7th Conference on Artificial Life and Robotics, Beppu, Oita, Kyushu, Japan, January 2002

• Humans depend to a very large degree on sight to identify and classify objects and their relationships.
  
  
• The visible world is constructed by the physiology of the eye, the optical nerve and the brain.
  
  
• Much of what follows may seem obvious to you. (Sugoi!) The purpose of this paper is to ask you to consider deeply the relationship between computer visualizations of a-life and the information which is conveyed using imagery.
  
  

After Polanyi (1968)...

Emergence/supervenience is the appearance of properties or behaviours of a set of components which may in principle be unpredictable by reference only to the rules governing their individual interactions, but which nevertheless arise through these same rules.

For example...

Reynolds' boids...

• The virtual flock (or school) emerges
• The movement / flocking or schooling "behaviour" emerges

Conway's The Game of Life and its glider...

• The glider's (dynamic) "topology" emerges
• The "movement" of the glider (tied inseparably to its topology in this case) emerges

Tufte (an information visualization expert) writes...

"Comparisons must be enforced within the scope of the eyespan"

• Is it possible that emergent phenomena will only be recognized if very particular visualization methods are utilized in specific cases?
  
  
• What are the features that need to be experienced "within the scope of the eyespan" for us to see a flock of boids or a mobile glider?
  
  
• How might an understanding of these issues assist us to engineer/observe further examples of emergent phenomena?

 

• A recognizable feature of a flock (or school) is the spatial proximity of its members at any moment in time.

Is this a flock?

• Creatures are spread through a volume which encompasses "the flock".

Is this a flock?

The usual visualization strategy for a computer-based "flock" is to compute boid locations in 3D Cartesian space, perspective project these co-ordinates into a 2D plane, and display the result on a flat screen.

Alternatives:

• Preserve 3D location and map this into 2 dimensions by perspective projection (as above) but provide a variable viewpoint

  
  
  
  "The Matrix" - multi-position, single-instant exposures/snapshots
  
  
• Preserve 3D location and visualize in 3 dimensions
  Employ stereoscopy or holography for example
  
  
• Tabulate or list all boid Cartesian co-ordinates!
  
  E.g. (1.0001, -2.534, 3.3356), (2.5654, -3.2111, 4.3254), (8.5129, -5.6778, 3.9533)...
  
  
• Write a description describing the shape and volume of the flock.
  
  E.g. The flock is almost spherical although it has a bulbous protrusion from one end where the density increases. There is one stray creature a little off to the left back from the other members of the flock. A small group of creatures is breaking away from the bottom into a tear-drop shaped sub-flock. The birds in this tear-drop (there are about seven of them) are less closely packed than...

Obviously, the usual method is not the best way to convey the volume of a flock!

But there are less effective methods.

 

Recognizable features of a glider:

• One of four possible configurations / patterns of active cells at any time (& mirrored versions)
  
  
• The pattern's isolation from other active cells

• The glider is constructed in (virtual) 2D space.
  
  
• The neighbourhood relations between cells in the 2D visualization represent neighbourhood relations between automata in the (virtual) connected network.
  
  
• If the visualization neighbourhood relations are randomized, but the virtual network neighbourhood relations are preserved (something which can't be done with physical/chemical systems), whilst there is still (logically) a "glider", it may not be perceived.
  

Alternatives:

• Any technique for representing/encoding a one-bit per-pixel image may be used.
  
  E.g. List active cells indexed from 0 at the top left of the grid: 1,5,6,7,8
  
  
• Describe the shape of the glider.
  
  E.g. The glider is pentagonal and looks like a child's drawing of a house, except that the left wall is missing. It looks a bit like a fish hook lying on its side.

"Panorama, vista and prospect deliver to viewers the freedom of choice that derives from an overview, a capacity to compare and sort through detail." - Tufte

Turing machine implemented in the Game of Life.
Taken from an image by Paul Rendell, April 2000

Frequently visualized state information includes:

• Direction of travel of individual members
  May be indicated by 'head/tail' or 'front' feature (e.g.a triangular body form)
  
  

Craig Reynolds' boids, 1987

• Speed of travel of individual members
  May use 'tail' length, motion blur etc.

Karl Sims' Panspermia, 1990

Frequently visualized state information includes:

• The current state (active/inactive) of a cell
  
  Active cells are usually coloured differently to inactive cells
  (Or inactive cells aren't "coloured" at all, these cells being considered quiescent / background)
  

• History of state transitions for a cell
  
  Recently de-activated cells may be coloured in a colour different to long-term inactive cells

In (virtual) biology, essential elements are processes, rather than static entities and relations

The time-series plot is a frequently employed device for displaying change. For example...

The ubiquitous "Fitness vs. Time" graph

Temporal relationships needing display include all boid and glider characteristics discussed above, including:

• Spatial relationships
• State-based relationships

A visualization tool called "small multiples" may be utilized with some degree of success...

Muybridge's "Animal Movement" photographs are also an example of what may be learnt using this technique.

• But, how do these visualization techniques compare with the power of animation for understandingthe "movement" of these patterns?
  
  
• Do "small multiples" convey the emergentproperties of the flock or glider?
  

Lack of attention to visualization principles will result in a lack of understanding.

But just as dangerously, visualization techniques give apparent "life" and substance to inanimate, abstract patterns.

Consider carefully what aspects of an abstract system need to be visualized and the best way to do this.

Ki o tsukete!

©Copyright Alan Dorin/Animaland 2002