In my previous essay, I explored my process of working with AI as both a software engineer and a musician, and the subtle sacrifices I ended up making. In this essay, I’m going to dig deeper into the impact of using AI for creative work.

Why do we create?

There are many theories, and the one that resonated with me was Donald Winnicott’s idea of creative apperception. According to Winnicott, we create to establish a sense of self because the alternative is compliance, and compliance makes life feel futile. If we approach anything in a way where we’re a participant rather than just a recipient, we’re operating in what Winnicott calls the potential space. Playing inside this space can be intrinsically rewarding because it preserves the feeling of being actively involved in one’s life.

What do we get from the process?

For one of my songs titled “Ghost In The Sea”, I remember I spent about two days just recording guitar solos. Some I liked, some I didn’t, some had one phrase worth keeping buried inside something that wasn’t. After giving myself some distance from the song, I went back with fresh ears to listen to the solos and pull out the sequences that ended up in the master. Would I trade this process for forty AI-generated solos? No way. To me, the creative process is a means of expression, flow, escape. I’m not usually chasing the result. I’m more interested in experiencing the process. That experience is what makes me feel most alive and most human. There’s a pressure that builds when perception, emotion, and curiosity have nowhere to go. When I bring into reality something that I imagined or felt, that pressure is released. I often end up learning something about myself or unlocking something in my mind along the way.

In 2019, the World Health Organization (WHO) published an extensive report on the role of the arts in improving health and well-being. Playing an instrument builds resilience to cognitive decline. In people with neurological disorders like Parkinson’s disease, dance has been found to improve balance and functional mobility. Drawing, painting, and crafting are linked to a reduction in stress. A 2024 study found that creative activity improves mood specifically because it satisfies three psychological needs from Self-Determination Theory (SDT): autonomy, competence, and relatedness. We’ll come back to the first two.

Do we get those same things when we introduce AI assistance?

When I think about my process for making music, ideas usually come from this autotelic state where I’m drifting off into space with no goal in mind. When I discover something interesting, I take that seed and try to cultivate it. That’s when the work begins. That’s when the rollercoaster of frustration and bliss starts moving. I push through the frustrating part because I’ve learned that the reward for the “that’s it” moment pays off. The moment when I’m able to articulate what I was feeling and bring into reality something that was inside.

This goes back to Winnicott’s potential space. This space is created by the friction between our inner fantasy and outer reality, and it’s what allows for play and creativity. With AI assistance, we often get solutions that prematurely resolve the creative paradox and collapse that space into non-existence. The potential space is the psychoanalytic arena where creation happens, while a sense of agency is the neurocognitive mechanism that allows us to inhabit and manipulate that space.

Researchers who study sense of agency split it into two parts. Feeling of agency (FoA) is modeled by the subconscious process where our brain is matching what it predicted would happen against what actually happened. Judgment of agency (JoA) is the conscious process of attributing an output to oneself or another. I think these map surprisingly well to the two psychological needs pinned above: JoA is to autonomy what FoA is to competence. I move my fingers, strum a chord, hear the sound, and register that it was me playing the guitar. FoA and JoA are aligned. My brain doesn’t need to pause, verify, or question who made the stroke. Why does this matter? A strong sense of agency leads to better mental well-being, and alignment between FoA and JoA supports the conditions for flow states. And who doesn’t want to be in a constant flow state?

When I create with AI assistance, FoA and JoA are usually not aligned, and that reduces the chances of me getting into any flow state. I describe an image at a high level in a prompt to an image generator, press enter, and wait. The sensorimotor process behind the prompt does not align with how the image was created—by a computer, not by me. The cycle of prompting, waiting, analyzing the output, re-prompting draws me into a monitoring state where I’m constantly just trying to decide how much of the work belongs to me versus the probabilistic generator.

If we’re reducing the size of the potential space, decreasing our sense of agency, and limiting our ability to enter a flow state, it seems unlikely to me that we’d reap the same benefits reported by WHO. Perhaps they’ll have to conduct another survey as the integration of AI technology into society evolves. If we consider the SDT study, autonomy and competence are much more volatile in AI-assisted work, depending on how it’s conducted.

Should we use AI in the creative process?

AI is trained on human production. We’re also trained on human production by the music we listen to, the books we read, the art we encounter. So why does it matter where an idea comes from? A study in Scientific Reports found that our sense of self-efficacy is higher when we define the idea and use AI to help refine it (rather than the reverse). What if no ideas come to you and you have a deadline? As Chuck Close put it, “inspiration is for amateurs, the rest of us just show up and get to work.” With enough focus and dedication, you’ll produce something. AI can help you get to something you were already reaching for but couldn’t quite grab. You can use it to learn — to understand why a chord fits a song, to see how a sentence could be restructured, to explore variations on an idea that started with you. As long as you’re engaging, asking, deciding, discarding, you’re doing the work. But be mindful about putting something out just for the sake of producing “content.” You’ll make it harder for the people who put in the time, the effort, the love, the energy to be heard. If creation is about establishing a sense of self, then consuming art is about witnessing another person’s self. If the “self” wasn’t present in the creation, what is there to connect to?

So…does this still feel like creating?

I haven’t really answered the question in the title. I think only you know the answer. I used AI to help edit this essay, so I’ve spent more than a thousand words evaluating a cost and then paid it anyway. I did the research and came up with the ideas. The clear presentation of my ideas mattered more to me than if I came up with every word myself. In an essay, I’m ok with this. In a song, I’m not. I’m not really writing a song for anyone but myself so clarity doesn’t matter and every word carries a weight I’m not willing to outsource. If you’re using AI to aid with the creative process, only you can know whether the result is true to you and whether it releases the pressure that was building.