The Perception Gap
The Perception Gap
Who still reads full-length scientific articles?
Data storytelling has gone digital and is ubiquitous, with ever more emphasis on graphics, visuals, and animation. Yet, few data scientists, writers or decision makers are proficient in the workings of the human visual system, opening a perception gap between the data and its representation.
Approximately 8% of men and 0.5% of women are colour-blind. In a general audience, it means about 1 person in 20 may have trouble understanding graphs and illustrations, or infer wrong information if visualisations are not specifically designed with colour-blindness in mind. This is larger than the margin of most elections.
From a perception perspective, however, colour-blindness is only the tip of the iceberg. “Perceptual biases” that originate directly from the Human Visual System include:
- Placing points close to each other (e.g., in a network graph) implies proximity and will lead observer to infer “closeness” even if the data does not support this.
- Areas are not well-interpreted by the human visual system. A small difference in area is often perceived as much larger, this is why pie charts are a bad idea.
- Movement has priority; movement in the periphery of the visual field trigger “fight or flight” responses directly, possibly distracting the viewer from the rest of the content.
- Humans are more sensitive to contrast in luminance (brightness), less sensitive to colour contrast along green-red axis, and even less sensitive in the blue-yellow direction. “Equal” colour differences are not perceived as such.
- Colours have symbolic meanings, which can change depending on culture, e.g., green = good, red=danger. Reversing or ignoring cultural meaning may otherwise distract from the visualization intent.
Current focus on digital storytelling as well as data visualisation of complex algorithms and data sets has increased the relevance of understanding human perception to ensure that visualisations are perceived in accordance with the data they represent. Humans are so good at pattern recognition that we are more than willing to interpret random “perceptual” noise as “meaningful patterns” when prompted by cues to our visual system.
As data and its analysis are becoming more commoditised, data visualisation has become a cornerstone of communication and decision making. However, the human visual system remains the same as it has been for centuries and needs to be taken into account to avoid creating a perception gap.
Dr Clement Fredembach is an expert in human vision, perception, colour science with a phd, postdoc, and researcher in perceptual image quality for Canon. Clement is presenting at The Art of Analytics - The Fusion of Data, Science and Art: Presentation and Exhibition on Tuesday 30 May as part of Vivid Ideas.