Overcoming the Crisis In Graphic Disciplines
Opinion by Alexander Yampolsky
There is a crisis in teaching graphic disciplines has been discussed for a long time now. Articles have been written on its causes, its consequences, and ways to overcome it. However, the marginalization of graphics departments in technical universities continues.
Traditionally, descriptive geometry is considered to be the theoretical basis of graphic disciplines. Its practical application has diverged in two directions.
Within the research direction, geometrically complex spatial problems are solved using operations on flat images (projections).
Within the communicative direction, projections are considered elements of language. The language of projections is considered the tool for creating common, consistent ideas of the spatial organization of objects.
Both directions are in deep crisis. The crisis of the research direction is a consequence of the expansion of computer technology, which devalues the drawing method approach to solving geometric problems. Nothing can be done about this, probably; we just need to leave the field to the mathematicians and the programmers.
The communicative role of descriptive geometry is related to the use of technical drawings. The drawing must be able to be
Transmitable -- capable of conveying our thoughts to others, and us perceiving the thoughts of others
Imaginative -- capable of evoking spatial representations of objects, based on their images on a plane
Feasible -- capable of manufacturing real objects according to the drawings
The theory states that the task set has a universal solution in the form of a projection (ie, a method of exact projections). The crisis in communicative direction is that this statement diverges from both practice and the standards in force for making drawings.
In essence, there is only one requirement, without which the others lose their meaning: the drawing must be understandable. Projection drawings do not meet this requirement. According to one country’s drawing standards named "Rules Preparing Working Documentation," all images illustrating the standard are clearly not projections, or else are projections about which it is impossible to say for sure whether they are accurate or not. In any case, another standard, "Application of Dimensions," prohibits measurements in drawings. The prohibition of measurements and the rejection of visual similarity with real objects indicates that all images in construction drawings should be considered as conventional symbols, the signs of a technical language.
Researching this language is something that graphics departments ought to be doing. It is something that might make it possible to get out of the impasse.
Concept-oriented Approach is the Basis of Drawing Language
A typical drawing is shown in figure 1, which I got from a set of standards, in this case for rebars. I choose it, because it illustrates the direction in which construction drawings are going.
The standard defines the drawing as a document intended for the execution of construction and installation works, or the manufacture of structures, products, and units. In essence, the drawing in Figure 1 is an technical task for manufacturing, in this case rebar meshes.
The parametric style of this task is obvious. It includes a generalized (conceptual) diagram with numerical and symbolic parameters tied to the elements of the diagram. The values of the symbolic parameters are specified in tabular form. This task in parametric form has all the characteristics of pictographic text (see Figure 2):
A distinctive feature of parametrics is that they are undemanding in geometric accuracy. The drawing of Figure 1 has no precise projections, not even any straight lines! Nevertheless, judging by the fact that the drawing was taken from a country’s standard and so used in thousands of real projects, its "inaccuracy" worries no one. We arrive at this conclusion: the accuracy of depictions does not affect accuracy during manufacturing.
As we can see in Figure 1, a single conceptual diagram represents a lot of objects that have the same set of key parameters. This design technique radically cleanses projects from information noise.
Let's look at another example of using this technique. In Figure 3a we have the reinforcement design of the floor overlap for a multi-story building. Figure 3a can be cleared of much information noise, as shown by Figure 3b.
It shows, in detail, the rebar mesh layout for just one typical area, which measures 6000 by 3000mm (approximately 20’ x 10’). In other areas, only the mark and the number of meshes are indicated.
A good drawing contains exactly as much information as is necessary for its understanding and subsequent implementation. Anything beyond what is necessary is information noise.
Parametric style, rejection of visual copying, elimination of information noise -- all these are direct consequences of the concept-oriented approach underlying the creation of drawings.
All regulatory and reference documents are based on the concept-oriented approach; as a result, they fit organically into the context of drawings. For example, in the drawing of Figure 1, all questions about how the rebars connections are resolved by referring to the standards document defining the technical requirements of welded joints.
