Lumos: Increasing Awareness of Analytic Behavior during Visual Data Analysis

Area of Science:

• Human-Computer Interaction
• Data Visualization
• Information Visualization

Background:

• Visual data analysis tools offer flexibility but can lead to unintended biases in data exploration.
• Users may overemphasize or underemphasize data subsets or attributes, impacting analysis outcomes.
• Lack of awareness regarding these analytic behaviors hinders effective data interpretation.

Purpose of the Study:

• To introduce Lumos, a novel visual data analysis tool designed to increase user awareness of their analytic behaviors.
• To investigate how Lumos's in-situ and ex-situ visualization techniques provide real-time feedback on data exploration.
• To evaluate the impact of Lumos on users' data exploration, decision-making, and awareness of their analytical practices.

Main Methods:

• Development of Lumos, a visual data analysis tool incorporating interaction history visualization.
• Implementation of in-situ (highlighting previously examined data points) and ex-situ (overlaying on baseline distribution) feedback techniques.
• Conducting a user study with 24 participants to assess Lumos's effectiveness in real-world data exploration scenarios.

Main Results:

• Lumos effectively increases users' real-time awareness of their visual data analysis practices.
• The tool promotes user reflection on their analytical intentions and behaviors during data exploration.
• Participants' subsequent interactions were potentially influenced by the real-time feedback provided by Lumos.

Conclusions:

• Visualizing interaction history in data analysis tools like Lumos enhances user metacognition.
• Real-time feedback mechanisms are crucial for mitigating biases in visual data exploration.
• Lumos demonstrates a promising approach to improving the transparency and intentionality of data analysis workflows.