- AI-enabled virtual tutors are ideally placed to support human tutors and teachers through Covid-19.
- Virtual tutoring is a scalable, affordable means of ensuring all learners are equipped with core knowledge and skills.
- Learning data is a by-product of virtual tutoring and serves as a powerful mechanism for real-time measurement of tutoring impact.
Tutoring has a lineage that stretches back thousands of years. Socrates tutored Aristotle and Aristotle, in turn, tutored Alexander the Great. In more recent times, ever since Benjamin Bloom’s 2-sigma study in 1984, the educational impact of one-to-one tutoring has been widely sought after.
Covid-19 has now brought tutoring centre stage onto the educational policy agenda. The Education Endowment Foundation, drawing upon their own more detailed research, are to be congratulated for persuading the UK Government to back a National Tutoring Programme as a means of supporting schools to mitigate the impact of Covid-19 school closures on pupils’ learning. From autumn 2020, the NTP will make high-quality tuition, both one-to-one and small group, available to schools, providing additional support to help pupils who have missed out the most as a result of school closures.
Virtual tutors, which we take to mean automated and AI-driven rather than talking heads on conferencing tools like Zoom, are now able to contribute greatly to the impact expected from human tutors. A virtual tutor emulates some of the core behaviours of human tutoring by providing individual pupil assessment, identifying knowledge gaps and delivering scaffolded content at the appropriate level for each user.
Augment rather than substitute
Virtual tutors can act as a safety net for the NTP, protecting against the inevitably challenging logistics of delivery of one-to-one and small group tuition in a Covid-19 affected context. Subject only to internet access, a virtual tutor is available anytime and anywhere.
The benefits of virtual tutoring extend beyond scalable instruction. Where human tutors are available, they can deploy virtual tutors to gain easy access to real-time insights into how their students are progressing. A virtual tutor augments the role of human instructors; it helps them to plan sessions by proactively identifying student misconceptions and suggesting areas of focus. A virtual tutor is not intended to replace human tutors but rather strengthen their face-to-face interaction with students.
In short, intelligently designed virtual tutors can provide the reliable, basic groundwork to support the aims of the NTP, and supply easily actionable data which the human tutors can leverage to ensure superior educational outcomes from the greatest number of students. The same data can be aggregated and analysed at any level to monitor the impact of large-scale tutoring interventions.
The scalability of the 21st century’s innovative virtual tutors is the answer to Professor Bloom’s “too costly for most societies to bear” verdict on one-to-one human tutoring. If the Government delivers on its promise of digital devices, internet connections and support for disadvantaged students, virtual tutoring can be delivered to all at a tiny fraction of the £700 12-week cost of one-to-one human tutoring estimated by the EEF.
A case study
Armed with individual logins to the virtual tutoring platform, Maths-Whizz, Martin offers small-group tuition to 24 students in the south of England. Martin requires the students to each spend 60 minutes per week in the virtual tutor section of Maths-Whizz in their own time. Ahead of each tutorial – usually involving 3 or 4 students – Martin reviews the Maths-Whizz reports of the students, noting the exercises in which they may have struggled and alerting himself to opportunities to enrich their understanding of the learning objectives they have worked on. If any student has encountered a significant roadblock, this will be easily identifiable from the student’s Maths-Whizz report and is quickly addressed by Martin.
The efficiency of the partnership with Maths-Whizz is demonstrated by Martin’s finding that he needs only to devote approximately 15 minutes per week in his tutorial to maths, leaving the remainder of his time with the students for English, Verbal Reasoning and other subjects. Over a recent 12-week period, Martin knows that each of his students spent 12 hours achieving an average 60 progressions in Maths-Whizz (equivalent to about 5 months of progress in the maths curriculum), fully in line with the EEF’s best expectations concerning the impact of small group tuition. Martin has a comprehensive record of the progress made in each maths topic for discussion with the parents and teachers of the students.
Our data reveals that students who use Maths-Whizz for 60 minutes per week increase their Maths Age, on average, by 18 months in their first year of use. To assist students to catch up from the learning losses suffered from Covid-19, Maths-Whizz is the ideal tool to find the particular gaps each student has developed and to help the NTP’s tutors both to fill them and to record their progress.
Arming the army
So many walks of life, from manufacturing to customer service to medical diagnostics, have leveraged intelligent technologies to improve their effectiveness. It is surely time to augment the intelligence of the proposed army of human tutors in the NTP with virtual tutors.
Professor Kurt VanLehn at Arizona State University has devoted his career to the study of intelligent tutoring systems. In his view, “the major limitation of intelligent tutoring systems is that they teach a certain class of cognitive skills such as solving well-defined problems in math…In retrospect, this limitation makes sense. Acquiring a high degree of cognitive skill requires hours and hours of deliberate practice. Such repetitious activity is ideal for computer tutors.”
Virtual tutors will not, and should not, displace humans from any instructional context. But they are poised to play a powerful role in ensuring all students are equipped with core mathematical knowledge and skills during Covid-19, and even beyond.