metarecognition

Metarecognition is a term used in cognitive science and artificial intelligence to describe knowledge about or monitoring of recognition processes. It refers to higher-order awareness of when a stimulus has been recognized, the accuracy of that recognition, and the strategies used to recognize it. It can involve judgments about confidence in a recognition decision, the monitoring of recognition performance over time, and the regulation of recognition strategies (for example, seeking more information when uncertain).

Metarecognition is often considered a facet of metacognition, particularly the metamemory aspect, which concerns memory-related knowledge and monitoring. In recognition tasks, researchers examine familiarity-based and recollection-based processes and how people report confidence or certainty about a prior encounter. Calibration refers to the alignment between confidence and accuracy, an index of metarecognition quality.

In artificial intelligence, metarecognition refers to a system's ability to assess its own recognition capabilities, detect when inputs fall outside its competence, and decide when to defer to human judgment or request additional data. It intersects with meta-reasoning and meta-learning and supports robust decision making, uncertainty handling, and safety.

Measurement methods include confidence ratings, post-decision wagering, and calibration curves in recognition tasks. Applications span education, clinical assessment of memory disorders, aging, and human–computer interaction. There is ongoing debate about terminology, with some scholars using metarecognition and metamemory interchangeably; others preserve a narrower focus on recognition-specific monitoring within the broader framework of metacognition.