AI Creativity Works Best in Moderation, Study Suggests

AI can help creativity, but only up to a point

Artificial intelligence is often framed in extremes when creative work is concerned. In one version, it is a threat that could flatten originality into algorithmic averages. In another, it is a powerful collaborator that can unlock ideas people would not have reached on their own. A report in New Scientist points to a more measured conclusion: both views may be partly right, depending on how much people rely on the tool.

The article centers on research by Hsuan-Che Brad Huang during his PhD at the University of British Columbia in Canada. The core finding is straightforward. People appeared to produce the most creative ideas when they used AI in moderation, not when they avoided it entirely and not when they leaned on it too heavily. The result suggests there is a practical middle ground for writers, designers, marketers, researchers, and other knowledge workers now trying to figure out where generative AI fits into their daily process.

That middle ground matters because creative work is rarely just about output volume. It also depends on surprise, judgment, ownership, and the ability to push beyond familiar patterns. If AI becomes too dominant in the process, the report suggests those qualities can weaken even when productivity appears to rise.

A “Goldilocks zone” for AI use

New Scientist describes the finding as a “Goldilocks zone” for creative AI use. In practical terms, the idea is that a small or selective use of AI can introduce fresh prompts, alternative framings, or unexpected combinations that help a person move past habitual thinking. But when the tool supplies too much of the substance, it can start to narrow rather than expand the creative process.

That conclusion fits a broader theory of how creativity works. Human thinking is shaped by experience, assumptions, and habits. Those constraints can be useful, because expertise helps people recognize quality and structure. But the same constraints can also trap people inside familiar routes. The value of an outside perspective, whether it comes from a colleague, a team, or a tool, is that it can jolt a person into considering ideas they would not have generated alone.

In that sense, AI can function less like a replacement for imagination and more like a mechanism for perspective shifting. It can propose plot directions, conceptual links, or thematic angles that interrupt routine thinking. The report argues that this benefit is strongest when the human user remains the active decision-maker rather than becoming a passive selector among machine-generated options.

Why too much AI can reduce creative quality

The same report also outlines the limits of large language models in creative settings. These systems are statistical tools that generate likely responses based on patterns in training data. That makes them useful for ideation, but it also means they tend toward familiar, blended, or average-seeming outputs. They can mirror common structures competently without reliably producing the kind of unusual, personal, or deeply idiosyncratic leap that people often associate with memorable creative work.

There is also a psychological cost. According to the New Scientist piece, heavy AI use can undermine a person’s sense of competence and ownership. If too much of the work arrives pre-shaped, the user may feel less connected to the result and less motivated to push it further. That can produce a kind of creative passivity: the person stops exploring and starts curating.

This distinction is important because many early AI workflows reward speed over depth. Asking a model to generate ten ideas in seconds feels efficient. But if those ideas become a ceiling rather than a starting point, the process may quietly trade away experimentation. The research highlighted by New Scientist suggests that over-reliance is not just a philosophical concern. It may directly weaken the originality people are trying to improve.

Testing the idea in practice

The article uses a simple writing exercise to make the point concrete. New Scientist columnist David Robson asked ChatGPT for film concepts based on a prompt about a shattered wine glass and a hidden memory card. The result, in his telling, was serviceable rather than remarkable. That is part of the lesson. AI can be useful for getting unstuck, but usefulness should not be confused with creative authority.

The practical takeaway is not that creative workers should avoid generative tools. It is that they should place them carefully inside a larger process. A writer might use AI to generate unexpected angles and then develop the narrative independently. A product team might use it to surface alternative framings before evaluating them through domain expertise. A designer might use it to challenge assumptions while preserving control over the final concept.

These approaches share a common principle: AI contributes stimulus, not final judgment. The person remains responsible for deciding what is interesting, what is derivative, what fits the audience, and what deserves to be discarded.

What this means for creative work now

The debate around AI and creativity often assumes a single answer for every field, but the New Scientist report points toward a more nuanced reality. Different tasks likely have different thresholds. Brainstorming may benefit from more model interaction than final drafting. Early ideation may tolerate generic suggestions better than brand work, fiction, or research communication, where voice and precision matter more.

Even so, the broader implication is clear. Organizations adopting AI for creative tasks should not measure success solely by how much labor the tool can absorb. They should also consider whether workflows preserve human agency, curiosity, and ownership. If the aim is better ideas rather than merely faster text, then moderation may be a feature rather than a compromise.

• Selective AI use can help break habitual thinking patterns.
• Excessive reliance may push work toward average outputs and weaker ownership.
• The strongest results may come when humans use AI for prompts and perspective, then shape the work themselves.

That balance may prove to be the most durable lesson from the research. The question is not whether AI belongs in creative work. It is how to use it without letting convenience hollow out the very human capacities that make creative work worth doing in the first place.

This article is based on reporting by New Scientist. Read the original article.