With that said, we can give a general formal definition of a working drawing:
A working drawing is a conceptualized parametrically-represented technical task, consisting of fragments of ordinary and pictographic text.
I would like to note that all the drawings illustrated above can be used, without any change, for producing digital models of products and structures.
Confirming The View of Drawings As Text
It is customary to divide project documentation into two parts, graphical and text. The first part, graphics, includes the drawings, and the second text-based documents. However, the drawings themselves consist of graphics and text fragments. If, as it turns out, there is no fundamental difference between these fragments -- that both graphics and text simply complement each other -- then we can arrive at uniform, homogeneous, essentially textual drawings and, accordingly, to uniform, homogeneous, essentially textual project documentation.
Below I have listed the properties that confirm the similarity of ordinary (alphabetical) text and graphic images (pictographic text) on drawings.
Discrete entities. The components of text are easily identified -- letters, words, sentences, paragraphs, chapters, and so on. Pictograms in drawings have the same properties.
In difficult cases, such as when pictograms overlap one another, we have graphical techniques that allow us to separate them.
Insensitivity to image accuracy. Printed and handwritten text is equally understandable in terms of meaning. Similarly, pictograms (for example, the reinforcement mesh drawing of Figure 1) do not require careful alignment of lines, observance of proportions, angles of inclination, etc.
On the other hand, there is a threshold of accuracy and precision required, beyond which text and pictograms become incomprehensible. Sloppy printing and careless drafting are unacceptable.
Distinguishability and conventionality, instead of naturalism. The linguistic signs of an alphabetical text (e.g. words) have conventional (contractual) connections with the objects they replace. The word “rebar” is not a physical steel rod; we agree that the word is a sign for the real-world object.
The same applies to the signs used for pictographic text – pictograms. Visual similarity of pictograms with the objects they represent is not mandatory; the only aspect that is mandatory is that we are able to distinguish between the pictograms, and so pictograms denoting different objects must be visually different from one another. See Figure 3, for example, and the pictograms of reinforcement meshes and openings on the overlap drawings.
Conceptuality. Below is some text typical of that found in drawings:
Reinforce the partitions with C1 mesh every 5 rows of masonry.
It contains fragments of concepts which are signs of language:
partitions
reinforcing
C1 meshes
rows of masonry (bricks)
This rather short notation is sufficient to cover all questions a builder might have when constructing reinforced partitions for a multi-story building. We can make the assumption that the mason's head contains the knowledge that allows him to reconstruct the full picture of the reinforcement task, based on these key fragments. He can fill in the details he needs, which are not mentioned in the text, such as protective layers, overlaps of reinforcement meshes, and so on.
Figure 4 shows a graphic conceptualization. In this case, the conceptual fragments are the elements of equipment that creates a load on the overlaps (shown by the dark colors), along with elements for openings in the overlaps (shown by light colors). All details that are not related to the main idea of the drawing are removed.
Hierarchical structuring. Division into sections, chapters, paragraphs, sentences, words are the integral properties of ordinary text, which makes it possible for us to understand it. Images (pictographic text) have the same properties. Figure 5 shows examples of structuring pictographic and traditional text.
The floor plan in Figure 5a can be considered to be a complex pictogram, consisting of pictograms of walls, rooms, equipment, and so on. In this case, the colors highlight pictograms of the rooms in a corner apartment (pink) and the areas located along the evacuation route from the apartment (hallway = orange and stairwell = green).
The text in Figure 5b lists the requirements for doors installed along evacuation routes. To make it easier to understand, the text is structured hierarchically using functions like "obligation," "prohibition," "execution," "equipment," "disjunction," and "conjunction."
