Understanding how social behavior evolves across social isolation is both an enormous priority and technical challenge as exploring the neural underpinnings of self-initiated behaviors in a quantitative and rigorous manner is difficult without experimenter-defined trial structures. Although loneliness has been declared an epidemic, there are no therapeutic targets and no mechanistic explanations for how social isolation negatively impacts our health. The dynamic time course of social isolation has been conceptualized in a model of social homeostasis (Lee et al., 2021). Multiple timescales for social behavioral dynamics have been observed in the PFC (Padilla-Coreano et al., 2022). Here, we investigate the neural representations of social interaction during a resident-intruder task with mice that have been acutely or chronically isolated (n=18). To uncover differences in their neural representations, we developed Shared Representation Discovery (ShaReD), a method for aligning high-dimensional paired multi-modal data across subjects. ShaReD identifies interpretable common feature combinations across all individuals in one modality (i.e. behavioral features) that are maximally correlated with specific feature combinations in another modality (i.e. neural activity) in each individual. We applied ShaReD to the neural and behavioral data, and found that neurons from each group were strongly correlated to multiple distinct behavioral features and that neurons from acutely isolated animals encoded information about the conspecific more strongly than chronically isolated or control mice (p<0.05). Our findings illuminate the distinct neural consequences of varying social isolation durations. This underscores ShaReD's utility in advancing neuroscience, particularly in examining neural correlates of behaviors in more ethological settings.