Abstract:

Existing research stresses the importance of equipping learners with skills that are relevant to the 21st century including how to effectively use digital technology in the context of learning (OECD, 2015; Redecker et al., 2012; UNESCO, 2011). As a result, educational resources and learning activities are increasingly being disseminated to students through digital learning environments (DLE) (Godwin-Jones, 2012). DLE have been developed with the objective of providing education that is adapted to the characteristics, needs, and behaviors of the user. Labeled as intelligent tutoring systems (ITS), these software are generally equipped with artificial intelligence algorithms that are capable of detection, comprehension and adaptation to produce learning sequences that are tailored to the progress of the learner. Despite the many years of research and case studies on the implementation of ITS, little is known about the effects of ITS on the quality of education, nor is there a general understanding of how to apply AI maximize learning outcomes, especially in school environments. Thus, our presentation aims to discuss results from an undergoing literature review that assesses current and recent developments in ITS research. In that context, our research question is: "How effective are intelligent tutorial systems on the learning of primary and secondary school (K12) students?"
To answer our research question, we targeted keywords associated with ITSs, education, and learning at the primary and secondary school level, and keywords associated with learners. We chose to work with commonly used databases in education, referring to Scopus and ERIC USDE (Education Resources Information Center U.S. Department of education) for the final search and the retrieval. ERIC USDE targets specifically education related papers and Scopus, a more general database, where computer science papers are referenced.
So far, we can gather from the extracted studies that they almost all measure something different, even beyond the grade and discipline of the subjects. Most measure the performance of the students, as is expected from the inquiry, but in over 70% of the studies they compare different characteristics of ITS instead of comparing their systems to offline learning. Most of the studies also takes place in a very short period, between one class period and up to four weeks. Thus, at this point we observe that Van Lehn (2011) and Kulik & Fletcher (2016) produced knowledge on general ITS effectiveness compared to other learning methods (human tutor or no tutoring), regardless of the learning environments (adult training, school environment, etc.). Interestingly, Kulik and Fletcher (2016) highlighted 3 studies in school settings where ITS showed no effect compared to the control group. This finding leads us to insist on the relevance of conducting the proposed scope review.
As stated in Kulik & Fletcher (2016), “It is important to note that none of the field studies in this review completely replaced all classroom instruction with tutoring. Instead, they replaced or partially replaced just one activity with tutoring.” Ultimately, while the teaching profession will be influenced by the rise of AI this movement shall not cause job losses among trainers and teachers. We are convinced that AI should not only be used to support the next generation of learners, but that this technology can also equip instructors with tools that will empower their ability to teach.

Keywords:

Education, AI, Classroom, Intelligent tutoring systems, AI performance, Learning, K12.