Stephen Rostanti Undergraduate Dissertation 2014/15

Can a robot teach?

Supervised by A.Sharkey

Abstract

An experiment was run in order to determine the effect of a humanoid robot called "Nao" on the learning of children with Autism compared with a traditional human teacher method and an e-learning method utilisng Brian Cox videos.  METHODS:  The quality of the teaching was measured using a multiple choice quiz with questions based on the taught content, the enjoyment and satisfaction which was gauged using a binary yes / no style questionnaire after the quiz had been undertaken. The videos were (1) Black Holes, (2) Searching for liquid water on Mars, and (3) Star death and the creation of elements.  RESULTS:  The results are as follows: T-Tests between the different conditions meant that we could not reject the null hypothesis for the quality of Nao's teaching. However, the null hypothesis for the human teacher could clearly be rejected, and as such, this teaching method was deemed the highest quality. Comparing just these two, the null hypotheses for Nao's and the video's quality of teaching could not be completely accepted. A Chi Squared test for the questionnaire results meant that we could not reject any of the null hypotheses for the enjoyment and satisfaction of any of the teaching methods.  CONCLUSIONS:  We conclude that the quality of Nao's teaching was lower than that of the human teacher's (p = 0.000123) there was no significant difference between the quality of Nao's teaching and the video's (p = 0.5388). We could not find enough evidence to suggest which condition yielded the most enjoyable and satisfactory learning experience (Probability level (alpha) value = 0.4830).