Overconfidence as Truth Approximation

Human reasoning and decision-making under uncertainty often deviate from normative standards of rationality. Over the past decades, cognitive scientists have extensively investigated heuristics and cognitive biases, such as overconfidence—the tendency to overestimate the probability that one's judgments are correct. Meanwhile, philosophers have explored different "cognitive utilities" that guide both scientific and everyday reasoning, including the concept of truthlikeness, i.e., how well an hypothesis, be it a statement or a numerical interval, approximates the whole truth about a target domain. In this paper, we integrate empirical findings with philosophical perspectives, showing how formal models of truthlikeness offer valuable insights for empirical research on overconfidence. In particular, by conceptualizing overconfidence through the lens of expected truthlikeness maximization, we argue that many instances of this phenomenon may be construed not as cognitive biases, but rather as rational strategies for approaching the truth under conditions of uncertainty.