ecological validity

Ecological validity is a concept from psychology and measurement science that describes how closely the setting, tasks, and conditions of an experiment or test resemble the real situations to which its results are meant to apply. It is used to judge whether findings obtained in controlled settings, such as a laboratory, hold in everyday environments, and it is closely related to external validity, which concerns whether results generalize to other groups or situations.

In AI, the term is used to assess how well benchmarks and evaluation methods reflect real-world usage. It helps explain why scores obtained from artificially constructed test items or constrained conditions may differ from a model's performance or behavior on the tasks of real users.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.