Contribution of eye-tracking to the study on perception of the complexity

Abstract

In studying the way human beings evaluate randomness and produce random objects, cognitive psychology showed that the mind finds it difficult to recognize true randomness as well as to produce it because it is influenced by numerous biases. Studying them can help to understand better the way it is structured. In parallel, mathematics showed that more random objects have a higher algorithmic complexity. Computing lately provided practical means to calculate the algorithmic complexity of objects of finite size and also produced new encounters between mathematics and cognitive psychology by allowing the latter to envision new models for the brain, inspired by algorithmic logic. In this context, our research applied eye-tracking techniques to the study of the perception of complexity. Forty subjects had to order images belonging to ten groups of four according to decreasing (perceived) complexity. The hypothesis was that images with the higher algorithmic complexity would be perceived as more complex as well and would cause longer fixation times. However, experimental results did not confirm these hypotheses as the correlation between algorithmic and perceived complexities was low, and the relation between complexity and fixation time was not linear but closer to an inverted “U” shaped curve. This may be due to contextual effects and to choose images with complexities too close to each other, as subjects found it difficult to order them as requested. Further experiments must then be carried out with conditions better controlled and modified parameters.