Spatial Programming: Connecting Across Disciplines
Since encountering spatial programming I have wanted to explore some tentative connections with similar projects in other fields, for the sake of sheer curiosity and cultivating a shared sense of purpose.
This is a draft and will be revised for clarity and completeness in a future version.
A general theme of Lu(ke)'s work is spatial programming.
In SandPond complex particle-based physics simulations can be generated using a spatial programming language named SpaceTode. SpaceTode has symbols that each represent things like atoms, space and transitions. SpaceTode was inspired by the graphical programming environment Stagecast Creator and the spatial programming language SPLAT.
Consider the following rule:
When an atom (
@) is above an empty space (
_) a transition (
=>) will follow in which the atom falls below that space.
This is encoded as:
@ => _
From a set of such rules elements can be defined, such as sand, rock and water, which can then be dropped into a 3D environment to create complex visualisations.
There is also CellPond, which presents a drag-and-drop interface for creating your own rules for generating 2D visualisations.
Programming has been used to make visualisations of space for many years.
When I wanted to make a simulation to show what affects the rate at which particles sink and float I wrote the code in a conventional style:
While this code (in part) generates a spatial visualisation, it cannot be considered an example of spatial programming.
The code is an abstract symbolic representation that bears no salient resemblance to the spatial arrangement that it generates.
Contrast this with a rule in SpaceTode, which encodes that an atom has a tendency to move to the left:
_@ => @_
Such rules are iconic, in they immediately offer us the possibility of a spatial interpretation.
In spatial programming, the rule looks like the result.
Spatial programming then does not just concern the use of code to generate interesting spatial visualisations, although that is an obvious use-case. Spatial programming fosters an intuitive connection between the signifier (some iconic representation) and the signified (a useful result that is generated).
Philosophers, logicians and computer scientists in this field attempt to develop formal systems for describing regions. Typically these regions are spatial but they are occasionally temporal, spatiotemporal or — most broadly of all — metaphysical.
Mereotopology begins with some primitive relation, typically parthood ($P$) or connection ($C$). To signify $x$ is connected to $y$ we can write:
Then axioms are constructed using this notion of connection; for example, it can be stipulated that for all x ($\forall x$) this relation is reflexive:
Every thing is connected to itself.
It can also be stipulated that connection is symmetric:
$$\forall xy (xCy \to yCx)$$
If a thing connects to a second thing, then that second thing must connect to the first.
The arrow here stands for logical implication. Arrows in SpaceTode might instead be understood to signify a causal implication, by virtue of their generating a side-effect in the real world (a cool visualisation).
In a spatial logic a set of these simple axioms are used to define ($:=$) other spatial relations, including disconnection ($DC$) and parthood ($P$):
$$xDCy \ := \ \neg xCy$$
A thing is disconnected from another if they are not ($\neg$) connected.
$$xPy \ := \ \forall z(zCx \to zCy)$$
An x is part of a y if anything that connects to x must also connect to y.
In this way all descriptions of space can be reduced to the idea of connection. Likewise, in SandPond complex spatial visualisations are generated from simple rules.
Spatial logics are like spatial programming languages in that they attempt to encode the concept of spaces in symbols. Unlike spatial programming languages, however, spatial logics emphasise symbolic rather than iconic representations. The formulas $xCy$, $zDCk$ and $jPi$ are just strings of characters. They each represent different topological spaces (x connected to y, z is disconnected from y, j is part of i) but do not resemble them.
Another key distinction between the two is that spatial programming is oriented towards the generative. Spatial programming languages are designed with a view to creating interesting things, whereas spatial logics are designed to clarify the underlying logic of our spatial descriptions.
Like spatial logic, the Dispersed System Formalism (DSF) aims to find a concise symbolic language for describing complex spaces in a consistent way.
DSF, however, is also a bit like spatial programming. Firstly, it is intended to be used as a tool to create interesting things, like food structures. Secondly, in its choice of symbols it is arguably more iconic than symbolic 3.
