Human-AI Co-Creation in Software Development: A Database-Centric Approach for Empowering Non-IT Students

With the rapid advancement of AI-native code
editors like Cursor, software development has become more
accessible to non-specialists. However, for liberal arts
students lacking deep technical backgrounds, reliance on
fragmented natural language prompts often leads to “garbage
code” and structural inconsistencies. This paper proposes a
Database-Centric Co-creation Approach as a collaborative
framework between humans and AI, positioning the
relational schema as a critical logical interface. In our
methodology, students first co-create a detailed database
schema through dialogue with generative AI, such as
ChatGPT, before proceeding to implementation. By injecting
this schema—a structured “Single Source of Truth”—into the
AI editor’s context, the schema serves as a powerful logical
constraint that mitigates AI divergence and ensures
architectural integrity. The effectiveness of this human
machine collaboration was demonstrated in a workshop
where a team of liberal arts students, with no prior IT
expertise, successfully developed a professional-standard
platform (“VetNearMe”) within only five days. Our findings
show that the database-centric approach shifts the human role
from “syntax-level coding” to “architectural decision
making,” effectively augmenting the creative capabilities of
non-IT students. This study suggests a new paradigm for
information education, emphasizing data modeling as the core
language for human-AI co-creation.