Humans have an extraordinary capacity to seek out novel information relevant for current and future goals. However, the use of simple choice tasks in decision-making research in which participants are not able to sample information freely has limited our ability to investigate the drivers of information sampling and their neural signatures. In this thesis, I present findings using active information sampling tasks combined with functional magnetic resonance imaging (fMRI) data and large-scale online participant samples. We find that activity in areas of medial frontal cortex (MFC) and parietal cortex predict upcoming decisions to sample in a multi-step decision-making task with active sampling. Furthermore, a large network of brain regions is shown to represent key features of sampled information. I then describe the development of a novel information sampling task, in which hypothesized drivers of information sampling can be manipulated. Running many pilot task versions on online participants reveals sampling behavior to be heterogeneous and the existence of many sampling strategies. A version of this task was used in combination with psychiatric questionnaires to investigate the link between compulsivity and information sampling in a large-scale online study. We find evidence for an effect of compulsivity on the integration of task features to drive sampling, but this is not replicated in a confirmation sample. Finally, another version of the developed task was used in combination with fMRI to investigate the neural signatures of different drivers of sampling. One driver of sampling was represented in brain activity: goal proximity. We find a similar pattern of brain activity predictive of upcoming decisions to sample, but this effect does not reach significance in MFC. The work in this thesis shows that areas known to frequently activate in decision-making tasks are likely to also drive information sampling to satisfy upcoming goals.