The most common fear about AI creative tools is that they'll turn us into passive consumers, copying whatever the machine spits out. A new study from Swansea University suggests the opposite is happening.
Researchers had over 800 participants design virtual cars using an online tool. Some worked alone. Others got access to AI-generated galleries showing a range of design possibilities, from high-performing options to deliberately weird and broken ones. The AI used a technique called MAP-Elites, an evolutionary algorithm that generates a wide spread of diverse solutions rather than converging on a single "best" answer.
The results, published in ACM Transactions on Interactive Intelligent Systems, were clear across the board.
People Worked Harder, Not Lazier
Participants who saw the AI galleries spent more time on their designs, not less. They produced higher-quality results. And they reported feeling more involved in the creative process. That last point matters because it cuts against the "AI makes people passive" narrative. These participants weren't copying the AI's suggestions. They were using them as springboards.
"When people were shown AI-generated design suggestions, they spent more time on the task, produced better designs and felt more involved," said Dr. Sean Walton, Turing Fellow and Associate Professor of Computer Science at Swansea, who led the research.
Bad Examples Were Actually Useful
The most counterintuitive finding: showing people flawed and unusual AI designs actually helped. Rather than confusing participants, the bad examples prevented what psychologists call "fixation," the tendency to lock onto your first decent idea and stop exploring. Seeing a range of outputs, including failures, encouraged people to take more creative risks.
This has practical implications for how AI tools should present suggestions. Most AI products today show you their best guess. This research suggests that showing you a diverse spread, including options the AI knows are imperfect, produces better human output.
What This Means for AI Tool Design
The study focused on a specific domain (car design) with a specific AI technique, so we should be careful about overgeneralizing. But the core finding aligns with what many daily AI users already sense intuitively: AI works best not when it gives you the answer, but when it gives you a starting point to react against.
For anyone building or choosing AI creative tools, the takeaway is straightforward. Tools that show diverse options, even imperfect ones, may produce better creative outcomes than tools that try to nail the "right" answer on the first try. The messy middle turns out to be where human creativity thrives.