Deskilling and overreliance: Two long‐term risks of AI assistance, and what designers can do about them

AI assistants are increasingly capable of handling tasks on behalf of users, but this delegation introduces two underappreciated long‐term risks. Deskilling occurs when users lose competence in delegated domains through lack of practice. Overreliance occurs when users accept AI outputs uncritically, amplifying errors rather than correcting them. This paper presents empirical evidence for deskilling, describes a controlled design intervention that reversed it in a navigation task, and proposes the prioritization of the “Human Goes First” cognitive forcing function to promote appropriate human‐AI collaboration and combat deskilling. The implications extend beyond usability to AI safety and trustworthiness.