Science in Society

Bryn Mawr College

Emergent Systems Working Group

April 9, 2003
Tim Burke
Something to do with Emergent Art

Speaker's Notes

Additions, revisions, extensions are encouraged in the Forum and/or at
Participants 09 April


Emergence and Creativity

Emergence has proved attractive to sculptors, visual artists working with computer-generated or mediated tools, multimedia performance artists, film makers and computer game makers for a number of reasons. There is a significant and rapidly growing field of artistic works that rely on emergence in one way or another.

In some cases, its uses have been strongly utilitarian and practical, meaning that emergence or autonomous agents (or closely related phenomena like genetic algorithims) have been judged the best way to solve some pre-existing representational problem. In other cases, emergence has interested artists and creators for its intrinsic aesthetic qualities.

I have identified four major principles behind the use of emergence in cultural work.

Organicism and mimesis

Many artists find emergence and autonomous agents attractive because they allow an artist to create a static or interactive work that has an organic, life-like feel, that seems "mimetic", mirroring the everyday sensory experience of the real world without directly representing any actual real-life object or scene.

Artists working with computer graphics have been long been aware that as the sophistication and power of computer animation and imagery has improved, a visual paradox arises. A computer-generated representation of a human being or real-life life form actually seems "less" life-like to human observers as it improves because the distance between how it looks and real-life narrows. Human visual processing will readily perceive life-like properties (like a face) in visual patterns that seem very unlifelike (say, the random patterning of lines on a tree or the shape of clouds) but as the image becomes more life-like, the gap between the inanimate and the organic is exaggerated, both in terms of how it looks and how it behaves. A life-like representation whose every action is prescripted actually tends to look "wrong" somehow to most human observers.

The utilitarian side of emergence and autonomous agents helps bridge that gap. In animated and special-effect driven films where it is important that audiences perceive the action of images of living beings, especially large groups, emergence has become critical. (loads slowly, but stampede appears relatively quickly). This directly uses a version of Craig Reynolds' Boids. (Go to Helm's Deep, click "Massive", short part of film)

However, a lot of artists are also interested in the organicism of emergent phenomenon in purely aesthetic terms. Rebecca Allen and her collaborators have built a program called "Emergence" for multimedia artists, and used it to create a work called "Bush Soul".

(View a bit of "Bush Soul")

Tim Blackwell makes "Swarm Music", using a version of Craig Reynolds' Boids. Blackwell describes emergence as a way to create an improvisational effect in computer-generated music.

Steve Grand, the creator of the computer game Creatures (and its sequels) makes heavy use of emergent principles in order to give players a sense that the "Norns" of the game were lifeforms. In his recent book "Creation", Grand takes a strong philosophical position within debates about artificial life and argues that his Norns are literally alive, not merely simulations of life.

Surprise and the unexpected

Blackwell's use of emergence is also connected to a second major principle governing the use of emergence in creative work, to create unexpected or unanticipated effects. It almost seems to me that this is a subset of organicism, that one of the reasons we readily perceive autonomous agents to be life-like.

However, the topic is of special importance in managing the creation of narrative or storytelling in virtual, simulated or free-form environments. There is a somewhat undeveloped but growing discussion of "emergent narrative" as an idea. The use of the word is not the strict form of emergence that the Working Group has explored so far: it is very difficult to identify what the agents and environment in a narrative system would be. The term is used to describe the post-facto narratives that can be told about the actions of agents in a creative or dramatic system governed partially or largely by emergent principles. Surprise and the unexpected are the key virtues of how such an environment deviates from a prescripted one.

Ruth Aylett, for example, has written about a project called "Virtual Teletubbies" where users in a 3-D graphically simulated virtual environment were able to observe and interact with simulations of the Teletubbies with limited autonomous agent programming. The environment was unconvincing with pre-scripted stories or routines: unexpected actions deriving from emergent behaviors were crucial to the environment working as its creators had hoped. As Aylett noted, though the approach had good results, "one of the risks of emergent narrative is that it may not emerge - the unpredictability that makes it interesting also makes it in some sense fragile. One might add that it is also inherently small-scale: like free improvisation it runs in continuous time, there is no ‘leaving out the boring bits’ as would be the case in a written narrative."

Here we are definitely talking purely about what human observers cognitively identify as "unexpected but coherent" events, rather than the rigorous sense of "unpredictable". Events in a narrative which are completely unpredictable--say, if Macbeth suddenly decides in Act I to start wearing women's clothing and takes to playing croquet obsessively--don't achieve any useful aesthetic result.

The use of even limited emergent principles to generate "aesthetically useful surprise" also plays a role in multiplayer computer games, and here one of the interesting things is the way that the rules of a game actually can turn human beings into autonomous agents, by constraining their behavior to (relatively) simple rules within an environment that cumulatively records their actions much as turtles do in Net Logo. This is something I've written about myself, in this paper, and also the economist Edward Castrovalva has written about this a bit in his work on the economics of virtual worlds.

Interestingly, the metaphor of choice among game programmers who describe their attempts to use emergent narratives is team sports: rule-driven, unpredictable in advance (in terms of the specifics of what will happen during any contest, not of the probable outcomes), and with "natural" sources of tension and drama. After such a game, a viewer can tell a story about what happened, but the story was not prescripted in any way (with the exception of wrestling, of course).

