More about HKUST
Bridging Data Analysis and Storytelling with Human-AI Collaborative Tools
The Hong Kong University of Science and Technology Department of Computer Science and Engineering PhD Thesis Defence Title: "Bridging Data Analysis and Storytelling with Human-AI Collaborative Tools" By Mr. Haotian LI Abstract: Working with data has become common across various disciplines, from natural science to business. For these data workers, communicating data insights and knowledge from data analysis through data stories plays a crucial role in enhancing collaboration in teams and raising public awareness. However, creating clear, coherent, and engaging data stories requires diverse skills and considerable time for human authors. To address this challenge, this thesis investigates how to introduce artificial intelligence (AI) to effectively reduce human effort and streamline data analysis and communication. In the first part of this thesis, we built theoretical foundations for human-AI collaboration in bridging data analysis and storytelling. We conducted an interview study to gain insights into the expected AI roles and challenges when telling data stories. Based on the findings from the interview, the human-AI collaboration in data storytelling tools is formalized as a framework with two dimensions: the roles of collaborators and the stages of collaboration. With the framework, various insights and opportunities in designing human-AI collaborative tools are unveiled through a comprehensive literature review. The second part of this thesis presents research on instantiating the theories into interactive tools for real-world applications. First, we designed Notable to bridge data analysis and data storytelling in computational notebooks with on-the-fly assistance, including automatic data fact documentation, story organization, and slide creation. Our second work explored enhancing the alignment between humans and AI in story organization with meta relations, which delineate connections between data story pieces using meta information beyond datasets, such as domain knowledge and narrative intent. With the two parts of this thesis, we hope to contribute knowledge and experience as cornerstones to boost effective and seamless collaboration between humans and AI for data analysis and communication in the coming era of large-scale AI systems. Date: Monday, 22 July 2024 Time: 2:00pm - 4:00pm Venue: Room 5501 Lifts 25/26 Chairman: Dr. Weiyin HONG (ISOM) Committee Members: Prof. Huamin QU (Supervisor) Prof. Cunsheng DING Prof. Xiaofang ZHOU Prof. Fugee TSUNG (IEDA) Dr. Duen Horng CHAU (Georgia Tech)