AI-Assisted Coding Short-Circuits the Creative Process: Study on LLMs and Programmer Creativity

Forum topic · 二一 · 2026-05-14

Summary

A study of 20 programmers found that using large language models (LLMs) significantly compresses the creative process in programming. Participants completed two sets of coding tasks, one with LLM assistance and one without. Researchers measured the creative process itself rather than code quality. Results showed that LLM use significantly shortened the idea-generation phase (p=0.0004), with programmers jumping to execution faster, and significantly reduced 'aha' moments of creative insight (p=0.002). LLM-assisted code was more correct and runnable, but the total quantity of creative ideas remained roughly the same. The authors argue that LLMs do not make programmers less intelligent—rather, by providing answers too readily, they remove the productive friction and struggle that drive divergent thinking and better design. Four human-LLM collaboration patterns were identified. The paper calls for generative AI tools designed to support the creative process, such as prompting users to compare alternative approaches before accepting the first solution. Source: Saghi, Huang, Chattopadhyay, 'Like Taking the Path of Least Resistance', arXiv:2605.13776.

"Creativity is fundamental to being human." That is the opening claim of the paper, followed by a simple but brutal experiment.

Twenty programmers each completed two sets of programming tasks — one with LLM assistance, one without. What was compared was not code quality (that has been studied to death), but the creative process itself.

Results:

With an LLM, the idea-generation phase shrank significantly (p=0.0004). Programmers jumped to the "execution" stage much faster.
Creative moments dropped significantly (p=0.002). Those "wait, there's a better way to do this" flashes of insight became rarer.
LLM-assisted code was more correct and more runnable — but the number of creative ideas was roughly the same.

The researchers identified four human–LLM collaboration patterns, supporting different problem-solving strategies. But the overall trend is clear: LLMs help programmers produce functional code faster, at the cost of skipping the divergent-thinking phase that might have led to better designs.

One participant summed it up: "Like taking the path of least resistance."

The paper's finding is not that LLMs make programmers dumber. It's that LLMs are too good at handing you the answer. When you have a tool that instantly says "just write it this way," you rarely feel the urge to ask "is there a better way?" Human creativity is largely driven by friction — being stuck on a problem, forced by difficulty, compelled to take detours. LLMs eliminate much of that friction. But friction is exactly where creativity thrives.

The design implication is clear: generative AI tools should be designed to "support the creative process" rather than merely "produce correct results." For example, after presenting a solution, the tool could ask "have you considered an alternative approach?" — or require users to compare multiple ideas before accepting the first one.

Reference paper: Zeinabsadat Saghi, Run Huang, Souti Chattopadhyay. "'Like Taking the Path of Least Resistance': Exploring the Impact of LLM Interaction on the Creative Process of Programming." arXiv:2605.13776, 2026.

AI-Assisted Coding Short-Circuits the Creative Process: Study on LLMs and Programmer Creativity

Summary

Tags