We are defined by our behavior--how we act and the decisions we make throughout life. The processes underlying decision-making are not fully understood, especially in regards to developmental and clinical populations. Cognitive processes have been proposed to reflect the function of two systems--one fast and automatic, the other slow and deliberative. These dual-systems models fail to acknowledge the complex interconnectivity of neural networks, but have provided a useful framework in psychology for understanding decision-making. An extension of this work uses computational reinforcement-learning algorithms to help characterize potential evaluative processes thought to underlie decision formation. Simple evaluations use error-based feedback from prior responses to track the values of options while increasingly complex evaluations supplement that information to guide goal-directed actions. Such goal-directed decisions are proposed to rely on the ability to form and recruit a cognitive representation of decision relevant information. Relatedly, information received through instruction is thought to have a prolonged biasing effect on habitual learning processes. While the signal underlying simple reinforcement-learning algorithms closely matches neural activity within a well-circumscribed circuit, more complex evaluative processes making up higher-order cognitive models of the world involve a distributed network of brain regions. Given that many of these regions and their connections show dynamic changes across development and perturbations in clinical populations, there are likely significant differences in the evaluative processes of decision-making among these groups relative to adults. Chapter One provides an overview of habitual and goal-directed decision-making. Chapter Two tests whether children and adolescents recruit task structure knowledge to make goal-directed decisions to the same extent that adults do. Chapter Three examines how instruction biases learning and decision-making across development. Chapter Four addresses how decisions involving delayed outcomes may be perturbed in anorexia nervosa. Chapter Five summarizes these results and offers a critical assessment of the current state of computational modeling of decision-making in developmental and psychiatric populations. Collectively, in this thesis, I report an initial attempt at using computational modeling as a method for understanding the underlying evaluative processes behind individual and developmental differences in decision-making, and discuss the advantages and disadvantages of this approach.