The Chart Changed What a Reader Could See
Early statistical graphics showed that a representation can make relationships apparent that remained difficult to notice in the same figures set as a table.
Put a column of annual figures in front of a reader and ask when the underlying pattern changed. A careful reader can compare the numbers one by one. Draw the same figures as a line and the change of slope appears almost at once. The information has not increased. The work needed to notice one relation has changed.
This matters to knowledge engineering because choosing a representation is never a neutral last step. The same claim may be difficult to inspect as prose, simple to compare in a table, and easier still to challenge when it is plotted against time. A system that stores every fact correctly can still fail the people who must understand its consequences.
Playfair designed a way to compare
William Playfair’s Commercial and Political Atlas of 1786 used line and bar graphics to present economic information; his Statistical Breviary of 1801 included a pie chart. Tables of economic figures already existed. The innovation was a visual arrangement that let readers compare quantities through position, length, and shape. The Library of Congress preserves and describes these works and their place in the history of statistical graphics. (Library of Congress on Playfair)
A chart does not think for the reader. It gives the reader a different task. Instead of remembering several entries while searching for a relation, the reader can inspect a stable field and ask where the pattern rises, falls, or diverges. That change can support discovery, but it can also mislead when scales, categories, or omissions distort the comparison. The designer has to make the visual structure faithful to the question.
Later graphic work made this more ambitious. Charles Joseph Minard’s 1869 drawing of Napoleon’s Russian campaign brought geography, direction, army size, time, and temperature into one account. The graphic is not important merely because it fits many variables on a page. It makes the loss of people along a route perceptible in relation to the geography and the falling temperature. A separate table for each variable would force the reader to perform that integration alone. (Library of Congress, “19th Century Colonization and Slavery in Charles Minard’s Flow Maps”)
Machine structure and human structure
Imagine a modern graph containing suppliers, components, tests, and applicable standards. A database can traverse the connections. A person investigating a failed decision needs to see which path the application took, what evidence supported each step, and where a different relationship would have changed the answer. Displaying every node at once is the visual equivalent of handing the reader all the tables at once. Completeness becomes noise.
The knowledge engineer needs a machine model with exact identities and relationships. The designer needs to decide which of those relations the person must inspect now, which can remain available one step away, and which uncertainty must stay visible. Both are selecting a representation for a task. They differ in the kind of intelligence that will use it.
This visual history does not mean every knowledge interface should be a chart. A list may be best for finding a known document. A timeline may expose a changed rule. A map may reveal jurisdiction. A comparison table may be more honest than an animated network. The form earns its place when it makes the necessary relation easier to examine without claiming more than the evidence supports.
By the twentieth century, graphic designers were deliberately organizing type, image, symbol, and sequence as ways of making public information understandable. One of those projects gave its central designer a title that sounds uncannily close to knowledge engineering: the Transformer.
