Minimal Creativity, Evaluation and Fractal Pattern Discrimination

Evaluation is considered to be an important component of any creative process. This paper explores how evaluation can be incorporated into our minimal model of creativity, which we have been developing using a combination of conceptual analysis and evolutionary robotics. Specifically, we consider how to extend our approach so that the robots themselves can evaluate mark patterns that they, or other robots, have made on the floor of their environment. Evaluation can, we suggest, be distinguished into a descriptive and a merit assignment component. To evaluate an object or event F is, (a) to discriminate some feature of F and (b) to assign some merit or de-merit to that feature of F . Component (a) is descriptive and (b) is what would more traditionally be called ‘evaluative.’ In simulation, our robots discriminate fractal from random patterns and demonstrate this by stopping on target regions of the arena floor that are covered with a fractal texture. We argue that in so doing they perform the descriptive component of evaluation. However, it is debatable whether the robots are performing the second, meritassignment component of evaluation. Currently, the discrimination mechanism is hard-wired and does not develop during an agent’s lifetime. In future experiments we will investigate how to artificially evolve agents that perform what might be described as ‘minimal evaluation’ by attempting to incorporate a preference element.