework decomposing errors into three distinct signatures: systematic mean shifts, variance inflation and serial correlation. A central finding is that platform design involves trade-offs: interventions targeting one error signature often exacerbate other distortions. By illustrating how biases interact, this paper provides a diagnostic roadmap for platform operators, regulators and policymakers to match interventions to the specific error signatures while compensating for secondary effects, ultimately enabling reputation systems to maintain trust, reduce information asymmetries and enable efficient online markets.