Skip to main content Skip to navigation

Using customized GPT to develop prompting proficiency in architectural AI-generated images

Project Overview

The document examines the application of customized generative AI models, specifically GPT, in architectural education to improve students' skills in prompt engineering for generating AI-driven images. It emphasizes the growing significance of prompt engineering as a crucial skill due to the increasing prevalence of generative AI tools in the field. The research highlights the effectiveness of structured prompting guides and the use of AI personas in fostering critical thinking and enhancing communication skills among architecture students regarding architectural concepts. Findings indicate that personalized interactions with AI not only bolster students' confidence but also lead to better articulation of design ideas, ultimately resulting in improved outcomes in prompt generation. This study underscores the transformative potential of generative AI in education, particularly in developing essential competencies that prepare students for the evolving demands of their professions.

Key Applications

Customized GPT models for prompting proficiency in AI-generated images

Context: Architecture education, targeting architecture students at various educational levels (bachelor's, master's, doctoral)

Implementation: Mixed-methods experimental design with control and experimental groups utilizing structured guides and AI personas

Outcomes: Significant improvements in word count, prompt similarity, and student confidence; enhanced critical thinking skills

Challenges: Steep learning curve for students; potential for confusion with AI feedback; need for clearer definitions of prompting parameters

Implementation Barriers

Technical

Participants faced difficulties with AI feedback, which was sometimes perceived as ambiguous or misaligned with their design intentions.

Proposed Solutions: Improve explainability of AI feedback by attaching rationale and allowing two-way interaction for students to query the feedback.

Cultural

Participants expressed a desire for more culturally diverse image outputs and clearer definitions of architectural terminology.

Proposed Solutions: Incorporate culturally inclusive training data and multilingual support to broaden accessibility.

Educational

Students experienced vocabulary limitations, word count constraints, and challenges related to insufficient background knowledge for articulating complex architectural concepts.

Proposed Solutions: Provide databases of sample keywords and real-time word count indicators to enhance usability.

Project Team

Juan David Salazar Rodriguez

Researcher

Sam Conrad Joyce

Researcher

Julfendi

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Juan David Salazar Rodriguez, Sam Conrad Joyce, Julfendi

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies