There is a vast literature that examines differences in overconfidence between people, depending on their knowledge level. In this line of research, the tasks used do not involve learning. As people approach these types of tasks, they generally have some knowledge about their abilities. Some are better, while others are worse, and perceptions of their abilities are assessed. Although there is a lot of research exploring differences in overconfidence depending on knowledge level, there is relatively limited research looking at how overconfidence changes in people as they learn. Much of my work is devoted to exploring the exact shape and timeline of that overconfidence, with an emphasis on probabilistic learning, in which people learn to read cues from the environment to predict some outcome. To explore this line of work first, I mapped out the trajectory of performance, confidence, and overconfidence as people engage in novel probabilistic tasks (Chapter 2). I found that people generally start with rather modest self-assessments, with their confidence tracking performance rather well, but then a problem develops. With just a little learning, confidence sky-rockets far above accuracy, a phase I refer to as the “beginner’s bubble” of overconfidence. There are several reasons for the development of this beginner’s bubble. First, overconfidence develops because people formulate faulty theories about how to approach a task, and once people have an idea about how to perform a task, even if it is wrong, it produces overconfidence (Chapter 2). Second, when people first engage in a task, they pay close attention to feedback to guide their decisions and are appropriately calibrated about their abilities. However, at a certain point, the experienced learner no longer carefully pays attention to feedback to discover why they made mistakes. In fact, overconfidence is not driven by feedback. It is best explained by how quickly people make decisions. The mere act of making a series of decisions, when similar ones have been made before, drives overconfidence (Chapter 3). In exploring mechanisms that lead to the development of the beginner’s bubble, I have also identified individual differences in overconfidence resulting from a specific behavior, the propensity to “jump-to-conclusions” (Chapter 4). Those who jump-to-conclusions are defined as those who collect less information to reach conclusions in problem solving tasks. People who jump-to-conclusions have divergent trajectories of overconfidence as they learn. They have an elevated confidence curve as they are learning, lower levels of overall performance, and are thus more overconfident. These individuals are also more likely to endorse false beliefs, like conspiracy theories. In this line of research, I have also modified metacognitive interventions that are ordinarily designed to reduce delusional beliefs, to quell overconfidence in a probabilistic learning task, without negatively impacting learning.