This chapter offers an overview of symbolic building representations in BIM, including their key differences to analogue representations and how these were implemented in CAD. It explains how a model is built out of symbols that may have an uneasy correspondence with real-world objects and how abstraction can be achieved using these symbols.

Symbols and relations in BIM

BIM^[1] is the first generation of truly symbolic digital building representations. CAD also used discrete symbols but these referred to implementation mechanisms: the geometric primitives that comprised a symbol in analogue representations. In BIM the symbols explicitly describe discrete building elements or spaces – not their drawings. BIM symbols usually appear as “libraries” of elements: predefined symbols of various types. The types can be specific, such as windows of a particular model by a certain manufacturer or abstract, e.g. single-hung sash windows or even just windows. The hierarchical relations between types enable specificity and abstraction in the representation, e.g. deferring the choice of a precise window type or of a window manufacturer to a later design stage, without missing information that is essential for the current stage: all relevant properties of the window, like its size, position and general type, are present in the generic window symbol at a suitable abstraction level.

Entering an instance of any kind in a model normally follows the following procedure:

• The user selects the symbol type from a library menu or palette
• The user positions and dimensions the instance in a geometric view like a floor plan, usually interactively by:
  • Clicking on an insertion point for the location of the instance, e.g. on the part of a wall where a window should be
  • Clicking on other points to indicate the window width and height relative to the insertion point (this only if the window does not have a fixed size)

Modifications of the instance are performed in three complementary ways:

• Changes of essential properties such as the materials of a component amount to change of type. This is done by selecting a different symbol type from the library menu or palette and linking it to the instance.
• Changes in the geometry of an instance involve either repositioning the reference points or numerically changing the relevant values in any of the ways allowed by the program interface: in dialogue boxes that pop up by right-clicking on the instance, in properties palettes, through dimension lines or schedules.
• Changes in additional properties that do not conflict with the type, e.g. the occupancy of a space or the stage where a wall should be demolished, are entered through similar facilities in the interface, like a properties palette. Some of these properties are built in the symbols, while others can be defined by the user.

BIM symbols make all properties, geometric or alphanumeric, explicit: the materials of a building element are not inferred from its graphic appearance but are clearly stated among its properties, indicated either specifically or abstractly, e.g. “oak” or “wood”. Most properties in an instance are inherited from the type – not just materials but also any fixed dimensions: each wall type typically has a fixed cross section. Changing type properties like materials means crossing over to a different type, not changes in the instance properties. This ensures consistency in the representation by keeping all similar windows truly similar in all critical respects. This is essential for many tasks, such as cost estimation or procurement.

Many of the relations between symbols are also present in BIM, even if they are not always directly accessible. Openings like doors and windows, for example, are hosted by a wall. Therefore, normally they can only be entered after the hosting wall has been placed in the representation and in strict connection to it: trying to move a window out of a wall is not allowed. Connected walls may also have a specific relation, e.g. co-termination: if one is moved, the others follow suit, staying connected in the same manner. Similarly, spaces know their bounding elements (which also precede them in the representation) and if any of these is modified, they automatically adapt themselves. Through such relations, some of the possibilities offered by graphs become available in BIM, albeit often in indirect ways. A door schedule (Figure 1) reveals that, in addition to its hosting wall, a door knows which two spaces it connects (or separates when closed).

The explicit symbolic representation of both the ‘solids’ out of which a building is constructed (building elements like walls, floors, doors and windows) and the ‘voids’ of the building (the spaces bounded by the building elements) is important. In analogue representations, the spaces are normally implicit, i.e. inferred by the reader. Having them explicit in BIM means that we can manipulate them directly and, quite significantly from the perspective of this book, attach to them information that cannot be linked to building elements: similarly to specifying that a window is made of oak wood, one can specify that a space is intended for a particular use, e.g. “office”, and even for specific activities like “small group meeting” or “CEO’s meeting room”. Such characterizations relate to various performance specifications, such as acoustics or daylighting, which can also be attached to the space and be used to guide and evaluate the design. Making spaces explicit in the representation therefore allows for full integration of building information in BIM and, through that, higher specificity and certainty. Spaces, after all, are the main reason and purpose of buildings, and most aspects are judged by how well spaces accommodate user activities.

