Social expectations play a critical role in everyday decision-making. However, their precise neuro-computational role in the decision process remains unknown. Here we adopt a decision neuroscience framework by combining methods and theories from psychology, economics and neuroscience to outline a novel, expectation-based, computational model of social preferences. Results demonstrate that this model outperforms the standard inequity-aversion model in explaining decision behavior in a social interactive bargaining task. This is supported by fMRI findings showing that the tracking of social expectation violations is processed by anterior cingulate cortex, extending previous computational conceptualizations of this region to the social domain. This study demonstrates the usefulness of this interdisciplinary approach in better characterizing the psychological processes that underlie social interactive decision-making.