Metacognition is mentally demanding

The ability to self-evaluate performance derives from metacognition, enabling us to monitor and control our behaviour. Metacognition is generally thought to accompany decision making automatically and effortlessly, with inefficiencies in metacognitive judgment attributed to intrinsic properties of the computations that support it. Contrary to this view, we show that metacognitive judgments are mentally demanding. Apparent inefficiencies can arise from avoidance of metacognitive effort when people engage in rational trade-offs between the costs and benefits of metacognition. We developed a flexible, effort based decision making paradigm that allowed participants to trade-off monetary rewards to make metacognitive judgments easier. First, individuals sacrificed rewards to simplify confidence self-evaluation, indicating that confidence ratings (a canonical metacognitive judgment) are mentally effortful. Second, we show that incentivising self-evaluation reduces confidence leaks, a well-characterised bias in metacognitive judgments widely regarded as a persistent metacognitive inefficiency. This result suggests cost-benefit trade-offs in metacognitive effort contribute to this phenomenon. Through model simulation within a signal detection framework, we show that stability of confidence criteria is a computation that can account for these trade-offs. Our results challenge the assumption that metacognitive judgments are effortless and motivate a re-evaluation of metacognitive inefficiency, especially in populations where metacognition is presumed to be inherently deficient.