Generative AI and Cognitive Offloading

Abstract Contemporary generative artificial intelligence software poses a threat to the development of critical thinking skills that sets it apart from the litany of prior technological developments that have raised similar concerns. This article adopts a metacognitive perspective on critical thinking in order to show that AI poses a risk of a critical thinking “death spiral” where AI use degrades critical thinking skill and thereby induces further harmful use. While the evidence on the harmful effects of cognitive offloading on critical thinking is not yet definitive, this article supports the case that AI risks undermining critical thinking abilities by appealing to two lines of evidence. The first is the alignment of features of contemporary AI software with psychological biases that will incline us to trust AI more as our reasoning capacities are degraded. The second draws on social epistemology to show that in the case of expertise identification, consideration of opposing perspectives, and use of evidential “stoppers,” AI undermines social practices and conventions that are constitutive of critical thinking. In light of these considerations, a precautionary principle requires that we do not teach AI skills as part of critical thinking skill development, at least until such time as we have stronger empirical evidence on how to do so without eroding the metacognitive capacity to recognize the right contexts for using rational strategies.