The ability to judge one’s own confidence is a core ability of human metacognition. Giving informative confidence ratings is crucial in many situations: Humans making decisions, either individually or in groups, rely on their own estimates of uncertainty. But finding adequate measures for the quality of confidence ratings is a challenge. The two major approaches to tackle this challenge will be contrasted in this symposium: model-based and model-free. On the one hand are computational process models of the formation of confidence ratings in humans. The first speaker, Matthias Guggenmos, provides an overview and categorization of these models. One of them is the most prominent model, which is an extension of classical signal detection theory (where perceptual sensitivity, d’, is measured). This metacognitive extension analogously measures metacognitive sensitivity, meta-d’. Together with nine others, this prominent model is evaluated on a collection of 13 experimental data sets by the second speaker, Manuel Rausch. His results should concern researchers in the field: The meta-d’/d’ model does not provide satisfactory results. The third speaker, Simge Hamaloglu, drills deeper into the model's mechanisms: As in classical signal detection theory, the meta-d’/d’ model estimates (metacognitive) criteria that determine the point where low turns into high confidence. She focuses on these criteria to differentiate when a stimulus is actually perceived versus only inferred from other cues. Contrasting these model-based approaches, on the other hand, classical information theory has inspired approaches to measuring metacognitive ability in a model-free way. The fourth speaker, Sascha Meyen, introduces this idea in which metacognitive ability is measured as transmitted information (in bits). Taken together, this symposium will pinpoint the contention between model-based and model-free approaches to measuring metacognitive ability. It will highlight challenges in terms of empirical fit and interpretability, and thereby guide future development of both approaches in tandem.