Accelerating data discovery through user-centered design

Duration

4 Months
January - May 2022

Team Members

2 CGHE Leaders
Lead UX Designer (me) 
Lead UX Researcher
2 Developers
Project Manager

Skills Learned

Literature Review
User Interviews
Design Space Critique
Survey
Interview
Affinity Mapping
Brainstorming strategies
Prototyping
Heuristic & Think Aloud Evaluation

Overview

Encouraging users to have a deep understanding of their data through automation, hierarchial exploration and rich data visualization

80% of a data scientists’ time is spent on low-level activities such as manually tweaking the data and understanding the problem space. Despite recent advances in automation through autoML tools like Google’s VertexAI data scientists today still rely on traditional and often time-consuming methods to explore their data. Existing tools are disconnected from their needs and workflow resulting in skepticism among practioners around modern tools that involve automation.

Bridging the gap

“...auto insight tools informed by user studies in specific domains is scarce. Without significant understanding of users the new applications identified may be divorced from real world needs.”

Stasko & Endert, Characterizing Automated Data Insights, 2020

Primarily focusing on redesigning the data exploration module of an existing autoML tool, the goal of this project is to leverage user-centered design to augment human capability in a controllable and trustable manner. Through 10 semi-structured user interviews and literature, following are the research areas I am exploring through my design:

As time has gone by, research has pointed out that there remains skepticism in the community when it comes to using modern tools that are focused around automating parts of the data discovery process. My initial research has surprisingly shown that practitioners would prefer using traditional tools over newer tools, even if it involves more manual effort and time.

As Dr. Stasko sums it up perfectly:

“...auto insight tools informed by user studies in specific domains is scarce. Without significant understanding of users the new applications identified may be divorced from real world needs.”

At Bell Labs, I'm working on understanding how practitioners explore datasets and ideating a user interface that bridges this gap between intelligent algorithms and user needs. The aim of this ongoing project is to envision, design and develop a data discovery module that helps accelerate exploratory data analysis using advanced algorithms, data visualizations and rich interactions. The UI I'm working on will function as an extensible framework, that allows for continuous addition of future algorithms. This would involve a careful design and development effort as I keep in mind that the UI I design is technically implementable and integrates well with the existing backend algorithms at Bell Labs.

I will be continuing this internship project as a part-time employee and as part of my masters thesis. We are also working towards a paper, and I intend to add more details here as and when our project gets cleared for publishing key information online.

Duration

4 Months
August - December 2021

Team Members

We worked in a team of 4:
1. Sejal Sarkar
2. Kshitij Gupta
3. Linda Lee
4. David Lacy

Role

UX Designer (Lead)
UX Researcher
(Phases 1 to 2)

Skills Learned

Literature Review
User Interviews
Design Space Critique
Survey
Interview
Affinity Mapping
Brainstorming strategies
Prototyping
Heuristic & Think Aloud Evaluation

Duration

4 Months
August - December 2021

Team Members

We worked in a team of 4:
1. Sejal Sarkar
2. Kshitij Gupta
3. Linda Lee
4. David Lacy

Role

UX Designer (Lead)
UX Researcher
(Phases 1 to 2)

Skills Learned

Literature Review
User Interviews
Design Space Critique
Survey
Interview
Affinity Mapping
Brainstorming strategies
Prototyping
Heuristic & Think Aloud Evaluation

Duration

1 year
June '23-'24

Team Members

1. Kshitij Gupta (me)
2. Senior data scientist (mentor)
3. Backend Developer

Role

UX Designer
UX Researcher
Frontend Developer

Skills Learned

Literature Review
User Interviews
Data Viz
Exploratory Data Analysis
Boosting user trust in AI 
Designing for humans-in-the-loop