Deskilling
- Slop
- Research term
Deskilling is the loss of practiced skill, or of room for judgment in a role, after work is reorganized around a tool, a process, or a new division of labor.
When a tool takes over routine cases, people may get less practice and less discretion. Expertise can shift toward system setup and exception handling instead. That may retire work nobody needs, or leave a gap in the team's backup capability.
Where you’ll meet it
Your team's accounting software has used AI to match invoices to payments for a year. At quarter end it can't read a supplier's new invoice format. The newer staff have never matched invoices by hand, and the one person who has is on vacation.
What to do
Use the invoices and payment records to finish the matching with someone who knows the process, if available. Then decide whether matching by hand still has a role in the backup plan. If it does, assign an owner and have newer staff practice on a few invoices each month.
Easy to mix up
Cognitive debt is a proposed later cost of letting AI do your understanding or remembering. Deskilling describes a role that has lost practiced skill, or room for judgment, after the work was reorganized.
Check yourself
Since AI took over drafting customer quotes, your team's quotes go out faster and with fewer errors. A manager says this proves no deskilling happened. Is that right?
Go deeperLabor research and control-room studies gave the idea two forms
Harry Braverman’s 1974 book Labor and Monopoly Capital argued that management can divide skilled work into simpler tasks. Studies of AI at work examine several different outcomes, including task performance and learning unfamiliar skills. Those results do not by themselves establish whether people lose existing expertise over time.
Lisanne Bainbridge's 1983 paper "Ironies of Automation," about plant control rooms, describes a recurring problem: routine operation may be automated while difficult exceptions remain for people. Operators may then have little practice for those exceptions. Monitoring a system that rarely fails can also leave them unready for a takeover.
Sources
- Labor and Monopoly Capital, monthlyreview.org
- Labor and Monopoly Capital, chapters 7 and 8 excerpt, static1.squarespace.com
- Generative AI at Work, arXiv
- How AI Impacts Skill Formation, arXiv
- Ironies of Automation, gwern.net