Any details on what this means?

Not surprising at all if the LLM users weren't able to recall their own llm-assisted writing, or had less ownership of it.

But it's a bit more surprising that one experimental treatment can have lasting impacts on overall cognitive levels over a four month period, unless those tests were directly related to the experimental treatment.

And, imo, it's also the biggest macro thing to worry about. Of course, LLM-assistance on a task will reduce our understanding of that specific task, but does it free up our time and mental energy to think about other tasks? Is the net benefit over all tasks greater than the cost on our ability to do specific tasks?