The concept of “Cognitive Debt” really resonates – it is remarkably easy to let speed replace scrutiny when deadlines press.
I have been experimenting with applying a conscious pause before finalizing AI-assisted drafts:
1. What did the AI assume that I didn’t specify?
2. What would my independent conclusion have been?
3. Where do the two perspectives diverge?
Taking just 30 seconds to run this check makes a noticeable difference in maintaining real ownership over the outcome.
How are others here building lightweight friction loops into daily workflows to keep critical thinking active?
4 replies
-
-
I like the idea of deliberately introducing friction rather than trying to make every AI-assisted process as seamless as possible.
From my perspective as a researcher and educator, the crucial issue may be whether we preserve what could be called productive cognitive resistance. Learning, interpretation, and critical judgment often emerge precisely because something is difficult, ambiguous, or requires us to formulate a position before seeing an answer.
Your three questions are a good example of a lightweight friction loop. I would perhaps add a fourth: What intellectual work did I avoid by using AI here?
That question matters to me because efficiency is not always the appropriate measure. In some situations, the struggle is part of the learning process itself.
I am increasingly thinking that AI literacy should therefore include not only knowing how to use AI effectively, but also knowing when not to reduce cognitive effort.
Best
René-
I think this is where the conversation about efficiency becomes much more interesting. We tend to assume that reducing cognitive effort is automatically a benefit, when sometimes the effort itself is where learning, judgment and even original thinking are developed.
For me, the distinction may be between unnecessary cognitive load and productive cognitive effort. AI can be incredibly valuable in reducing the first, but if we continuously outsource the second, we may become more efficient while gradually weakening some of the capabilities we actually want to preserve.
Perhaps AI literacy therefore isn’t only about knowing when to use AI, but also knowing when not to use it. Even when we could…
@Rene, I think your fourth question deserves a place in @Kaare´s check.🩵
-
-
I really like the idea of deliberately putting a little friction back into the process, Kaare.
We usually design AI workflows to remove friction, but perhaps some friction is exactly what protects our judgment…I liked especially your second question: “What would my independent conclusion have been?” Because if we stop forming that independent position first, comparison eventually becomes difficult.
There is nothing genuinely ours left to compare the AI’s reasoning against!!And 30 seconds is such a small intervention. It shows that maintaining cognitive ownership doesn’t necessarily require using AI less. It may simply require becoming more intentional about where we pause before accepting its direction.
I’m curious whether that 30-second check has ever made you reject or significantly change an AI-generated conclusion. That would be interesting to hear.😁
very thoughtful