Traditionally, belief-updating biases have been studied in isolation. But people can make more than one kind of mistake at a time. This paper addresses the challenge of identifying and testing multiple probabilistic updating biases simultaneously. It does so by constructing a theoretical model that encompasses a wide range of biases previously proposed in the literature (mostly) in a disconnected fashion. Next, the model is tested in the laboratory to assess whether the presence of any of those biases intensifies, diminishes or fully disappears, after including the complete set of biases.

At the population level, base rate neglect emerges as a persistent influence, while an individual-level analysis uncovers significant heterogeneity: All tested biases are present to a certain extent and each participant exhibits at least one identifiable bias. Notably, motivated-belief biases (optimism and pessimism) and sequence-related biases (gambler's fallacy and hot hand fallacy) are identified as key drivers of biased inference, whereas the presence confirmatory biases is scarce. This study contributes to the belief-updating literature by providing a methodological toolkit for researchers examining links between different conflicting biases, or exploring connections between updating biases and other behavioural phenomena.