*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Automated Yet Transparent Data Insights
Terrance Law
Ph.D. Student in Computer Science
School of Interactive Computing
Georgia Institute of Technology
Date: Monday, May 4th, 2020
Time: 3:00 pm to 5:00 pm (EDT)
BlueJeans: https://primetime.bluejeans.com/a2m/live-event/zegywwpc
Committee:
Dr. Alex Endert (advisor), School of Interactive Computing, Georgia Institute of Technology
Dr. John Stasko (advisor), School of Interactive Computing, Georgia Institute of Technology
Dr. Duen Horng (Polo) Chau, School of Computational Science and Engineering, Georgia Institute of Technology
Dr. Jian Zhao, Cheriton School of Computer Science, University of Waterloo
Abstract:
Visualization systems have begun to employ functionality that automatically extracts and communicates data insights. For example, given a sales data set, these systems may automatically generate textual descriptions of data insights such as "Your market share of motorcycles is the highest in Georgia" and "The sales of motorcycles in Georgia have been decreasing over the past six months." Researchers have been investigating various applications of automated data insights such as exploratory data analysis, visualization interpretation, and question answering. However, such investigations have been detached from the workflow of visualization users in practice.
A practical concern of these automatically generated data insights is a lack of transparency: Systems that recommend automated data insights often do not reveal how and why the insights are generated. In a typical scenario, visualization users manually mine insights from data and rigorously assess the validity of these data insights. With the validation, they can confidently communicate the insights to other stakeholders and utilize the insights to inform decisions. In the face of data insights that are automatically generated, however, visualization users tend to be skeptical because they do not know how the insights are garnered. Users may not be willing to trust data insights that they cannot explain for communication and for making important decisions. Opaqueness in automated insights can pose hindrance to their adoption.
My work intends to investigate the needs for transparency in automated data insights and the effectiveness of explanations in supporting transparency. First, I proposed an organizational framework of the types and purposes of automated insights based on a systematic review of 20 systems that recommend automated insights. I then interviewed 23 professional visualization users from 19 organizations to understand how they perceive automated insight tools. I will conduct a crowdsourced study to investigate the effectiveness of why and why-not explanations in promoting user trust, understanding, and knowledge transfer. My thesis research is expected to offer a more structured understanding of automated data insights and provide guidance to designers who aim to create transparent and trustworthy automated insight tools.
