Pedagogy and Professional Development

Accuracy of Image-generative AI in Interpreting Images Generated from Prompt-Based Inputs

---

Abstract

Purpose: By utilizing ChatGPT Dall-E 3, this research aims to meticulously evaluate the accuracy of images created by AI in the context of a visual merchandising class.

Method: 126 window display images were input into Dall-E 3, using two different prompts to investigate the image and pick one element of design and one principle of design.

Findings: Only 51 responses were entirely correct (20.24%), 124 responses (almost 50%) were half correct, and the remaining 77 responses (30.56%) were incorrect. Significant challenges were identified in image interpretation, indicating that Dall-E struggles to accurately identify the intended design elements and the principles of the images it generates. Moreover, the results revealed Dall-E 3 tends to fabricate information when generating new elements.

Implications: Dall-E 3 can help a student generate a display image and interpret it from different design concepts.

Keywords: Dall-E 3, ChatGPT, Image-generative AI, visual merchandising

How to Cite:

Ho, X. & Ma, Y., (2025) “Accuracy of Image-generative AI in Interpreting Images Generated from Prompt-Based Inputs”, International Textile and Apparel Association Annual Conference Proceedings 82(1). doi: https://doi.org/10.31274/itaa.21936