The goal of this dissertation is to investigate the structure and predictability of aesthetic preferences at the individual level. While much of the existing literature in empirical aesthetics focuses on group-level trends and averages, the studies presented here demonstrate that individual aesthetic judgments are not only consistent but also predictable—often more accurately than models based on aggregate data. The improved predictive accuracy observed in these studies arises from three central findings. First, individuals make use of low-level visual features in distinct ways. As shown in Chapter 2, features that strongly predict beauty ratings for some individuals are ineffective for others. This variability suggests that aesthetic evaluations are not fully governed by universal rules but also shaped by personalized visual sensitivities. Second, Chapter 3 reveals that aesthetic preferences are not randomly distributed across the population. Instead, people cluster into taste cohorts—groups of individuals who systematically share similar evaluative patterns. These cohorts reflect meaningful structure in aesthetic taste and challenge the assumption that individual differences are purely idiosyncratic or noisy. Third, Chapter 4 shows that individuals apply stable internal evaluation criteria when judging images. This internal consistency enables others to learn and predict a person’s aesthetic preferences, even when objective visual features offer little guidance. Observers are able to infer and approximate another individual’s aesthetic profile through brief exposure to their previous ratings. Together, these findings suggest that aesthetic preferences are both individualized and intelligible. They argue for a shift away from one-size-fits-all models toward frameworks that respect the structure of individual differences. Such an approach has implications for personalized recommendation systems, computational models of taste, and our broader understanding of how people experience visual beauty.