Contextuality. It is impossible to understand ordinary text without taking into account the context in which it is "pronounced." Drawings as a whole and graphic images on drawings have the same property. The drawing in Figure 1 was understood because
It is part of the "Structural Solutions" section of project documentation
It is a reference drawing for the reinforcement drawings of structures
It contains the word "Meshes" in the title block
So, we can see that similarity in a number of features gives grounds to speak of drawings as a kind of text in natural language. Both languages, drawings and natural speech, are created for collaboration based on understanding. We can assume that the common properties of these languages – conceptuality, hierarchical and contextual dependence, and so on (see above) – are key to understanding any text in any language.
3D Models Instead of Drawings?
Graphics departments see a way out of the crisis by abandoning descriptive geometry and reorienting the educational process towards teaching methods of 3D modeling. It is common to express the opinion that traditional drawings are useless.
There is no doubt that 3D modeling successfully copes with solving the analytical side of design, solving problems involving strength, heat, geometry, and so on. However, the ability of digital models (as well as their real prototypes) to explain themselves is far from obvious. Explanations are impossible without conceptualization. But in the so-called digital models, unlike linguistic ones, there are not and cannot be abstract objects (concepts). There are only specific floors, specific walls and overlaps, specific reinforcement bars. Figure 6 shows the consequences of literal replicating the real world: information noise.
A less radical view is that drawings are needed, but they can be obtained automatically from 3D models. Following this logic, to obtain a drawing like the one in Figure 1, we first have produce digital models (digital twins) of three reinforcement meshes (C1-10, C1-11, and C1-12). A predictable question is, "What should the technical task for producing these models look like?", which emphasizes the absurdity of the idea of extracting schematic drawings from details 3D models.
In real design, 3D models are used to obtain missing knowledge about an object. The process of knowledge acquisition consists of the following steps:
A model is built based on the initial specifications
The model is tested
The specifications are adjusted, based on the data obtained from analyzing the model for strength, and so on
The process is repeated until acceptable results are achieved. The result is recorded in a form that is generally understandable, for example, in the form of drawings.
Feasibility of Drawings
Drawings are feasible only insofar as it is possible to establish a connection between the task contained in the drawing and the existing technology for executing the task. The technology for executing can be the following
Step 1. Mental – creates a mental model
Step 2. Computer – builds a digital model
Step 3.Production – produces the real world objects
To establish a connection with an existing technology, it is sufficient to provide information containing a conceptual diagram, key parameters, and a link to the technology. For example, for a typical floor slab (this is a complex, heavily reinforced structure), the following should be indicated:
On the plan – the outline, overall dimensions and binding of the slab to the coordinate axes
On the section – the elevation mark(s) of the slab
In the specification – the series number, designation, and standard size of the slab
If there is no special technology for a complex task, then we use the hierarchical decomposition method. The task is broken into simpler tasks by referring to detailing drawings. The simplification procedure is repeated until all tasks in the drawing receive technological support.
If we establish that drawings are text in a natural language, then the procedure of linking the text with the technology for creating anything is precisely what is called the interpretation (understanding) of the drawings.
Human Interpretation. In the case of human interpretation (human understanding), we are dealing with the translation of natural language into a low-level (at the level of human brain cells) individual language. As a result, we obtain a technology (algorithm) for producing mental objects. This is the algorithm allows the mason to imagine in detail a partition that does not yet exist, and then turn his image into the real structure.
This kind of a mental algorithm is able to handle problems. For example, when a crack appears in the wall, the engineer adjust his mental model, tests it, and issue a conclusion, such as, "Patch the crack with mortar" or "Order a foundation reinforcement project."
Machine Interpretation. When we turn to machine interpretation (machine understanding, knowledge computation), we see that it is the automatic translation of natural language into low-level (executable) machine language. The output is a procedure that creates digital models. Let’s look at a machine understanding algorithm with the example of interpreting raster drawings. The sequence of operations can be as follows:
Load raster image
Extract (recognize) alphabetic and pictographic fragments
Convert raster fragments into graphic primitives, such as texts lines, polylines, and dimensions
Determine the semantic role of each primitive: view boundaries, view names, coordinate axes, structural elements, and so on
Link dimensions with nodal points of structural elements, and determine the exact coordinates of the nodal points
As a result, we obtain a geometrically accurate vector drawing from an inaccurate raster image. The "understanding" procedure can be continued, eventually arriving at a list of low-level commands for constructing a 3D model. I wrote about a program that does machine interpretation of accurate vector drawings and then constructs a 3D model in 2012; see "3D interpreter of construction drawings" at http://3d-int.ru.
