Discovering effective tutorial strategies in human tutorial sessions


We analyze in this paper dialogue based tutorial interactions between human tutors and tutees to understand what distinguishes effective tutorial sessions from non-effective tutorial sessions. In other words, we investigate what effective human tutors do differently from less effective tutors. Towards this end, we characterize and explore human tutors' actions in tutorial sessions by mapping the dialogue based interactions, which are streams of utterances, into streams of actions, based on the language-as-action theory. Next, we use human expert judgment measures, evidence of learning (EL) and evidence of soundness (ES), to identify effective and ineffective sessions. We finally perform a number of tutorial analyses using various methods such as profile comparison, sequence logo analysis and discriminant sub-sequence analysis to present several interesting patterns.

Publication Title

Proceedings of the 31st International Florida Artificial Intelligence Research Society Conference, FLAIRS 2018

This document is currently not available here.