PhD Defense: Towards Rapid, Effective, and Expressive Visual Data Exploration

Talk
Mehmet Adil Yalcin
Time: 
10.19.2016 15:00 to 17:00
Location: 

AVW 3450

The need to make precise data-driven decisions is ever growing with the increasing availability and impact of data in our everyday lives. Quickly gaining an accurate and deep understanding of a given dataset, and exploring its rich relations and potential insights, requires cognitive processes that are heavily influenced by the design of our tools. Seeing and interacting with data remains the human-centered approach to identify relevant data features, trends, and outliers, i.e. the promise and potential of interactive data visualizations.
In this thesis, I focus on the design and cognitive aspects of visual and interactive data exploration. First, I present the Cognitive Exploration Framework which identifies six distinct cognitive stages. This framework provides a high-level structure to existing visualization and interaction design guidelines as well as the evaluation of analytic tools through a perspective on cognitive barriers and activities. Next, in order to reduce the complexities in making decisions to create interactive data visualizations, I present a minimalist yet expressive data exploration model for tabular datasets based on aggregated data summaries and linked selections. I demonstrate that this model applies to categorical, numerical, temporal and spatial data, as well as set-typed data. Based on this model, I developed Keshif as a new out-of-the-box, web-based tool to bootstrap the critical data exploration process for tabular datasets. I have so far applied Keshif to over 175 datasets across many domains, aiming to serve journalists, researchers, policy makers, businesses, collectors, and those tracking personal data.

Using tools with new designs and capabilities require learning and seeking for help, whether you are a novice or an expert. This is especially critical for tools supporting the complex cognitive processes in data exploration and analytical sensemaking. To improve self-service training for visual data interfaces, I present a data-driven contextual in-situ help system, HelpIn, which contrasts with separated and static videos, tutorials, and manuals. The last, but not least, focus of this thesis is graphical perception of data visualizations, another critical cognitive stage, in addition to proposing new visualization designs. To this end, I present a graphical perception and design evaluation contrasting the use of treemaps for non-hierarchical numerical data against two multi-column bar chart designs, wrapped bars and piled bars, a novel design introduced in this thesis.

Overall, this thesis contributes to our understanding on how to create effective visual and interactive data interfaces by focusing on human-facing challenges including design, cognition, perception, and the iterative, highly-dynamic nature of data exploration across many disciplines. The many facets of this work aim to support future work to lower barriers and provide the power of data analysis to a broad public, which remains a cognitive, design-driven, social, and educational endeavor.

Examining committee:

Chair: Dr. Ben Bederson

Co-Chair: Dr. Niklas Elmqvist

Dean’s rep: Dr. Ira Chinoy

Members: Dr. Amitabh Varshney

Dr. Catherine Plaisant