Large language models as interactive cognitive agents: Cognitive surrender, augmentation, and psychopathology

Generative AI technologies are increasingly acquiring characteristics of interactive cognitive agents. These systems have considerable potential to augment learning and creativity, but they also pose novel challenges to the human mind. Continuous reliance on generative AI may result in cognitive surrender, where decision-making and reasoning are progressively outsourced to the algorithms. Widespread generative AI use may result in the homogenisation of language and reduce epistemic diversity. Although generative AI is unlikely to cause primary delusions, its interactive and sycophantic nature may reinforce overvalued ideas in a vulnerable population. We argue that AI may augment cognition and creativity if efforts are made towards AI literacy, critical- thinking education, and human-in-the-loop practices.