Justice sensitivity (JS) reflects personal concern and commitment to the principle of justice, showing considerable heterogeneity among the general population. Despite a growing interest in the behavioral characteristics of JS over the past decades, the neurobiological substrates underlying trait JS are not well comprehended. We addressed this issue by employing a machine learning approach to decode the trait JS, encompassing its various orientations, from whole-brain resting-state functional connectivity. We demonstrated that the machine-learning model could decode the individual trait of other-oriented JS but not self-oriented JS from resting-state functional connectivity across multiple neural systems, including functional connectivity between and within parietal lobe and motor cortex as well as their connectivity with other brain systems. Key nodes that contributed to the prediction model included the parietal, motor, temporal, and subcortical regions that have been linked to other-oriented JS. Additionally, the machine learning model can distinctly distinguish between the distinct roles associated with other-oriented JS, including observer, perpetrator, and beneficiary, with key brain regions in the predictive networks exhibiting both similarities and disparities. These findings remained robust using different validation procedures. Collectively, these results support the separation between other-oriented JS and self-oriented JS, while also highlighting the distinct intrinsic neural correlates among the three roles of other-oriented JS: observer, perpetrator, and beneficiary.