Shaping conversations: Custom GPTs to spark reflection in design

This paper documents the development of a customized GPT designed to facilitate reflection on cognitive biases in design decision-making. Leveraging the OpenAI GPTs platform, a specialized conversational agent was developed, integrating a structured knowledge base of twelve categories of cognitive biases with prompt engineering strategies for progressive disclosure. The study details the construction of the GPT, including the design of system instructions, the structuring of the knowledge base, and the implementation of conversation starters. By analyzing real conversations with design students, the iterative refinement process and the challenges of implementing effective conversational principles are documented. Although the system did not fully achieve the intended goals of progressive disclosure, highlighting the inherent limitations of the platform, the results provide practical guidance for the development of specialized GPTs and identify key considerations for contextual adaptation and cognitive load management. This work contributes to understanding how conversational AI platforms can be tailored to support reflection in specific contexts, particularly within the field of design.