Recommender systems as co-design tools: empowering human-AI hybrid intelligence in complex design networks
Danwen Ji, Man Zhang, Yongqi Lou · Journal of Engineering Design · 2025
Collaborative design (Co-design) is essential for addressing complex, multidisciplinary challenges, but traditional tools often struggle to support dynamic, multi-stakeholder environments. This paper presents a graph neural network (GNN)-based recommender system that enhances co-design by predicting potential connections within heterogeneous collaboration networks. The system integrates multi-dimensional network analysis with deep learning on small-scale data to provide actionable recommendations, assisting designers in identifying key collaborators and projects. Developed within the NICE 2035 community network, the system organises diverse entities into structured network graphs and employs the GraphSAGE architecture to predict and prioritise network linkages. By translating complex network insights into intuitive, goal-oriented suggestions, the system empowers practitioners to streamline workflows, enhance collaboration, and explore innovative strategies. Evaluation through workshops and interviews with 10 design practitioners demonstrated its ability to simplify decision-making, foster creativity, and support multidisciplinary engagement. While participants praised its efficiency and accessibility, challenges were noted in transparency, data dependency, and the need for deeper contextual insights. These findings highlight the potential of recommender systems to bridge data-driven and human-centered approaches, enabling Hybrid Intelligence in design. Future work will focus on scalability, user customisation, and transparent explanations to enhance applicability across diverse, complex design domains.