Decisions feel daunting, as the weight of selecting one path over another engenders a sense of unease and hesitation. This is a natural consequence of the fact that most decisions in the real world involve uncertainty. A choice involves two or more alternatives and usually resolves in the selection of a subjectively preferable option. Very often, however, decision-making requires consideration of multiple options and their numerous possible outcomes, as well as determining what ’preferable’ stands for. This makes choices risky.’ With this thesis, I studied the neural principles associated with risk, and I tested fluctuations of risk-taking as predicted by a novel computational model and social identity theory. In the first experiment, I used social stimuli as cues and recorded trial-by-trial fluctuations in EEG to try and capture brain responses to estimates of risk and risk prediction errors. The results reveal distinct spatio-temporal EEG component sassociated with the computation of risk and violations of expected risk. In the second experiment, I tested mediators of trial-by-trial risk-seeking. More specifically, independent of participants’ risk propensities in real life, I implemented a task to drive risk-taking choice in some trials and risk avoidance in others. I showed that innon-social contexts positive reward prediction errors can predict risk-taking, and I found brain responses associated with this process. In the final chapter, I discuss the same task in which I tried to induce risk-seeking with the addition of a social factor aiming to test predictions from social identity theory. I showed that the mere online presence of an in-group and an out-group member was enough to alter behaviour during the task, although possible explanations can span from exploratory behaviour in some groups of participants, to overall arousal or increased stress in the participants while being observed. Together these experiments show the importance of incorporating social factors into studies of decision-making, the benefit of computational methods for a better understanding of risky decision-making and model-based neural responses, and the importance of accounting for individual differences when studying value based choice.