Session: DTM-02-01: Artificial Intelligence in Design

Paper Number: 118069

118069 - Exploring Human-Centered Design Method Selection Strategies With Large Language Models 

Human-centered design (HCD) is supported by a range of open-source materials, many of which promise designers a greater ability to select the optimal method for a desired design outcome. Despite the widespread availability of such information, a clearer understanding about how designers translate information on design methods into method selections is needed to better unify models of design cognition with theories on design methods and support human-AI teaming in design. In this work, we compare HCD methods advised by the publicly-available neural network “large language model” GPT-3.5 to 405 novice design team method selections occurring in five separate years of a design project-based learning course at a large public university.  We observe that GPT3.5 appears to represent design method knowledge held in method repositories like TheDesignExchange well. We also observe that while there are strong similarities in human and GPT3.5 method selections, GPT3.5 appears to struggle to distinguish between HCD phases in terms of method selection recommendations, and also recommends methods that focus on specific aspects of HCD phases. These findings highlight the unique contribution of human design cognition in design decision-making, and also emphasize the promise of human-AI teaming in design method selection with further research on phase and method context. 

Presenting Author: Vivek Rao Berkeley

Presenting Author Biography: Vivek Rao has been a lecturer at UC-Berkeley's Haas School of Business since 2018, teaching courses on design, innovation, entrepreneurship and foresight methodologies across the three MBA programs. In addition to his work at Haas, he leads research on design theory and methodology at UC-Berkeley's Department of Mechanical Engineering and serves on the admissions and educational committees of the Masters of Design program at UC-Berkeley’s Jacobs Institute for Design Innovation, where he co-developed and has co-taught the degree’s foundational course, ‘Technology Design Foundations,” for three semesters. His research has received support from the National Science Foundation, Center for Long-term Cybersecurity, and the Odebrecht Foundation, and has earned multiple awards, including the 2020 ASME IDETC Best Paper Award in the Design Theory and Methodology track. He earned a BS, MS and PhD, all in Mechanical Engineering, from UC-Berkeley, and prior to returning to academia, he worked as an engineering designer and strategist at global innovation consultancy IDEO.

Authors:

Vivek Rao Berkeley
Timothy Yang UC-Berkeley
Euiyoung Kim Delft University of Technology
Alice Agogino UC Berkeley
Kosa Goucher-Lambert UC-Berkeley