Delaware’s AI Commission had a public meeting Oct. 23, as state agencies, educators and employers consider how artificial intelligence should be used. The state has also expanded access to AI and digital skills training.
These efforts deserve one common measure: Can people recognize an AI-generated error and correct it before someone else pays the price?
Imagine a public service employee reviewing an AI-drafted explanation of a benefits rule. Give the employee a simulated case with a misleading answer and ask him or her to identify the authoritative source, mark the error, explain the correction and escalate uncertainty to the appropriate human decision-maker. The case should contain no real applicant’s information.
The agency could publish anonymized results for these exercises: how many mistakes staff noticed, how quickly they found the source and what changes followed. This would give the commission and residents something more concrete than reports of how many employees completed a course.
A technology demonstration can look flawless when nobody introduces a wrong answer. Delaware’s public institutions should deliberately test the points where the technology fails.
Successful adoption should mean faster service backed by the ability to correct a mistake, with clear responsibility when the answer affects a resident’s life.
Dr. Gleb Tsipursky
CEO, Disaster Avoidance Experts
Columbus, Ohio
Reader reactions, pro or con, are welcomed at civiltalk@iniusa.org.