BIM symbols and things

BIM has many advantages but, in common with other symbolic representations, also several ambiguities. Arguably the most important of these concerns the correspondence between symbols and real-world things. Building representations in BIM are truly symbolic, comprising discrete symbols. Unfortunately, the structure of building elements often introduces fuzziness in the definition of these symbols, similarly to the one-to-many correspondence between graphemes and phonemes we have seen in alphabets. In general, there are two categories of ‘solids’ in buildings. The first is building elements that are adequately represented by discrete symbols: doors and windows, for example, are normally complete assemblies that are accommodated in a hole in a wall. Walls, on the other hand, are typical representatives of the second category: conceptual entities that are difficult to handle in three respects. Firstly, walls tend to consist of multiple layers of brickwork, insulation, plaster, paint and other materials. Some of these layers continue into other elements: the inner brick layer of an external wall may become the main layer of internal walls, forming a large, complex and continuous structure that is locally incorporated in various walls (Figure 2).

Secondly, BIM retains some of the geometric bias of earlier building representations, especially in the definition of elements like walls that have a fixed cross section but variable length or shape. When users have to enter the axis of a wall to describe this length or shape, they inevitably draw a geometric shape. BIM usually defines symbols on the basis of the most fundamental primitives in this shape. Even if one uses e.g. a rectangle to describe the axis, the result is four interconnected yet distinct walls, each corresponding to a side of the rectangle. Similarly, a wall with a complex shape, but conceptually and practically unmistakably a single structure, is analysed into several walls, each corresponding to a line segment of its shape (Figure 3).

Thirdly, our own perception of elements like walls may get in the way. Standing on one side of a wall, we see only the portion of the wall that bounds the room we are in. Standing on the other side, we perceive not only a different face but possibly also a different part of the wall (Figure 4). As a result, when thinking from the perspective of either space, we refer to parts of the same entity as if they were different walls.

The inevitable conclusion is that some symbols in BIM may still require further processing when considered with respect to particular goals. One may have to analyse a symbol into parts that then have to be combined with parts of other symbols, e.g. for scheduling the construction of the brickwork in Figure 2. Other symbols have to be grouped together, like the internal wall in Figure 3. Such manipulations should not reduce the integrity of the symbols; it makes little sense to represent each layer of a wall separately. At the same time, one has to be both consistent and pragmatic in the geometric definition of building elements. In most cases, acceptance of the BIM preference for the simplest possible geometry is the least painful option: the vertical internal wall in Figure 4 should be represented as a single entity and not split into two parts in order to simplify the adjacency of walls to spaces. Looking at it in any way beyond the three perspectives indicated in the figure and the spaces that frame them, it cannot be anything else than a single building element.

Paradigmatic and syntagmatic dimensions in BIM

Even with the issues discussed above, the symbolic character of BIM has obvious advantages for the paradigmatic dimension: each symbol is explicit and integral in a building representation. The same holds for the syntagmatic dimension, in three different respects. The first concerns the practical side of developing a building representation in BIM: in common with most computerized programs, BIM editors can record the sequence of user actions and so make the history of a representation accessible and transparent. This allows users to undo actions and backtrack to earlier states.

More significantly, the sequence of user actions is often organized in prescriptive procedures because their order is not trivial. As we have seen, one has first to select the type of a new symbol in a model and then indicate the geometry of the instance. Such procedures ensure consistency in the symbols and their structure, as well as register many relations between symbols, for example the anchoring of a window or a wash basin to a hosting wall.

The third advantage of BIM for the syntagmatic dimension relates to 4D modelling: the addition of a time property to symbols, for example the moment the symbolized element should be constructed. This supports the scheduling of construction, demolition and other real-world activities from within the building representation, and reduces inconsistencies or other errors that emerge from poor communication between building representations and scheduling activities or software.

Abstraction and grouping in BIM

BIM symbols cover a wide range of abstraction levels, from generic symbols like “internal wall” without any further specifications to highly detailed symbols, representing e.g. a very specific wall type, including precise descriptions of materials from particular manufacturers. Usually a building representation in BIM starts with abstract symbols, which become progressively more specific. It is also possible to backtrack to a higher abstraction level rather than sidestep to a different type on the same level, e.g. when some conflict resolution leads to a dead end and one needs to reconsider their options. This typologic abstraction is one of the strong points of BIM but also something one has to treat with care because a model may contain symbols at various abstraction levels. Managing the connections between them, e.g. deciding on the interfacing between a highly specific window and an abstract wall, requires attention to detail. On the positive side, one can use such connections to guide decision making, e.g. restrict the choice of exact wall type to those that fit the expectations from the window.

