The gap i see is the definition of "AI use" is not clearly delineated between passive (usage similar to consumption) vs active.

Passive AI use where you let something else think for your will obvious cause cognitive decline.

Active use of AI as a thought partner, and learning as you go yourself seem to feel different.

The issue with studying 18-22 year olds is their prefrontal cortex (a center of logic, will power, focus, reasoning, discipline) is not fully developed until 26. But that probably doesn't matter if the study is trying to make a point about technology.

The art of learning fake information from real could also increase cognitive capacity.