When making a decision, we experience a sense of confidence in the accuracy of this decision. It has been argued that confidence is used for cognitive optimisation (e.g., maximizing gains while minimizing costs), however, direct empirical evidence for this hypothesis is lacking. Inspired by Bayesian accounts, I propose that beliefs about which we are highly certain (i.e., in which we have high confidence) are more resistant to change than beliefs about which we are uncertain (i.e., in which we have low confidence). I will test this hypothesis on three different time-scales. On the within-trial level, I will examine whether subjective confidence determines the extent to which novel information influences our decisions (WP I). On the between-trial level, I will test whether the amount of evidence that is acquired before a decision is made, depends on subjective confidence (WP II). Finally, on a longer time-scale, I will test whether the degree to which our beliefs are altered based on external feedback depends on our subjective confidence in these beliefs (WP III). Using a mixture of neuroimaging recordings (MEG/EEG), pupil recordings, and model-based analysis of behavioural data, I aim to reveal the cognitive and neural architecture of how confidence influences cognitive optimisation.