Millions of people are now asking chatbots to summarize books, draft emails, and even explain political events to them. But what looks from one perspective like a productivity revolution may also be something more discomfiting: the quiet outsourcing of judgment itself.

Writers on artificial intelligence have long claimed that it poses an existential risk because, for example, it may become so powerful that it turns against human beings. But AI may create a different kind of existential risk, as philosopher Nir Eisikovits notes — not in the apocalyptic sense often imagined but in relation to the question of what it means to be human. One of the most underestimated dangers of these systems lies in the growing tendency for users to delegate the task of forming judgments to the algorithmic outputs of chatbots, thereby risking the gradual erosion of our capacity for independent thought.

The negative side effects accompanying the use of large language models (LLMs) are vividly illustrated by the phenomenon of “cognitive debt.” From an economic perspective, the short-term productivity gains achieved through the use of AI systems are difficult to dispute. By delegating numerous tasks previously performed by humans to AI, significant efficiency gains can be observed: workflows are accelerated, processes are rationalized, and organizational routines are overall made more efficient.