The World Economic Forum recently estimated that more than 1.2 billion children in 186 countries have been forced out of their classrooms by the COVID-19 pandemic. Education has changed as a result, with demand for remote learning systems accelerating the rise of e-learning digital platforms.
Whether the adoption of online learning will continue post-pandemic remains unknown, but even a slight shift could heavily impact the huge global education market. Lower-cost online educational approaches such as massive open online courses (MOOCs) have flourished over the past decade and now boast many millions of users.
In a domain as sensitive as education, quality is essential. Assessing the performance of online learning systems however is difficult due to low levels of active learning, little feedback from instructors, and fewer peer discussions. Because different students have different interests, needs and abilities, Canadian education technology startup Korbit Technologies has introduced a personalized AI-powered learning experience that it says can help all students learn faster and better in a cost-effective way.
A spin-off from the Mila lab (Quebec’s Artificial Intelligence Institute) led by Turing Award honouree Yoshua Bengio, Korbit announced a seed round of over CA$2.65 million last October. The company says over 7,000 students have enrolled since the launch of its first course on machine learning taught by Bengio and other Mila professors last year.
“Korbit’s mission to democratize education through AI will transform the education world, and I’m excited to help them achieve this monumental goal,” Bengio said in a press release last year.
Korbit CEO and co-founder Iulian Vlad Serban completed his PhD in AI and personal assistants technology at Mila and has since been applying his expertise to the construction of an AI tutor. Serban, Bengio, and a team of researchers from Korbit Technologies, Mila, University of Cambridge, and École de technologie supérieure in Montreal published a paper earlier this month detailing the company’s large-scale, open-domain, mixed-interface, dialogue-based intelligent tutoring system (ITS) — also named Korbit.
Korbit uses machine learning, natural language processing, and reinforcement learning to provide interactive, personalized online learning. Korbit is currently capable of teaching topics related to data science, machine learning, and artificial intelligence, although the researchers say it can easily scale to thousands of subjects by automating, standardizing, and simplifying the content creation process.
ITS are AI-powered computer programs that deliver real-time and personalized tutoring to students. Traditional ITS implement or imitate the behaviour and pedagogy of human tutors for example in dialogue-based ITS systems using natural language conversations to tutor students.
A 2018 study by University of Memphis professor Andrew Olney however found that ITS are extremely expensive to produce — with some groups estimating that it takes 100 hours of authoring time from AI experts, pedagogical experts, and domain experts to produce one hour of instruction.
Korbit differs from other ITS in that a teacher can develop new learning modules in only hours, and its state-of-the-art cloud-based microservice architecture is built to scale to millions of students. The company says the highly modular platform is designed to enable easy content creation and will soon be expanded with many more topics.
To facilitate learning across a wide range of STEM subjects, Korbit uses a mixed-interface which includes videos, interactive dialogue-based exercises, question-answering, conceptual diagrams, mathematical exercises, and gamification elements.
The researchers conducted multiple studies — between October 7 and December 22 last year, in an A/B testing setup with 612 participants — to evaluate the Korbit ITS. Testing compared Korbit against the xMOOC ITS with regard to average time spent by students, returning students, students who said they would refer others, and learning gain.
In the evaluations both student learning outcomes and student motivation were substantially higher with Korbit ITS compared to typical online courses.
Student responses in the Korbit feedback questionnaire suggested improving solution correctness identification and explanation accuracy and providing more relevant and personalized feedback in the case of incorrect solutions.
The paper A Large-Scale, Open-Domain, Mixed-Interface Dialogue-Based ITS for STEM is on arXiv.
Journalist: Yuan Yuan | Editor: Michael Sarazen