In DSF the symbol $@$ represents a relation, equivalent to the notion of parthood (topological enclosure). To say $x$ is enclosed in $y$ we write:
Look at that $@$: it resembles a thing (the character 'a') enclosed in another thing (a circle). It is iconic, like the symbols we saw in SpaceTode. The symbol resembles the spatial relation it represents.
Unusually for the area of science from which it emerged, papers on DSF include actual (symbolic) code, which can be used to generate DSF formulas. Unfortunately the code is written in the proprietary Maple software, but I have slapped together a basic version in Python below. The code basically loops through a list of atoms and relations to create the possible binary formulas that result:
# Components of DSF formulas
# Generate binary DSF formulas
# Prints: x@x x@y x@z x+x x+y x+z x/x x/y x/z y@x ...
Here is how a rule might be written in DSF for a system that has an emulsion (oil-in-water) structure:
O / W => O σ W
In DSF, the symbols $/$ and $\sigma$ mean "dispersed in" (think: oil droplets in water) and "on top of" (think: icing on cake), respectively.
The $\sigma$ looks like a cap on top of an 'o' (iconic), while the $/$ signifies the mathematical notion of division (symbolic).
The use of $/$ is conventional for dispersed systems in science; given that it refers to discrete particles ($ n \geq 2$) the colon symbol "$:$" might be a suitable iconic alternative.
The rule states that if oil is dispersed in water the oil will transition to a state in which it is floating on top of the water.
In SpaceTode, this would be something like:
W => O
I am reminded that people in seemingly-distant areas of inquiry often work on similar problems with analogous tools4.
The historian Ursula Klein has written extensively on the role of Paper Tools in science.
A paper tool is anything that can be put on paper, or any other surface, which helps us think about difficult things and imagine new possibilities.
When scientists are asked about the importance of something like a chemical equation they tend to speak about accuracy, precision and quantification.
While these are important features, Klein thinks that such paper tools also enhance our capacity to imagine and invent.
Going to back to TodePond for a moment, once we understand that the rule for sinking is:
@ => _
then it is easy to imagine the rule for floating:
_ => @
The spatial nature of the language both eases the cognitive leap to the resultant visualisation but also the creative progression to new rule-sets. These are positive features of SpaceTode as a paper tool.
Another paper tool — the chemical equation — is not simply a tool for labelling things that already exist in the world, it also allows us to question and imagine how they might be combined.
Once we acknowledge the existence of the elements carbon, hydrogen and oxygen we can formulate a question as follows:
$$C + H + O = \ ?$$
A synthetic chemist recognises the creative potential of such equations, yet they are unlike spatial programming languages in that they are symbolic and not iconic.
Things could have been different...
What we now recognise as standard chemical notation has its origins in symbolic innovations made by chemists from Sweden (Berzelius) and France (Lavoisier).
The English chemist Dalton had presented an alternative notation that was more iconic in nature. He was especially focused on representing the spatial arrangements of atoms, having just proposed an influential atomic model.
The reasons behind the ultimate rejection of his icons in favour of the modern symbolic version were often prosaic. For example, Dalton's icons were more difficult to type-set using the technology of the time.
Later developments in chemistry, such as the discovery of the benzene ring structure, relied greatly on a spatial intuition that could not be realised through a strict adherence to the symbolic notation that was available to chemists at the time.
While symbolic systems can be very powerful — see mathematics, chemistry and programming — they can sometimes present barriers to learning and creation.
Spatial programmers like TodePond are showing how effective and joyful more iconic languages can be when designing visual effects.
Others have proposed that teaching with concrete icons rather than abstract symbols may benefit learners who typically struggle with mathematics.
Spatial programming could be a generative topic for thinkers and practitioners from different fields to explore alternative means of representing space.
Mereotopology is a combination of mereology (the study of parts) and topology (the study of connection).
Formal ontology combines the classical philosophical study of being with the formal apparatus of logic, usually with a view to developing coherent systems of classification (formal ontologies).
The repeated distinction I make between the iconic and the symbolic broadly follows the semiotics of Peirce.
There is no DSF tool that generates visualisations of the structures its formulas represent. This is why I made DSF-racket. While less visually-enthralling than TodePond's creations it is similar conceptually and attempts to make salient the connection between iconic formulas and the visualisations they generate.