Evolutionary intelligence : a heuristic method of constructing conscious learning behavior and application to language development in infants
Date of Issue2016
School of Physical and Mathematical Sciences
The topic of learning receives diverse perspectives in psychological cognitive and behavioral analysis. In this thesis, an evolutionary computational model is proposed to integrate both the cognitive and behavioral point of view. The constructed artificial intelligence (AI) was made to play Tic-Tac-Toe against experienced human opponents. The observed learning behavior of the AI, exhibited signs that could suggest ‘conscious learning’ behavior. This dialectic of learning is based on the hypothesis that learning is an evolutionary process (behavioral) mediated by certain internal memory organizing mechanism (cognitive). Genetic algorithm and cellular automaton are the two fundamental logic behind this model. Here in the game model, the AI is able to both successfully figure out rule(s) and strategies of the game. Adopting this approach, a simplified version of the AI is constructed to model infant language acquisition. In this infant model, we are able to mimic an infant’s initial phase of language development through the first three levels of consciousness via social interactions. This result could be of paramount importance in identifying infants with difficulty in acquiring language.