Карта сайта
Версия для печати

Тренды визуализации данных подвергаются изменениям с развитием науки о восприятии

11 сентября 2013 Эффективная визуализация данных обусловлена соответствием таким составляющим как человеческое восприятие, память и подсознание. Автор предлагаемого материала детально исследует общие черты лучших примеров визуализации. (Материал опубликован на английском языке)
Steve Jobs popularized the saying "Simplicity is the ultimate sophistication." That dictum applies in spades to data visualization. The best visualizations leverage simple design principles to transmit useful intelligence to the human mind without delay and with maximum impact.

Data visualizations draw their power from the semi-conscious interplay among human perception, memory, and cognition. The best visualizations are designed to ensure that viewer perception of key relationships is automatic, retention and recall are effortless, and grasping of the larger meaning is immediate. Visualizations can also be a bond for communal consensus, helping people catalyze a shared understanding of quantitative relationships on topics of common interest.

Data visualizations stumble when their designers run afoul of the core principles of human perception. Advanced visualization is built on deep multidimensional correlations among diverse variables. Unfortunately, the technology can produce visuals of mind-numbing busyness. In many data analytic visualizations, we often see a wide range of linked graphs crammed into many dashboards, but often each element--report, scorecard, plot, etc.--includes far more detail than the average human can absorb. And every new visualization tweak--interactive, immersive, data-rich, 3-D, geospatial, multisensory, dynamic, context-sensitive, etc.--threatens to make the mess ever less comprehensible.

I came across a great article recently that uses the science of human perception in defining the core design principles for quantitative data visualizations. The core principle, per author Graham Odds, is this: "to maximise the power and efficacy of a visualisation, we should seek to encode as much information as possible in the pre-attentive features perceived during bottom-up processing." Essentially, he defines a "pre-attentive attribute" as a fundamental visualization feature that is understood intuitively and immediately by potential viewers, even when they're not paying full attention to the visual's content.

Odds translates that into practical terms as follows.

First, avoid using numbers and text as the core visual elements in your graphic, since they are not understood and retained intuitively without conscious thinking. That principle applies to all potential viewers, not even by the most educated, mathematically inclined, and data-savvy people. This information is retained in very-short-term bottom-up "iconic memory," from which it feeds into "visual working memory," which interacts with "long-term memory" to distill the larger meaning of the visual. However, you should use numbers and text in your scales and in labels, footnotes, and other ancillary roles that add relevant detail and anchor the context to your visual presentation within relationships drawn from long-term memory.

Second, if you want to depict precise quantitative relationships, rely on such pre-attentive features as the length of visual elements and their positioning in two-dimensional grids. In other words, bar charts, line charts, histograms, and scatterplots--utilizing a common aligned scale--are usually the best for presenting the absolute and relative magnitude of continuous variables. If quantitative precision is your intention, Odds advises against using the relative size, width, intensity, slope, angle, area, volume, and/or blur of entities, because these "pre-attentive" features make it harder to glean such information at a glance.

Third, if you want to depict inter-entity relationships, use the "Gestalt" visual grammar of grouping--enclosure, proximity, connection, and similarity--to make these affinities crystal clear at first glance.

Last but definitely not least, observe the "seven, plus or minus two" rule. That refers to the limitations of visual working memory that keep most people from holding more than that many visual elements in their minds at any given time. That doesn't mean you must limit your data points to seven plus or minus two. It simply means that you must be careful not pack your visualizations with more than a handful of each type of design element (lines, shapes, colors, text labels, etc.)--or run the risk of distracting the viewer from the core patterns you're presenting.

Keep those principles in mind as you push advanced visualization techniques into new data-rich applications. If you overload your visuals with distracting "eye candy," you're likely to induce a sort of analytical "sugar shock" in your poor users.

Source:  The Big Data Hub