Wednesday, July 5, 2023, 12pm

Complex sensemaking, whether for scientists, designers, or lawyers, involves discovering diverse domains and iteratively curating their core structures to gain new insights. Typical bottom-up processes involving collecting, synthesizing, and integrating examples are cognitively demanding and require frequent context-switching between different tools and stages. Bootstrapping explorations with existing synthesis (e.g., extracting research threads described in scientific papers, re-mixing moodboards on a theme, re-using core argument structures from legal cases) or generative artificial intelligence (GenAI) can reduce some costs but may introduce new interaction and cognitive challenges. Furthermore, users are often left alone when leveraging their curated data for creative insights.

My research aims to address this gap by developing systems and interaction techniques that facilitate users’ bootstrapping from existing synthesis and GenAI, while enhancing downstream discovery using user-curated data. I focus on two significant domains: scientific research (completed work) and industrial design (proposed work).

In the first part of the proposal, I explore how users can curate a specific form of ‘research threads’ (i.e., sentences describing and citing specific groups of prior research) while reading an individual paper, and ‘expert committees’ (i.e., a group of authors representing a particular perspective on a topical area) to make sense of the knowledge landscape and discover diverse and relevant results. I propose two approaches that each centers threads and committees as first-class objects in the interaction design. These approaches leverage citation graphs, AI-based retrieval, and GenAI-based summarization to enable broader exploration beyond relevant research examined in the source papers. The objects also act as boundary objects, translating users’ intents to AI during exploration and augmenting the discovery process. They help users to find other significant papers, identify high-level themes that emerge from them, and generate relevance explanations for retrieved results that enhance users’ engagement.

In the second part of the proposal, I introduce a simple schema that abstracts users’ synthesis during scientific research ideation. Using this schema, I develop an analogical search engine for scientists to retrieve analogical papers addressing similar high-level challenges as the user query despite differences in low-level specifics, and demonstrate that they trigger creative adaptation ideas. In the proposed work, I plan to extend this paradigm to the industrial design domain, which involves designers’ continuous explorations to gain design inspirations and emphasizes processing of information from the visual modality during exploration.

Thesis Committee:
Aniket Kittur (Chair)
Sherry Tongshuang Wu
Nikolas Martelaro
Michael Terry (Google)

Additional Information

Zoom Participation. See announcement.

Event Type: Thesis Proposals
Room Number: Virtual Presentation - ET
Building: Remote Access - Zoom
Speaker's Name: HYEONSU BUTTWEILER KANG
Speaker Website: www.hyeonsukang.com
Speaker's Professional Title: Ph.D. Student, Human-Computer Interaction Institute, Carnegie Mellon University
Talk Title: Supporting Synthesis with AI and Interaction Design to Enhance Sensemaking and Creativity
For More Information: hyeonsuk@cs.cmu.edu
Affiliations: Human-Computer Interaction Institute (HCII)
Organization(s): SCS
Event Website Title: Event Website
Event Website URL: www.hcii.cmu.edu…