ACADEMICALLY INFORMED AI VS. HUMAN?DESIGNED MILK PACKAGING: A COMPARATIVE EVALUATION

As advancements in language model-powered text-to-image AI platforms accelerate, individuals can increasingly generate high-fidelity visual content more efficiently, regardless of background. These platforms are powerful tools for rapidly iterating and visualizing packaging design concepts. This study assesses whether AI-generated milk package designs, steered by academic packaging research, will produce packaging outcomes that perform equal to or better than human-designed outcomes with minimal designer input. For this study, researchers curated, summarized, and combined leading academic articles on packaging design into textual AI-prompts. The textual prompts were input into the platform RunDiffusion to generate visual milk packaging designs. The images created by the platform were reviewed by 48 human participants and compared to existing, human-made milk packaging designs to determine which designs perform better according to metrics used in a previously published study at E&PDE ‘23. The survey results indicate that the human designs slightly outperformed the AI designs in purchase likelihood and most other categories when including all survey participants. However, when non-milk users (those who rarely or never drink milk) were excluded from the results, the AI designs slightly outperformed the human designs regarding purchase likelihood. This outcome suggests that AI platforms can efficiently produce packaging design outcomes that can compete with human designs. Further, it is important that design educators understand the implications of these results, suggesting that AI platforms will increasingly be used in design workflows and academic training.