‘Smart glasses’ that give teachers detailed real-time information on how each student is learning can help pupils do better in class and help ensure lower achievers don’t get left behind, according to a ground-breaking experiment by US academics.
The research, published at the London Festival of Learning, is the first experimental study to show measurable benefits of a complementary mix of human and machine intelligence to support classroom learning in real-time.
Researchers from Carnegie Mellon University’s Human-Computer Interaction Institute in Pittsburgh, Pennsylvania, equipped teachers with the futuristic viewing aids to enable them to instantly see how each student in their class was progressing while working with a computerised mathematics tutor.
The ‘mixed reality’ transparent glasses, called Lumilo, allowed teachers to check up on individual pupils’ and whole class performance during lessons using the lenses as a private data display, while simultaneously keeping an eye on the classroom as normal to pick up on students’ body language for signs of confusion or frustration.
By programming in class seating plans, teachers could see a personalised display pop up on their glasses above each student’s head, with the option to make a tap gesture in mid-air to see a detailed ‘deep dive’ set of individual data. They were alerted by the glasses when pupils were struggling with particular skills (and especially when they appeared to be truly ‘stuck’), when they seemed to be avoiding asking for help despite prolonged struggle, when they were attempting to abuse the programme’s design or ‘game the system’ or had simply stopped working for over two minutes. Teachers were then able to provide one-to-one support as needed.
The approach combined the benefits of sophisticated computerised learning technology (known as intelligent tutoring systems, ITSs), which provides students with step-by-step guidance and allow them to work at their own pace, with complementary targeted support from human teachers. While research has shown that ITSs are highly effective for learning overall, this approach enabled teachers to step in where artificial intelligence (AI) was less effective, for example in providing social and emotional support, supporting student motivation, or coming up with fresh ways to explain a concept.
By alerting teachers in real-time to situations the ITS may be ill-suited to handle on its own, Lumilo facilitates a form of mutual support or co-orchestration between the human teacher and the AI tutor. This research illustrates the promise of artificial intelligence systems that integrate human and machine intelligence to support student learning.
Said Ken Holstein, lead author on the study together with Bruce M. McLaren and Vincent Aleven.
The research focused on 286 middle school students aged 10 to 13 and eight experienced teachers in 18 classrooms drawn from four public schools in separate school districts in a large US city and its surrounding areas. The pupils were using Lynnette, an intelligent tutoring system for linear equations which has been shown in studies to significantly improve users’ equation-solving ability, and the teachers helped design the Lumilo display.
As well as assessing the results when teacher and AI tutor worked together using the smart glasses, the experiment also examined an ITS classroom in which teachers had no glasses and no access to real-time analytics, together with a third variant in which teachers wore smart glasses but only had access to basic monitoring support (allowing the teacher to see what was on individual students’ screens from anywhere in the classroom, without any additional analytics).
Tests taken by pupils before and after the lessons revealed they learned more from working with the intelligent tutoring system when teachers were using Lumilo, with the best results when teachers had access to the most detailed analytics and not simply monitoring support alone.
Whereas prior research has shown that classroom use of AI tutors typically widens the gap between the most and least prepared pupils (often dubbed the ‘rich get richer’ effect), the researchers concluded that ‘the real-time analytics provided by Lumilo appear to have served as an equalising force in the classroom: driving teachers’ time towards students of lower prior ability and narrowing the gap in learning outcomes between students with higher and lower prior domain knowledge, at no cost to the latter group’.
Even where teachers had only basic data available through the glasses, students were less likely to try to abuse or ‘game’ the maths programme, suggesting that even the knowledge that a teacher is monitoring their activity has a motivational effect, the study found.
The researchers now hope to build on the work, examining the longer-term effects of real-time teacher analytics, investigating how such systems may best be designed for teachers with varying levels of experience, and further exploring how such ‘teacher-in-the-loop’ AI tools might serve to benefit otherwise underserved groups of students.
Student Learning Benefits of a Mixed-reality Teacher Awareness Tool in AI-enhanced Classrooms, by Kenneth Holstein, Bruce M. McLaren and Vincent Aleven (Carnegie Mellon University, Pittsburgh), was presented at the International Conference on Artificial Intelligence in Education as part of the 2018 London Festival of Learning.