Craig's List

Op Ed

Mind Games

L. N.

Be Frank

Mark Up

Sun Burns

Art Tech Statement

Typically, a caricature is drawn using many line segments. A Cable-drawn Caricature is more constrained as the lines are all linked in a one-line wire sculpture. “What is not constrained is not creative” (to quote Johnson-Laird [1]).

Technically, my own bent was to ask how [2] art “works” in the minds of artists and audiences – and specifically to see if cognitive computing might improve my Cable-drawn Caricatures. I started by sketching conventional caricatures, each made with lots of lines. I then turned to science for the missing links.

Mathematically, an optimal solution would minimize the total length of links. This is known as the Traveling Salesman problem: each city (like a line segment) must be visited once and only once (by a link to and from), minimizing the total path length. Iterative algorithms can compute this optimal solution, so I wrote a program that takes a conventional caricature as input and makes a Cable-drawn Caricature as output. Unfortunately, the optimal solutions did not look very good to me – I guess my eye was not buying what the Traveling Salesman was selling.

Psychologically, it turns out that the hard part is not linking the lines – but drawing a “likeable and linkable” set of lines in the first place. By “likeable” I mean a set of lines that captures the likeness of the subject I am drawing. This is the art of conventional caricature. By “linkable” I mean a set of lines that is still a good likeness when linked in one line. This is the art of Cable-drawn Caricature. Realistically, both arts lie beyond the current state of science in cognitive computing.

Coincidentally, a similar situation arose in MITRE research on weapon-target pairing for time-critical targeting. Here the basic problem was also a version of the Traveling Salesman, and an optimization algorithm had been implemented in a command/control system. Our research found that targeteers did not like or use the system, for much the same reason as I did not like or use computer solutions in my Cable-drawn Caricatures. In each case the system was getting an optimal answer in a sub-optimal context, because inputs to the program did not reflect the best knowledge and judgments of a human user. Instead we designed “Pairing Pictures” [3], a support system that leverages perceptual visualization and mathematical optimization together to help warfighters develop weapon-target pairing solutions.

Currently, my work at MITRE continues along the lines of visualization and optimization in Cognitive Computing. To interact effectively, a system must know what looks good to a user. To collaborate effectively, people must agree on what is “likeable and linkable”. These are matters of style [4] – in art and war.

Further Reading  

[1] Freedom and Constraint in Creativity (Philip N. Johnson-Laird). The Nature of Creativity (R. J. Sternberg, editor). New York: Cambridge University Press (1988).

[2] www.ask-how.org.

[3] Structure Mapping in Visual Displays for Decision Support. Command and Control Research and Technology Symposium. http://www.dodccrp.org/events/2006_CCRTS/html/papers/182.pdf.

[4] Creature Double-Feature: On Style and Subject in the Art of Caricature. AAAI Fall Symposium on Style and Meaning in Language, Art, Music, and Design. http://music.ucsd.edu/~sdubnov/style2004.htm.