Symbolic representations also have considerable capacities for bottom-up mnemonic abstraction on the basis of explicit relations between symbols, ranging from similarity (e.g. all vowels in a text) to proximity (all letters in a word). As it typical of digital symbolic representations, BIM allows for multiple groupings of symbols to produce mnemonic structures of all kinds, e.g. selecting all instances of the same door type in a design, identifying all spaces with a particular use on the second floor or determining which parts of a design belong to the north wing. For the latter, some additional input from the user may be required, such as drawing a shape that represents the outline of the north wing or labelling every symbol with an additional wing property. No user input is required for relations built into the behavioural constraints of a symbol, e.g. the hosting of openings in walls.

Through the combination of standard symbol features (like their properties) and arbitrary, user-defined criteria (like the outline of a wing), one can process the representation at any relevant abstraction level and from multiple perspectives, always in direct reference to specific symbols. For example, it is possible to consider a specific beam in the context of its local function and connections to other elements but simultaneously with respect to the whole load-bearing structure of its floor and wing or of the whole building. Any decision taken locally, specifically for this beam, relates transparently to either the instance or the type and may therefore lead not only to changes in the particular beam but also reconsideration of the beam types comprising the structure, e.g. a change of type for all similar beams. Reversely, any decision concerning the general type of the structure can be directly and automatically propagated to all of its members and their arrangement.

The automatic propagation of decisions relates to parametric modelling: the connection of symbol properties so that any modification to one symbol causes all others to adapt accordingly. In addition to what is built into the relations between types and instances or the behaviours like hosting, one can explicitly link instance properties, e.g. make several walls remain parallel to each other or vertical to another wall. One can also specify that the length of several walls is the same or a multiple of an explicitly defined parameter. Changing the value of the parameter leads to automatic modification of the length of all related walls. Parametric design holds significant promise. People have envisaged building representations in which it suffices to change a few values to produce a completely new design (or variation). However, establishing and maintaining the constraint propagation networks necessary for doing so in a reliable manner remains a major challenge. For the moment, parametric modelling is a clever way of grouping symbols with explicit reference to the relation underlying the grouping, e.g. parallelism of walls. Still, even in such simple cases, the effects of parametric relations in combination with built-in behaviours can lead to unpredictable and unwanted results.

In views which replicate conventional drawings, BIM software often also incorporates visual abstraction that mimics that of scales in analogue representations. By selecting e.g. “1:20” and “fine” one can make the visual display of a floor plan more detailed than with “1:200” and “coarse”. Such settings are useful only for visual inspections; they alter only the appearance of symbols, not their type or structure.

 

The LoD in BIM is also related to abstraction. LoD specifications attempt to standardize the specificity of information in a model or preferably in a symbol, as a model may contain elements at various LoD. Many LoD standards have been proposed but strict adherence to them is a throwback to analogue standards regarding drawing scale. Such adherence fails to appreciate that information in a model has a reason and a purpose: some people have taken decisions concerning some part or aspect of a design. The specificity of these decisions and of the resulting representations is not accidental or conventional. Rather, it reflects what is needed for that part or aspect at the particular stage of a project. The LoD of the model that accommodates this information can only be variable, as not all parts or aspects receive the same attention at the same stages.

Specificity should therefore be driven by the need for information rather than by convention. If information in a representation is at a higher specificity level, one should not discard it but simply abstract in a meaningful way by focusing on relevant properties, relations or symbols. A useful analogy is with how human vision works: in your peripheral vision, you perceive vague forms and movement, e.g. something approaching you rapidly. If you turn your eyes and pay attention to these forms, you can see their details and recognize e.g. a friend rushing to meet you. As soon as you turn to these forms, other parts of what you perceive become vague and schematic. In other words, the world is as detailed as it is; your visual system is what makes some of its parts more abstract or specific, depending on your needs. By the same token, the specificity of a building representation should be as high as the available information allows. Our need for information determines the abstraction level at which we consider the representation, as well as actions by which we can increase the specificity of some of its parts.

Implementation mechanisms in BIM

Despite its symbolic structure, BIM uses the same implementation mechanisms as CAD: the same geometric primitives that reproduce the graphic appearance of analogue representations. The key difference is that these primitives are just part of pictorial views, in which they express certain symbol properties. The type of a door, for example, is explicitly named, so that we do not have to infer its swing from the arc used to represent it in a floor plan; the width of a wall is a numerical property of its symbol, so that we do not have to measure the distance between the two lines indicating the outer faces of the wall. On the contrary, this distance is determined by the width property of the symbol.

As we have seen, however, implementation mechanisms still influence the structure of a building representation in other respects: a wall is still partly determined by drawing its axis and so by the geometric shape one draws. On the whole, therefore, one should consider BIM as largely immune to undue influences from implementation mechanisms but at the same time remain aware of persistent geometric biases in both building representation in BIM and in the mindset of BIM users.