Effective Visual Representations (Test Talk)

Data visualization is a pipeline that transforms the data into visual encodings, which convey insights to the humans. To make such a transformation effective, it has to take into account data specifics, task requirements, and not at the least human perceptual and cognitive capabilities. The last mentioned ones play an important role in the visual mapping stage of the pipeline, where the visual representations are assigned to data. This step determines the expressiveness and effectiveness of the visualization, i.e., it determines whether the visual representation conveys all relevant aspects and whether the underlying data characteristics are clearly communicated to the viewer. One can think of three strategies to achieve effective visual encodings. One approach can be based on analyzing the work of visual artists, most notably illustrators and infographics designers, and subsequently formalizing and extending their work into the realm of interactive data visualization. The illustrators have refined their visual designs over centuries and while they often do not include explicit knowledge about human visual perception, the artists’ intuition in design leads to visually effective encodings. The second approach to achieve effective visual mappings is based on present knowledge from the vision science and psychology that describe how the human visual processing works. The outcome of these sciences then serves as a guideline for the design of visual mappings. While century-old Gestalt principles are an important factor in the visual mapping design, the most recent outcome from vision sciences is not directly coupled with the visualization design. The visualization research progress is thus dependent on progress of vision and psychology sciences. There is, however, a third approach for developing visually effective visual mappings and it is directly based on studies of human perception and cognition. The third method of obtaining visual mapping designs is determined only by a statistical description of the human visual performance and not by explicit visual processing knowledge as it was in the second approach. The statistics of human visual performance can be obtained through psychophysics studies assessing the human understanding of particular visualization mappings. In contrast to the above two methods, this approach results into the quantitative characterization of the effectiveness of a visual mapping. It does neither strongly depend on prior work of creative crafts nor perceptual sciences. Statistics can be moreover used for revising the visual mapping over multiple iterations so that optimization strategies could be employed to find a potentially best solution that maximizes the match between the stimulus and the sensation. Employing concepts such as feedback loop from control theory, after several iterations of psychophysics studies and revisions, an entirely novel visualization mapping is developed, with known, numerically characterized effectiveness. Thus such visual mapping strategy can be the first step in predicting the understanding of specific visualizations viewed by lay or expert audiences.