Not Just Novelty: A Longitudinal Study on Utility and Customization of AI Workflows

Tao Long, Katy Gero, Lydia Chilton

 DIS 2024  Generative AI workflows solve real and complex problems by chaining AI outputs together with human interaction. However, it's uncertain how useful they are after novelty wears off and whether users take advantage of the potential for customization. Thus, a study with 12 users over three weeks found a 4.3-session familiarization phase. Post-familiarization, perceived system usefulness increases notably, driven by user customization of workflows. This suggests a future where AI systems can be tailored to individual needs, enabling design for appropriation.

     PDF            arXiv       

 Generative Disco: Text-to-Video for Music Visualization

Vivian Liu, Tao Long, Nathan Raw, Lydia Chilton

 arXiv 2024  Visuals enhance our music experience while existing text-to-video tools produce uncanny ones. Our system, Generative Disco, helps generate music visualizations with start and end prompts and design patterns (transition and holds) using LLMs and text-to-image models. With robust user control to mitigate the uncanny, professionals found the system enjoyable, easy to explore, and highly expressive.

     PDF            Video            Tweetorial           Code           Demo           BibTeX      

ReelFramer: Co-creating News Reels on Social Media with Generative AI

Sitong Wang, Samia Menon, Tao Long, Keren Henderson, Dingzeyu Li, Kevin Crowston, Mark Hansen, Jeffrey Nickerson, Lydia Chilton

 CHI 2024  News outlets struggle to create news reels that match the style of TikTok. Our system, ReelFramer, uses LLMs and text-to-image models to help journalists explore framings, generate scripts, character boards, and storyboards. Our user study and co-design workshop found it eases the burden of writing, makes the process rewarding, and generates informative and entertaining end results.

     PDF            Video            Tweetorial           BibTeX      

Tweetorial Hooks: Generative AI Tools to Motivate Science on Social Media

Tao Long, Dorothy Zhang, Grace Li, Batool Taraif, Samia Menon, Kynnedy Smith, Sitong Wang, Katy Gero, Lydia Chilton

 ICCC 2023  Tweetorials are popular, but STEM experts struggle to make them engaging. Our system, HookIncubator, utilizes scaffolding and AI chaining with LLMs for co-creating compelling hooks. Our evaluation indicates reduced cognitive load and well-supported science contextualization. Users edits LLM outputs for personal style and clarity, enhancing ownership and motivating more readers.

     PDF            Video            Project           Code           Demo           BibTeX      

Social Dynamics of AI Support in Creative Writing

Katy Gero, Tao Long, Lydia Chilton

 CHI 2023  To address the lack of work considering the social-technical gap that LLMs poses on writers, we discover three elements that govern a writer’s interaction with support actors: [1] what writers desire help with (motivation, planning, translating, reviewing) and [2] how writers perceive potential support actors (availability, individuality, trust) and [3] the values writers hold (intention, authenticity, creativity).

     PDF            Video            Tweetorial           Press           BibTeX      

Challenges and Opportunities for the Design of Smart Speakers

Tao Long, Lydia B. Chilton

 AAAI AI4Home 2023  VUIs have opened up many new types of interaction, while they are still underused. We identify opportunity spaces for designers and developers to explore such as focusing on 1) information support while multitasking (cooking, driving, childcare, etc), 2) incorporating users’ mental models, and 3) integrating calm design principles.

     PDF            Video            BibTeX      

Perspectives from Naive Participants and Social Scientists on Addressing Embodiment in a Virtual Cyberball Task

Tao Long, Swati Pandita, Andrea Won

 CSCW 2023 EA  To address the lack of research on UCD methodologies for social science paradigms, we identify shared considerations and disagreements between naive participants and experienced researchers for future systems: 1) intuitiveness, 2) inclusivity, 3) minimalism.

     PDF            Code           Funding 1           Funding 2           BibTeX      

Multistakeholder Perspectives on Digital Tools for U.S. Asylum Applicants Seeking Healthcare and Legal Information

Aparajita Bhandari, Diana Freed, Tara Pilato, Faten Taki, Gunisha Kaur, Stephen Yale-Loehr, Jane Powers, Tao Long, Natalie Bazarova

 CSCW 2022  A lack of information for asylum applicants negatively impacts their healthcare engagement. We identify four challenges for immigrants using current digital tools: lack of information, inaccessibility, emotional barriers, and lack of cultural sensitivity. The study recommends multiple stakeholders' perspectives when designing future tools.

     PDF            Video            Project           Press 1          Press 2          Press 3           BibTeX