How Can I Use AI Without Losing My Ability to Think for Myself?
How can I use AI without losing my ability to think for myself? You can use AI while continuing to think for yourself by keeping your own judgment in the work. Attempt an explanation before asking for one. Check what the system gives you. Practice the abilities you want to retain. And before accepting an answer, be able to explain why you accept it.
That sounds straightforward.
Then the machine hands you something beautiful.
And suddenly your own rough attempt looks like it should apologize.
I’ve been thinking about this through drawing, which is where a surprising amount of my thinking ends up these days.
I could already draw. I could imagine a scene. I understood color, atmosphere, and what I wanted someone to feel when they looked at it.
Getting the whole thing to hold together on the page was less reliable. AI helped with that. I could describe an image and get back a composition close to what I had been carrying around in my head.
There it was. Cleaner than my drawing. More polished. Already finished, apparently.
Except I still wanted to draw.
So I brought the image back through my own hand. First with pencil and a light box. Later through digital illustration.
I removed things. Simplified backgrounds. Changed the emphasis. Let a line be a little less perfect because I liked what happened when it was mine.
Over time, I found myself reconstructing more of the work independently. That does not prove AI improves artistic skill or prevents skill loss. It is what I noticed in my own practice.
And it gives me a better question to ask than whether the machine made a good picture.
What was happening to my ability to make one?
That question sits at the center of my practitioner theory, Human-in-Command: how do I use assistance while staying responsible for the work and attentive to what I am learning?
Its practical framework, A.R.C.H.I.E.™, names six stages: Anchor, Recognize, Construct, Hold, Implement, and Evaluate. They give me a way to look at what I am doing while the assistance is happening.
Keep some of the thinking yours to do
A finished output does not tell you how much you understood along the way. You can have an excellent explanation sitting in front of you and still struggle to explain it yourself.
Anchor begins with making your own starting point visible. Before asking AI to write the argument, try stating what you believe and what supports it. Before asking it to solve the problem, make an attempt. Give yourself something to compare with the answer.
The attempt can be rough. We seem to have become very impatient with rough. But a rough attempt can show you exactly where you need help.
Recognize means working out what kind of difficulty you are actually facing. In my drawing, that was structure. I already had an image in mind and a sense of its emotional tone. Recognizing the particular difficulty let me use assistance there while continuing to make the other choices myself.
You can make a similar distinction when writing or learning. Do you need information? Help arranging it? Someone to question the conclusion you have drawn?
Construct is where you decide how to work on that difficulty and what role AI will play. For a drawing, that might mean deciding to use AI for a starting composition while leaving yourself the choices about what to remove, simplify, or emphasize. For a difficult reading, you might plan to write your own explanation first, ask AI where it sees gaps, then check those gaps against the original text. Before beginning, you have decided which parts of the thinking you will practice yourself.
Stay long enough to decide what you believe
Then comes Hold.
I think this is where a very ordinary temptation enters the room.
The answer is here.
It sounds reasonable.
You would quite like to be finished.
Hold means remembering what you set out to do when something makes you want to change course. Suppose you asked AI to find weaknesses in your argument, and it returns a flattering rewrite. It reads beautifully. You still need to know whether the argument holds up. Hold means keeping that question open even when you like the answer you received.
The same pause matters when an answer sounds convincing. Has something changed that gives you a reason to accept it? Or are you simply relieved to have an answer?
Suppose AI tells you a survey proves audiences want shorter performances.
You open the survey.
It asked about ticket prices.
The citation exists. The conclusion still has not been established.
That moment of comparison is work you can practice. Asking AI to produce another explanation can help you investigate, but asking the same system to critique itself does not independently verify its answer.
Before accepting a recommendation, try finishing this sentence:
“I am choosing this because…”
What comes after “because” matters.
Can you explain the evidence? The tradeoffs? Why this choice fits what you are trying to do?
If you cannot, you have found a place where the thinking needs more attention.
Take it back into your own hands
The last two ARCHIE stages are Implement and Evaluate.
Implement is where you carry out the approach you chose. Evaluate is where you examine what happened and decide what needs to change on the next attempt.
For me, that meant drawing the image and examining the result. Over repeated attempts, I could look at what I was changing, what still needed support, and what I could increasingly do myself.
For you, it might mean closing the AI conversation and explaining the idea in your own words. Writing the next paragraph yourself. Trying a similar problem without the generated solution beside you.
You get to decide which abilities deserve that practice. You do not have to turn every convenience into a test of character.
But if an ability matters to you, leave room to use it.
The next time AI helps you finish something, put its answer aside for a moment.
See what you can carry into the next attempt.
In my case, there was still a drawing to make.
And I still wanted to be the person holding the pencil.
Comments