Carnegie Mellon University
Browse

Accelerating Innovation through AI-Powered Conceptual Abstraction and Interaction Design

Download (18.09 MB)
thesis
posted on 2024-09-23, 21:18 authored by Hyeonsu KangHyeonsu Kang

 Isaac Newton famously said, “stand on the shoulders of giants,” to emphasize the  importance of pre-existing synthesis for making new challenges tractable in a single  human brain. Newton himself learned partial abstractions from Ptolemy, Copernicus,  Kepler, and Galileo, as well as Descartes’ analytic paradigm, which he used as foun dations for his calculus problem. However, rapidly accumulating knowledge makes it  increasingly difficult to be aware of existing approaches and innovate upon them.  

In this thesis, I argue that what we need are new tools to help people synthesize use ful cross-cutting abstractions from knowledge, effectively organize knowledge with  those abstractions, and use them to find novel cross-domain insights. I present four  systems toward this goal, where I explore several kinds of abstractions to enable new  interaction capabilities. These include ‘research threads’ for supercharging people’s  reading experiences with AI to enable seamless interaction with thread-level abstrac tions while reading, the purpose-mechanism schema and how AI can help users find  cross-domain analogies, and ‘active ingredients,’ a mechanism abstraction that helps  designers engage with and transfer insights from biology to mobility design.  

Through controlled laboratory studies, I demonstrate the value of these abstractions  in elevating people’s focus during reading and exploration to a higher level (e.g.,  from individual papers to how notable threads divide a research field; from individual  species to active ingredients of mechanisms), thereby gaining efficiency and helping  them broaden their pursuit of problem-solving strategies. The end result is more cre ative ideas. 

 In a world of abundant knowledge and large language models, the structuring and  distilling of conceptual insights will be the defining characteristics of driving value in  knowledge work. By putting powerful techniques that empower conceptual interac tion with information into the hands of everyday people, I envision a future where  innovators everywhere deeply engage with insights that overcome domain bound aries and develop novel ideas that address personal challenges they face to bring  forth positive effects for the world.  

History

Date

2024-08-01

Degree Type

  • Dissertation

Department

  • Human-Computer Interaction Institute

Degree Name

  • Doctor of Philosophy (PhD)

Advisor(s)

Aniket Kittur