Semi-Automatic Wizard-of-Oz

A hybrid Wizard-of-Oz study methodology in which part of a prototype system is implemented with real automation while a human researcher (the 'wizard') intervenes only for components that are not yet reliable enough to run autonomously. For example, an AI agent may generate responses automatically while a wizard manually converts noisy speech input into clean text to avoid speech-recognition errors masking the phenomenon under study. This approach preserves experimental realism while isolating the behaviour of the mature subsystem, making it useful for evaluating emerging assistive technologies before full end-to-end reliability has been achieved.