Sidra

Metacognitive thinking will become increasingly important for effective collaboration with AI systems. We identify four characteristics that differentiate human-AI interactions from human-human interactions. These are (1) the accuracy of our cognitive heuristics for predicting the behaviour of AI systems, (2) the AI’s limited capability when dealing with novel and ill-defined problems, (3) the lack of a natural, reciprocal feedback mechanism in AI systems and (4) the inability of AI systems to engage in metacognition. Drawing upon the dual-process theory of human thought process, we argue that these characteristics will diminish the efficacy of the system one mode of human thinking, which relies on cognitive heuristics.