Optimizing Input in Generative AI Powered Public Redesign
Authors
Mohamed Farook, Farika Magjabeen ; Jackson, Cody ; Shalah, M A
Term
4. term
Education
Publication year
2026
Submitted on
2026-06-04
Pages
18
Abstract
Generative artificial intelligence (GenAI) can help citizens imagine and share ideas for public urban spaces. While AI-generated images are already used in planning, less is known about how different input methods affect people’s ability to express design intentions and work with AI-assisted tools. This thesis examines four input methods—text, voice, inpainting (painting over parts of an image to change them), and drag-and-drop—within a web-based platform for participatory urban redesign. A mixed-methods user study with 29 participants assessed how well each method supported expressive ability, user satisfaction, and ease of use. The results show that text and voice best supported communicating design intentions, while drag-and-drop was experienced as the most intuitive and usable interaction. Qualitative findings further suggest that input methods shaped not only how participants expressed ideas but also the kinds of redesign concepts they generated. Overall, the study highlights the importance of interaction design in AI-supported participatory planning and offers guidance for developing multimodal tools that broaden public engagement in urban redesign processes.
Generativ kunstig intelligens (GenAI) kan hjælpe borgere med at forestille sig og dele idéer til offentlige byrum. Meget forskning har set på AI-genererede billeder i planlægning, men mindre på hvordan forskellige måder at give input på påvirker folks evne til at udtrykke designintentioner og bruge AI-assisterede værktøjer. Denne afhandling undersøger fire inputmetoder – tekst, tale, inpainting (at male over dele af et billede for at ændre dem) og træk-og-slip – i en webbaseret platform til borgerinddragende byomdannelse. Et mixed-methods brugerstudie med 29 deltagere vurderede, hvor godt hver metode støttede udtrykskraft, brugertilfredshed og brugervenlighed. Resultaterne viser, at tekst og tale gav den bedste støtte til at kommunikere designintentioner, mens træk-og-slip blev oplevet som den mest intuitive og anvendelige interaktion. Kvalitative analyser antyder desuden, at inputmetoderne påvirkede både måden, deltagerne formulerede idéer på, og hvilke typer redesignkoncepter de skabte. Samlet understreger fundene, at interaktionsdesign er afgørende i AI-understøttet borgerinddragende planlægning, og de giver vejledning til at udvikle multimodale værktøjer, der kan engagere flere borgere i byomdannelsesprocesser.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords