Conference on Cognitive Computational Neuroscience

Many everyday situations, such as investing money, or allocating attention in exams, involve making uncertain choices using limited resources. In these cases, both choices and the resource allocation fluctuate over time. However, the mechanisms underlying such variability and its impact on performance remain poorly understood. Here, we employ a novel ecological paradigm to test the presence of a consistency bias in sequential multiple-choice scenarios with limited search capacity. We manipulated the environment so that homogenous resource allocation was not always the optimal search regime. The results showed that humans follow search strategies characterised by intentional fluctuations that exceed those predicted by the optimal model, but they have little impact on performance. By extending the optimal model, we show that, overall, the observed variations in resource allocation strike a balance between the gathering of sufficient relevant information for some choices and the risk to leaving some others to chance. These fluctuations prioritise anticipation and information gathering, both beneficial behaviours in a highly dynamical world.