Evolution, cumulation and path dependence

Computer games bring up another artistic and cultural use of emergence, to allow interactivity between audiences and creative works (or between different parts of an artwork) to have some form of memory, so that successive interactions between audiences and the artwork build over time without each interaction needing to be remembered or recorded in a discrete form.

Computer games, especially massively multiplayer online games (MMOGs), make extensive use of emergent effects to create a sense of cumulative change over time. Again, MMOGs effectively use human players as rule-constrained autonomous agents. One of the most striking explorations of this is a 2003 presentation at the Games Developers Conference by Raph Koster, currently one of the lead designers of the upcoming MMOG Star Wars: Galaxies. Koster explicitly uses the network theory of Barabasi and Watts, and the "artificial societies" work of Axtell and Epstein, to talk about how cumulative structures in persistent gameworlds emerge from simple rule-conditioned interactions between players.

Some game engines also use emergent principles to change the game world or elements of it over time, most notably the work of the company Computer Artworks, founded by William Latham and Mark Atkinson. Latham is an early pioneer of "evolutionary art", having written a book about it in 1992. They have marketed various versions of a software package called Organic Art designed to use evolutionary principles and two computer games, Evolva and The Thing, designed to use some agent-based effects to create a sense of cumulative effect over time.

Interactive art, especially installations of various kinds, has been around for a while, but a very large percentage of artists with an interest in interactivity have been drawn to emergent principles because of an aesthetic interest in evolutionary or path-dependent changes over time in the nature of a complex artwork or performance.

David Rokeby's work, some of which uses his motion-capture software Very Nervous Systems, uses emergent principles to express cumulative effects of the interaction between disparate elements of his installations and in some cases audiences or viewers as well, as in the work N-Cha(n)t (2001).

Evolution is a central aspect of some of the work of Christa Sommerer and Laurent Mignonneau, most explicitly in their piece Life Spacies. (This I think demonstrates some of the ways that genetic algorithms and agent-based emergent systems are very closely allied in cultural and creative work.)

Many of Simon Penny's installations are attempts to create emergent behavior over time through the cumulative interaction of separate artistic agents. See for example Sympathetic Sentience

Ken Rinaldo's 2000 installation Autopoesis is another important example of the use of emergent systems theory to create cumulative interactive effects.

Auto-authoring: complexity management

Evolutionary and cumulative effects from the interaction of audiences and artwork, or disparate elements in an artwork or performance, is largely an aesthetic rather than practical driver of the incorporation of emergent systems into cultural work.

A more utilitarian motive is that emergence is a practical strategy for getting aspects of complex cultural works to "auto-author", or generate themselves, freeing the artist or designer to work on more critical or specific systems. Again, this is a principle that applies in particular to computer games, but not only to them. In fact, many of the installations and interactive works linked to in this review use emergent principles to generate a significant portion of their content.

With respect to narrative in general, this is actually a fairly old, pre-emergence concern among AI researchers, both those interested in language recognition and in AI-human communication, getting AIs to create narratives either on their own, or through interaction with human agents. The papers at the 1999 meeting of the AAAI explored some of these issues in depth.

Andrew Stern's program Facade (scheduled for a fall commercial release) seems to me to follow some of these principles in terms of trying to manage complexity while retaining an authorial presence, but I get the impression from reading about it that it is not truly governed by emergence, but by more tightly "top-down" kinds of controls. It's hard to say for sure. In fact both Stern and Michael Mateas' work on narrative intelligence and "expressive AI seem to me to follow non-emergent strategies for managing the complex interaction of multiple AI elements. Mateas is very clear about this in his own work, arguing that merging AI and artistic work creates fuzzy boundaries between "GOFAI" ("good old-fashioned AI") and behavioralist or interactionist AI.

That seems to me to be characteristic in general of what happens when an artist tries to think of ways to achieve auto-authoring for a utilitarian goal of managing complexity in a cultural work: rather than the fairly "pure" cases of emergence using autonomous agents that inform the aesthetic principles of some of the work discussed previously, the solutions here tend to be hybrid, fuzzy, informed by whatever works best, which often includes highly limited uses of emergent principles and insights.

This may also reflect something that I picked up on at the edges of my readings, namely, that "complexity management" as an issue in software design and business organization is largely a mechanistic, top-down, systems-design field, in other words, the opposite of emergence. When artists are explicitly interested in getting parts of a large multimedia piece to generate or create themselves, they may find that some of the tools which come to hand are not based on principles of emergence.


One interesting thing I turned up was the Emergent Art Group. As I understand it, they're artists who are committed to creating art collectively, governed in their interactions by emergent principles, as if they were agents themselves and their art was the result of agent-based interactions with an environment. It's a bit like the way that players in multiplayer computer games and sports exhibit "agent-like" behaviors by virtue of being confined to simple rulesets within constrained environments, but this group essentially elevates this to an aesthetic principle, rather like Dogma 59 does for filmmaking.

Good summaries of art, creative work and emergent systems Dan Collins, "Breeding the Evolutionary"


Home | Calendar | About | Getting Involved | Groups | Initiatives | Bryn Mawr Home | Serendip Home

Director: Liz McCormack - | Faculty Steering Committee | Secretary: Lisa Kolonay
© 1994- , by Center for Science in Society, Bryn Mawr College and Serendip

Last Modified: Wednesday, 02-May-2018 11:57:04 CDT