Human-AI Collaboration in Spatial Planning: Designing an Interface for Visualizing Multi-Objective Trade-offs for Non-Expert Users

Abstract

In public-space planning, non-expert users can often spot that something is off. However, understanding the multi-objective trade-off logic behind a planning decision is a different story. This gap limits how deep public participation can actually go. Prior work on explainable AI (XAI) has mostly centered on explanation and user trust. Multi-objective decision support tools, meanwhile, tend to be built with expert users in mind. How non-experts might actually make sense of a complicated trade-off structure has gotten much less attention. To help close that gap, this study designs an explainable interface for non-expert users in public-space planning. It’s tested through a preliminary questionnaire and a between-subject user study comparing a conventional interface against the explainable one. The comparison looks at users’ understanding of trade-offs, their decision confidence, and how transparent the reasoning behind a recommendation feels to them. The explainable interface came out ahead on all three fronts. Users understood the pros and cons of different plans more clearly, along with the trade-offs among competing objectives and the reasoning behind recommendations. Their confidence increased too, without any drop in their sense of autonomy. Taken together, these results support explainable interfaces as a way to improve non-expert understanding of public-space plans. They also offer some empirical grounding for discussions of algorithmic transparency and public communication in planning.