New Paper Warns AI Adoption Erodes Professional Expertise
A new research paper warns that rapid corporate AI adoption is creating a "tragedy of the cognitive commons" that could wipe out deep human expertise in fields like software and law.
Nolan Lovett of the NATO Special Operations University has published a paper in the Human Resource Development Review outlining how individual corporate AI adoption threatens collective professional expertise. Lovett argues that when companies automate entry-level roles, they capture immediate efficiency gains but deplete the shared talent pool. This "tragedy of the cognitive commons" prevents junior workers from gaining the foundational experience needed to develop deep domain knowledge. The long-term consequences of entry-level cuts starting in 2023 may not fully manifest until between 2030 and 2045.
Recent research supports this warning. A Federal Reserve Board study found that growth in programming jobs has nearly halved since ChatGPT launched, while a March 2026 Anthropic study noted a 0.5 percentage point drop in the job-finding rate for workers aged 22 to 25 in highly exposed fields. Furthermore, cognitive costs are becoming measurable. An MIT study using EEG scans showed that brief AI use weakened neural connectivity, leaving over 80 percent of participants unable to recall their own AI-assisted writing. Another Anthropic study found software developers with AI access scored 17 percent worse on knowledge tests, particularly when using the technology as a mere answer generator.
The cognitive decline is especially visible in educational settings. A Swiss study of 666 participants linked AI use to reduced critical thinking, particularly among young adults. In China, students using AI saw homework grades rise by 18 percent, but their exam scores dropped by up to 24 percent over a two-year period. For practitioners in highly vulnerable, lightly regulated fields like software engineering, financial analysis, and legal research, these findings suggest that over-reliance on AI will erode the very expertise required to validate AI outputs. To counter this, Lovett recommends implementing AI-free training environments and testing human competence before AI tools are introduced.
This is our own summary of reporting by The Decoder
