Jurassic Mark: Inattentional Blindness for a Datasaurus Reveals that Visualizations are Explored, not Seen

TL;DR

This study demonstrates that visualizations are explored over time rather than instantly perceived, showing inattentional blindness to prominent data features during quick glances, emphasizing the active nature of data perception.

Contribution

The paper introduces an experimental paradigm revealing that visual data features are not immediately perceived but require active exploration, highlighting the iterative filtering process in visual data interpretation.

Findings

Participants missed a conspicuous dinosaur in scatterplots when instructed to ignore it.

Inattentional blindness was significantly reduced when the task did not incentivize ignoring data.

Visualizations are perceived through active, iterative exploration rather than instant recognition.

Abstract

Graphs effectively communicate data because they capitalize on the visual system's ability to rapidly extract patterns. Yet, this pattern extraction does not occur in a single glance. Instead, research on visual attention suggests that the visual system iteratively applies a sequence of filtering operations on an image, extracting patterns from subsets of visual information over time, while selectively inhibiting other information at each of these moments. To demonstrate that this powerful series of filtering operations also occurs during the perception of visualized data, we designed a task where participants made judgments from one class of marks on a scatterplot, presumably incentivizing them to relatively ignore other classes of marks. Participants consistently missed a conspicuous dinosaur in the ignored collection of marks (93% for a 1s presentation, and 61% for 2.5s), but not in a control condition where the incentive to ignore that collection was removed (25% for a 1s presentation, and 11% for 2.5s), revealing that data visualizations are not "seen" in a single glance, and instead require an active process of exploration.