top of page

INTEGRATING

Low-Rank Adaptation IN BRAINSTORMING SESSIONS

RESEARCH

Course

Exploring Design Intelligence

Date

Jul 2024

Extra Material

This study examines the integration of AI into the design process, with a specific focus on enhancing brainstorming sessions through personalised AI image-generation tools. The research evaluates the effectiveness of employing a Low-Rank Adaptation (LoRA) model within a stable diffusion framework to align the design team's vision and improve collaboration. The experiment involved three students participating in two brainstorming sessions: one utilising generic AI tools and the other using the personalised LoRA model.

 

The results show that the personalised LoRA led to improved collaboration, a stronger emphasis on design details, and more frequent use of metaphors, all of which suggest enhanced creativity. Although some participants felt limited by the LoRA, they ultimately became more productive. The images generated by LoRA were rated slightly higher than those from generic AI tools.

bottom of page