Robot tutor and pupils' educational ability: Teaching the times tables

Abstract Research shows promising results of educational robots in language and STEM tasks. In language, more research is available, occasionally in view of individual differences in pupils’ educational ability levels, and learning seems to improve with more expressive robot behaviors. In STEM, variations in robots’ behaviors have been examined with inconclusive results and never while systematically investigating how differences in educational abilities match with different robot behaviors. We applied an autonomously tutoring robot (without tablet, partly WOz) in a 2 × 2 experiment of social vs. neutral behavior in above-average vs. below-average schoolchildren (N = 86; age 8–10 years) while rehearsing the multiplication tables on a one-to-one basis. The standard school test showed that on average, pupils significantly improved their performance even after 3 occasions of 5-min exercises. Beyond-average pupils profited most from a robot tutor, whereas those below average in multiplication benefited more from a robot that showed neutral rather than more social behavior.