Thinking with Machines (Part 3)

What I am Observing in My Own Practice

My original question asked “How do humans maintain and develop cognitive capability in AI-mediated environments?”

In my own practice of “thinking with AI”, what I have observed is a balancing act. However, it’s not as much balancing the amount of AI use vs. my own independent thinking as it is finding the best uses of AI that preserve authorship and higher cognitive functions.

This post will make it personal: What have I noticed when attempting to think with AI in a deliberate way?

1. AI is changing how I think.

I no longer think completely alone. I can bounce ideas, ask for feedback, essentially get another perspective. Many intellectual activities that once happened entirely inside my own head now involve dialogue with AI.

2. Cognitive labor is being redistributed.

I spend a lot more time with evaluation, decision-making, metacognition and less time searching.

I’m not doing less thinking. I’m doing a different kind of thinking. Less searching, less synthesis, less initial organization; more judgment, evaluation, direction-setting, metacognition.

In the context of a learning sprint, I set the direction, I identify the relevant questions. A combination of AI-led and more traditional research identifies relevant resources. AI may help synthesize. I curate relevant resources and annotate. AI helps identify key concepts and patterns across notes much more efficiently than I ever could. I often validate connections across concepts myself with a concept map. AI helps with definitions of concepts but I annotate with connections based on my own experience. I draft a Substack post and I ask AI for feedback.

3. I think more about process than I used to.

I regularly have conversations around boundaries, authorship, workflows, learning sprints, and prompting practices that go well beyond “writing good prompts”. In other words, using AI has made me more conscious of how I think. I am often questioning my approach, attempting to create clear processes, boundaries, but staying flexible. I play a very active role in thinking before I prompt, bringing in my own experience, my own judgment, setting intention and direction.

4. Maintaining authorship matters.

Early on, I allowed AI to generate content for me. Now, I write first drafts. If I’m completely stuck, AI can help me get unstuck but asking AI to write something for me from scratch has never worked well enough and I don’t like editing AI drafts. I much prefer writing my own first draft and getting AI to react to it. Beyond the personal preference for this process, there is an element of required authenticity that comes with maintaining authorship.

5. I still believe augmentation is possible.

My experience has not led me to conclude that AI inevitably weakens cognition.If anything, it has made me more reflective about how I learn, write, and think.Whether that ultimately qualifies as cognitive augmentation remains an open question, but it is enough to keep me exploring. In fact, I continue to develop and implement challenging learning sprints, I maintain a human-first writing routine, and I design and teach new courses related to AI and technology.

As someone who teaches AI to older adults, these observations matter.If the way we use AI influences cognition, then teaching people HOW to use AI may be just as important as teaching them what AI can do.The older adults who sign up for my AI-related classes may or may not be aware of the controversy around the use of AI on cognitive functions.The next post will explore what this all means for teaching AI.