Qinyi Zhou

CoLyricist is an AI-assisted lyric writing tool designed around the workflows of experienced lyricists. Interviews with 10 lyricists identified four recurring stages: theme setting, ideation, drafting, and melody fitting. The system offers tailored support for brainstorming, drafting, rhyming, and fitting lyrics to a melody while preserving the writer’s intended meaning. In an exploratory study with 16 experienced and novice lyricists, participants valued the workflow-aligned support; novices particularly appreciated melody fitting, while experienced lyricists valued the ideation tools.

Conversational language models can present misinformation in fluent responses, making overreliance difficult to detect from task outcomes alone. We studied interaction logs from 77 participants working across three tasks with an LLM containing plausible misinformation. By encoding and clustering segments of user interactions, we identified behavioral patterns associated with overreliance. Lower overreliance was linked to careful task comprehension and fine-grained navigation; higher overreliance was linked to frequent copy-paste, skipping initial comprehension, repeated LLM references, coarse locating, and accepting misinformation despite hesitation. These findings inform the design of timely interventions during human-AI interaction.

Cuby and His Shadow Friend is a game-based STEAM learning book that introduces light and shadow to children aged 6-9 through parent-child co-reading. Developed with feedback from teachers, researchers, and parents, the book uses tactile, visual, and auditory activities, including a UV lamp that reveals story details and a prism for exploring refraction. User feedback suggests that these multi-sensory features encourage curiosity and understanding, while identifying elements that could be improved for independent use.

Product metaphors, which involve creating products that convey meaning through metaphorical associations, are a powerful tool in product design. However, according to our formative study, novice designers often struggle to establish coherent links between target and source, to manage the complexity of diverse mapping possibilities, and to balance product usability with metaphorical expression. To address these challenges, we introduce ProductMeta, a creativity support tool designed to support novice designers in exploring and developing metaphorical product designs. ProductMeta incorporates domain knowledge and decomposes the design process into iterative modules and framework-based interfaces, fostering both divergent and convergent thinking. Through user studies, we demonstrate that ProductMeta enables novice designers to generate diverse and contextually relevant design ideas by facilitating structured exploration. We conclude with design implications for human-AI co-creation.

Nowadays, novice users often turn to digital tutorials for guidance in software. However, searching and utilizing the tutorial remains a challenge due to the request for proper problem articulation, extensive searches and mind-intensive follow-through. We introduce "Docent", a system designed to bridge this knowledge-seeking gap. Powered by Large Language Models (LLMs), Docent takes vague user input and recent digital operation contexts to reason, seek, and present the most relevant tutorials in-situ. We assume that Docent smooths the user experience and facilitates learning of the software.