Human behavior is flexibly regulated in accordance with specific goals of cognitive tasks. One notable instance is the goal-directed modulation of human metacognitive behavior, where logically equivalent decision-making problems can yield different patterns of introspective confidence depending on the frame in which they are presented. While this observation highlights the important heuristic nature of metacognitive monitoring, the computational mechanisms of this phenomenon remain elusive. Using a two-alternative dot-number discrimination task, we aimed to investigate the underlying computational rules of the confidence framing effect. Participants made decisions under “choose more” or “choose less” frames and reported their confidence. We replicated the previous observation of the confidence framing effect, demonstrating distinctive confidence rating patterns between the decision frames. Furthermore, formal model comparisons revealed that the confidence framing effect can be attributed to a simple heuristic that assigns greater weight to the evidence for the chosen alternative compared to that for the unchosen alternative. This computation appears to be based on internal decisional evidence constituted under specific task demands rather than physical stimulus intensity itself, a view justified in terms of ecological rationality. These results shed fresh light on the adaptive nature of human decision-making and metacognitive monitoring.