Abstract
Dehaene has posited that humans possess a set of core geometric abilities regardless of culture, language or education. In this paper, we develop a computational model (called CoreGeo) of core geometric abilities based on a more general theory of fractal analogical reasoning. Fractal analogical reasoning enables the calculation of confidence in an answer and the automatic adjustment of level of resolution if the answer is found to be ambiguous. We present results from running CoreGeo on Dehaene’s test of core geometry problems. We also compare our results with Dehaene’s results from different cultures and show that CoreGeo performs about as well as humans at core geometry tasks.