As a final result, we can obtain a control program for producing real objects with CNC machines.
Overcoming the Crisis Through a Linguistic Turn
The term "linguistic turn" refers to the history of philosophy and is associated with a rethinking (bringing to the forefront) of the role of language in solving philosophical problems. The turning point was the idea that philosophy contributes not to the multiplication of knowledge, but to its understanding. The original purpose of philosophy was to pursue science (as in Ancient Greece), but then science became its own discipline, leaving philosophy to pursue understanding. The linguistic turn in relation to graphic disciplines consists of a rethinking of the role of language in solving problems in building design.
There are two processes at the heart of design:
The first is acquisition
The second is dissemination of knowledge
It is impossible to say which is more important: the ability to find the right technical solution, or the ability to explain your intentions, to convey your knowledge to others.
The only means we have of transmitting knowledge is through language. When we implement the linguistic turn, we create a paradigm shift: from the paradigm of "It’s all about geometry" to the paradigm of "It’s all about language." In essence, we are talking about abandoning the language of projections in favor of the means and methods of natural language.
The crisis in the graphics departments does not mean a crisis of real drawings and real design. Drawings, like projects in general, have always been linguistic models. However, the lag of theory behind practice is has its harms. For example, the idea of drawings as 2D models, which underlies modern CAD, is an anachronism from the time of Gaspard Monge. What it leads to can be seen by comparing drawings taken from this country standard of 1980 and 2018 in Figure 7.
Practical Consequences Of The Linguistic Turn
Technical documentation as a whole can be represented as a multi-level pyramid of knowledge. Each lower level is a concretization of the one above it:
At the top of the pyramid are federal laws and government regulations
Below are industry standards, norms, and rules
Below those are typical technical solutions
At the base of the pyramid is detailed working documentation
The linguistic turn cancels the definition of project documentation as a special type of technical documentation. All types of documents, from government regulations to detailed drawings of working projects, fall under one general definition: an alphabetic-pictographic text in a natural language. As a result of the linguistic turn, we arrive at a single knowledge base, including normative, reference, design, archival, and other documentation. Working with this basis provides ample opportunities for research in the most in-demand areas of real design. Below is a list of areas.
Collaboration and its basics. Knowledge, understanding, and explanation; conceptualization, hierarchical decomposition, and contextual identification of objects; sign systems and the role of language; language standardization; formats for storing and disseminating knowledge.
Sources of ready-made knowledge. Textbooks, reference books, regulatory documents, standard projects, archival documents.
Investigations during the design process to obtain missing knowledge. Simulation modeling; modeling results in the form of data; data processing and transforming data into knowledge.
Algorithms for preparation and reading of drawings. Textual, natural language essence of drawings; ordinary and pictographic text in drawings; parameterization; mental, computer and production technologies for implementing drawings.
Machine knowledge processing. Machine-understandable documents; machine natural language processing, machine translation; relevant knowledge retrieval; decision support systems; automatic checking of drawings; machine interpretation of drawings.
This list can become part of an academic discipline that examines general issues of design technology. As far as I know, it is not taught in engineering faculties. Graphics departments have every reason to occupy this niche.
© 2025 by Alexander Yampolsky
Model vs drawing, and the challenge of 'clearing away information noise' is fascinating. The conflict between model everything, and model/visual clutter is real. Seems there should be a tech solution—along the lines of (the dreaded) Level of Definition—where there is also a Level of Representation that goes from Symbolic to Model.