A Framework for Exploring the Evolutionary Roots of Creativity Hrafn Th. Thórisson Latrastrond 7, 170 Seltjarnarnes, Iceland [email protected] This paper describes a development framework aimed at quantifying creativity and a simulation apparatus that allows its mechanisms to be traced to its evolu- tionary roots. The theoretical foundation of the framework is described, along with methods for addressing fundamental questions regarding the origins and nature of creativity and its relationship to logic. The method employs virtual worlds created using cellular automata. The worlds are inhabited by creatures with a cognitive system designed to resemble natural insects such as bees or ants. Creativity is operationalized by measuring the number of distinct plans used by the creatures; diversity of the environment is defined by the number of components that make up the environment and the total number of possible in- teractions between them. Primary hypotheses to be addressed are that environ- mental complexities govern the evolution of mental creativity and that creativ- ity is bound by innate rules of the environment. 1 Introduction Creativity is one of the most important abilities of humans. Our creative skills and ability to express it distinguishes us in obvious ways from the rest of the animal king- dom. Through creativity the human race has invented complex technology enabling it to acquire almost all abilities evident in other animals, as well as to go beyond them (e.g. interstellar travel, infrared vision). Considering the importance of creativity to our existence, surprisingly few studies have been directly aimed at understanding the general underlying structure of creativity mechanisms. By exploring the roots of creative mechanisms I hope to be able to understand by which means it is best to con- struct a general creativity system which can be applied in various different intelli- gence systems. To get around the inherent limitations of human-subject experiments I am building a simulation environment for exploring alternative theories of creativity and setting up an environment where higher-level psychological variables and cognitive functions, such as perception and planning, are under direct control of the experimenter. The primary hypotheses of interest here are that the evolution of creativity mecha- nisms are directly related to the complexity/diversity of the environment, and that the underlying mechanisms of creative behavior are essentially the same in all creatures, despite the obvious human advantages mentioned above. Depending on (a) the num- ber of individual components in the environment and their interactions, (b) the crea- tures' ability to perceive their environment, and (c) the creatures' set of operators which can by applied to modify their behavioral pattern (in particular, to make plans), an increase in environmental diversity will result in a larger set of distinct behavioral patterns (plans) and, if the world is random, there will be no persistent structure evi- dent in the composition of plans. 1 A secondary hypothesis to be addressed states that creativity is bound by rules inherent in the environment. The hypotheses are grounded in the conjecture that creativity is evolution’s answer to perceptually apparent unpredictability. This is different from some hypothetical or “actual” unpredictability – the discussion here revolves around unpredictability from the standpoint of the creature. A simplified version of the conjecture can be stated as follows: Creatures inhabiting a simple, closed and static world would tend to evolve to become completely robotic due to predictability of the environment; if an environ- ment is very simple the cognitive system of the creatures require little or no effort to evolve mechanisms for survival. In a complex world, where events in the environ- ment are not evident entirely by observation of the current situation, evolution must provide organisms with a mechanism to predict in uncertain situations. With the in- crease of interacting components in the environment, more complex cognitive efforts are thus required to produce and assess reactions to the current situation. The paper is organized as follows: The related research and the theoretical founda- tion of the background hypotheses is discussed, followed by a short introduction to the terminology used in this paper. Then we present an example simulation environ- ment for testing these hypotheses. At last we propose mechanisms for quantifying plan diversity and environmental complexity that will serve as the basis for quantita- tive measures of creativity. 2 Related Research Peter Carruthers [2] has discussed the pretend play evident in children, where they enact adult situations – such as using a banana as a phone. He proposes that the func- tion of the extensive creative play of children, also evident in the behavior of other mammalians where the young engage in pretend-play such as hunting and fighting, is to train the young in imaginative thinking for use in adult activities. I propose an ex- tension to this hypothesis: During the initial stages of the development of more ad- vanced animals such as mammals, creativity is yet to be bound by intrinsic rules of the environment. The gradual adjustment of creative mechanisms and internal repre- sentations to environmental structures impose logical reasoning and the ability to un- derstand. The process of understanding, as defined by Baas & Emmeche [3], is related to, or equal to, explanation. Understanding complex adaptive systems (such as the Earth’s environment) requires the discovery of causal relationships between different levels of organization. These different levels of organization are components and 1 The complexity of the environment and the organism’s ability to perceive the environment are equally important aspects of the evolution process, since varying complexity of the environ- ment will make no difference if a creature is unable to perceive the variation. Vision is believed to constitute a significant proportion of information reception in humans, and I have selected the visual mode for my simulated creature's only sensual organ. emergent, multi-leveled hyperstructures [3] which constitute the environment in a hi- erarchical manner, e.g. from atoms to raindrops to rivers. Recent research has shed light on patterns spontaneously emerging at different scales in the synaptic-firing of neurons in the human brain [4]. They describe these patterns in the following manner: “These songs [synaptic firing patterns] resemble spiking correlates of sequential behavior, like bird songs or spatial navigation, and have compressing dynamics, as if the circuit was replaying and modifying previously learned sequencestivity ... the neocortex can spontaneously generate precisely reverberating temporal patterns of activation, dynamic ensembles that could represent endogenous building blocks of cortical function." [4]. These patterns could indicate the application of internal rule-structures to neural clusters representing external phenomena, aimed at the discovery of new conceptual relationships. The present framework provides a future foundation for observing and comparing any behavioral patterns which emerge through the composition of primi- tive actions. Stephen Thaler [5] has created what he calls a “creativity machine” by perturbing connections in artificial neural networks and thereby creating ‘noise’ which manipu- lates learned concepts. A system inhabiting an ability to recognize abstract rules in the environment and use efficiently to manipulate conceptual structures (neural clusters) would behave similarly except for applying ‘noise’ in the form of “rule hyperstruc- tures” to produce more ideas which correlate with the environment. 3 Theoretical Foundation Earth's environment is governed by rules that can be traced down to a sub-atomic level, resulting in its emerging attributes such as land and sea, water circulation and ecosystems. As creatures evolve in worlds governed by these rules their plan-making mechanisms come to reflect them. If a particular world should on the other hand be random-based, then a creature’s planning ability might produce more diversity, but there would be no logic as to when or where the plans are applied. This makes it im- portant to measure the complexity of the world in addition to the diversity of the creatures. I propose to measure the structural complexity by the use of cellular auto- mata [1] where, through simple interacting rules, complex overall behavior can be produced (see section 5). The fact that there are general rules in the environment, many of which are inde- pendent of (perceived) scale, indirectly lends support for the proposed hypotheses. By adapting to – and adopting – the rules inherent in the environment, creatures might manipulate and subsequently apply rules internally (imagination) to different situa- tions or problems. An externalization of such an internal representation would be re- garded as creative behavior if a new causal relationship is discovered and applied (e.g. when a creature discovers new means to fulfill it’s goal). It is safe to presume that animals do evolve consistently with these rules since failure of their adoptation would tend to prove fatal to the creatures and thereby result in the destruction of their genes. Creativity’s roots, according to this line of thought, derive from the interaction of the environmental structures and their components: If the interactions of these struc-