Collaborative Systems for Design Inspiration

Inspiration is a crucial activity in design and innovation, in which potential desirable solutions are explored and refined to later provide directions and inspiration in later stages of design. Designers use a plethora of inspirational methods and tools. Among them are mood boards, a visual collage of pictures, text, and objects, that is usually created collaboratively in e.g. fashion design, architecture, and marketing. Mood boards help designers identify, select, and curate visually inspiring content, to express their existing ideas but also to inspire new ideas through their combination. As mood board design becomes increasingly digital, the availability and variety of online material presents ever greater opportunities to assist designers. However, it also poses new challenges. Through interviews with professional designers we identified three of them in particular: 1) turning tacit ideas into expressible search terms, 2) synthesizing and reflecting on visual material, and 3) finding external inspiration. While existing methods for visual inspiration hint at a great potential to support conceptual innovations, the computational support to address these challenges remains lacking. My work contributes knowledge, interaction techniques, and co-creative algorithms to assist users with these challenges. First, it introduces collaborative systems that enrich images with semantic information, to help designers express vague, visual ideas and translate them into usable search terms. Second, to support visual reflection, this thesis introduces multiple levels of semantic abstraction of visual material, to inspire designers to find higher-level concepts in their own work. Third, collaboration is an integral part of physical mood board practice, yet digital mood boards are often crafted at an individual level, which deprives designers of many opportunities to challenge and expand their ideas. An artificial agent was developed that creates mood boards jointly with a designer, based on a cooperative contextual bandit algorithm. This approach conferred it the ability to make its own decisions about whether to explore or exploit the visual contents of the current mood board, and our participants, all professional designers, genuinely valued its contributions. Thanks to a grounding-based interaction approach, it had the ability to justify its contributions and to inquire about sudden changes in the designer's choices. That resulted in a system that was perceived as a contributing agent, rather than merely a tool. Finally, beyond mood board creation with individual designers, the developed collaborative systems also contributed to creative collaborations between human designers. Within these collaborations, artificial agents played a role complementary to that of designers, and were appreciated in particular when ideas were sparse, when designers felt ''stuck',' or had trouble expressing their ideas. My work highlights the immense potential of intelligent collaborative systems for inspiration-seeking and creative processes, and opens new ways to assist designers in the era of digital ideation.