Maintaining an accurate internal model of our changing environment is essential for efficient decision-making. Previous studies of perceptual decision making were focusing almost exclusively on overly simplistic situations, in which observed changes could be accounted for by a single parameter of the internal model. We extended these investigations to more realistic situations in which changes in external conditions could be explained by multiple, equally feasible variants of the complex internal model through the adjustment of its multiple parameters simultaneously and obtained three results. First, using Bayesian ideal observer analysis and a novel sequential 2AFC visual discrimination paradigm, we developed a method in which we could use observers’ behavioural response biases to identify the internal representations they used during decision making. Second, we found by computational modelling and verified by a set of behavioural experiments that in such complex tasks, observers’ interpretation was strongly modulated by the specific dynamics of the observed and latent aspects of the sequential input. Third, we showed that this behaviour could not be accounted for by simple models dominant in the literature but could be qualitatively captured by assuming that observers rely on hierarchical representations with detailed dynamics of each parameter of their automatically developed internal model and they use this information for readjusting their model to properly account for the changes in the input sequence. To verify that our Bayesian model fits were correct, we developed a new method, a strong form of cross-validation: First, we demonstrated that the parameters of the abstract Bayesian model naturally map onto the parameters of a process-level sequential sampling model, then we showed that this process-level model could in turn explain idiosyncratic reaction-time patterns present in the behavioural data that were out of the scope of the original Bayesian model. Importantly, our results are compatible with a fully Bayesian view of perceptual decision making, in which uncertainty at various levels of the complex internal model representing and interpreting the external input is optimally accounted for. Our approach provides a new way of investigating complex human decision making.