Same Tool, Two Counterparts

A scene that plays out in every studio. At the same screen sit an intern and a graphic designer with twenty years of practice. The same program suggests patterns to both. She clicks through six variants and beams. He clicks back three times, changes a detail, then another, and looks increasingly put out. It is tempting to dismiss this as a matter of character, curiosity versus stubbornness. A new study suggests something else is going on.

Joana Rovira Martins, Pedro Martins and Ana Boavida (2026) had 30 people work with PatternPursuit, a system that breaks down works by the Portuguese graphic designer Sebastião Rodrigues into graphic raw material and turns it into patterns for wrapping paper (p. 2). Fifteen were experts with more than five years of professional practice. The other fifteen formed a mixed group of junior designers and people from unrelated fields (p. 9). Important for what follows: PatternPursuit is not an image generator that turns a sentence into a finished picture. It recombines existing material step by step (p. 17).

The two groups used the same tool and experienced two different ones. The non-experts tried out more ideas, mostly four or five or more, before deciding. The experts mostly stayed with two or three and went back more often to change things (p. 13). Twelve of the fifteen experts described the system as semi-autonomous; nine of the fifteen non-experts saw it as something they directed. The non-experts also rated their control and their freedom of choice higher (p. 13). For them, the authors write, the system extends their abilities; for experts, it is “a semi-autonomous partner that challenges authorship” (p. 3). One expert: “I felt that it controlled many of the images it gave me, so I felt I was a bit limited” (p. 16).

On the Creativity Support Index, the non-experts scored 83 points, the experts 71. The non-experts rated their own patterns higher for quality and value; on novelty, the difference was not significant (p. 14). But these are self-assessments, and the sample is small, as the authors themselves note (p. 17). The correlations are more telling. Among the non-experts, perceived novelty and perceived quality went together, and the authors conclude that they conflate originality with quality. Among the experts, the value they gave a result was linked to the control they had over it (p. 16). The experts experience the machine more as a counterpart, and are nonetheless, or perhaps for that very reason, less satisfied.

Other studies measure differently and arrive at a more sober picture. Wang and colleagues (2025) had 21 students and 21 designers with more than three years of professional experience design a character, first conventionally, then with generative tools. Three external industry experts rated the results. For the novices, AI helped mainly with generating ideas; for the experienced, rather with quality and refinement, and the experienced stayed ahead (p. 13). Because everyone worked first without AI and then with it, practice effects cannot be fully separated (p. 7). And Chulvi and colleagues (2026) found a significant difference between image generators and students only for variety, not for novelty or quality (p. 8).

Which leaves the question of what the dissatisfaction means. Lee and colleagues (2025) surveyed 319 knowledge workers. Those with more confidence in AI reported thinking less critically. Those with more confidence in themselves did so more, even though they found it more effortful. This does not establish causation, and here, too, respondents are describing themselves. Work shifts, the authors argue, from execution to oversight: to checking and integrating. So the graphic designer’s grumbling can be read differently: not as resistance to the tool, but as judgement still at work.

Bahaa Mustafa and Abeer Ibrahim (2026) describe what is at stake for the young in a review of 33 sources (p. 1). They distinguish three stances graphic designers take towards generative AI: Resisters oppose it, on ethical grounds among others; Adapters use it while holding that judgement and invention remain human; Integrators rebuild their professional identity around co-creation with the machine (p. 6). Experienced designers are probably mostly Adapters, and it works: what AI takes over was never the heart of their work and does not carry their self-image (p. 7). At the start of a career, that very routine is where learning happens. Three days on a logo teach balance and proportion; thirty seconds with a prompt do not (p. 7). Those who never practised making things themselves cannot later judge what the AI delivers (p. 8). The typology has yet to be tested against data (p. 9).

For a studio, this is a practical question, in practice and in teaching alike. The enthusiasm of the younger is no proof of good work, and the hesitation of the older no evidence of stubbornness. So assess not the feeling at the end but the reasoning before it. Whoever picks the fifth pattern says in one sentence why not the fourth. And before the machine makes a suggestion, what is non-negotiable is fixed: material, format, occasion. At SOL LAB, we call this language before spec. Then the intern has something to test variants against, and the graphic designer something that gives him back control.

The tool is the same. What it does depends on who sits in front of it and what that person already knows. A studio that only measures how satisfied everyone is mostly measures experience.

Sources

Chulvi, V., Ruiz-Pastor, L., Berni, A., Royo, M., & Castillo-López, A. (2026). Creative potential of image-generative AI models for conceptual engineering design tasks. Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 40, e8. https://doi.org/10.1017/S0890060426100286 (full text)

Lee, H.-P., Sarkar, A., Tankelevitch, L., Drosos, I., Rintel, S., Banks, R., & Wilson, N. (2025). The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (CHI ’25) (pp. 1–22). ACM. https://doi.org/10.1145/3706598.3713778 (full text)

Mustafa, B., & Ibrahim, A. (2026). Professional identity of graphic designers in the age of generative AI: Threats, adaptation, and a conceptual typology. American Journal of Arts and Human Science, 5(4), 1–11. https://doi.org/10.54536/ajahs.v5i4.8133 (full text)

Rovira Martins, J., Martins, P., & Boavida, A. (2026). Understanding how expertise shapes human-AI collaboration in graphic design with PatternPursuit. Design Science, 12, e34. https://doi.org/10.1017/dsj.2026.10078 (full text)

Wang, N., Kim, H., Peng, J., & Wang, J. (2025). Exploring creativity in human–AI co-creation: A comparative study across design experience. Frontiers in Computer Science, 7, 1672735. https://doi.org/10.3389/fcomp.2025.1672735 (full text)