How a tutoring app is using Spotify-inspired tech for education

Image credit: Pexels

Music lovers are probably familiar with Spotify’s song recommendation engine. The app analyzes users’ behavior and suggests more songs based on their preferences. Spotify’s popular Discover Weekly function runs on this algorithm, and Spotify Radio uses the same formula.

Netflix and YouTube use the same technology for their apps, relying on similar algorithms to offer content that users might like, based on previous behavior.

But what if this were applied to a tuition chatbot? Opet Foundation’s AI-equipped companion chatbot does exactly that, with the help of a hybrid filtering recommendation engine. Scheduled to launch in October 2018, the chatbot is designed to help students learn by putting forward questions and study material based on users’ aptitude for different subjects.

How recommendation engines work

Generally, there are three types of recommendation engines: collaborative filtering, content-based filtering, and a hybrid of the two systems.

In collaborative filtering, the AI learns what users enjoy based on their behavior. An example is Amazon’s “people also bought” feature, which is drawn from other customers’ purchases. Their preferences are analyzed, and more material will be recommended based on what they have already picked out. The machine doesn’t understand the content – it simply makes suggestions based on what other people also like.

Image credit: Pexels

In content-based filtering, items are given scores depending on certain qualities. For instance, music-streaming service Pandora uses characteristics of a song or an artist to recommend similar tunes. Attributes as specific as “aggressive drumming” and “ambiguous soundscapes” are given values, which then become the basis for leading users to other songs with similarly valued attributes.

However, companies like Spotify and Netflix use a hybrid system that combines both filtering methods, providing more accurate recommendations by balancing both user behavior and attribute preference.

Adapting it for education

Opet’s tuition chatbot app builds upon Spotify’s existing application program interface, but the way it works is opposite to the digital music service’s recommendation engine.

Ameya Kulkarni, Opet’s chief technology officer, explains that if the AI recognizes that a user is good at a particular subject, such as trigonometry, for example, it “won’t nudge you to solve more math questions relating to trigonometry.” Instead, it will encourage the user to spend more time on other subjects that are accessed less often, as it assumes that the student needs to improve in those topics.

This approach allows students to learn in a more holistic manner, instead of simply focusing on material that they’re already strong in. The Opet chatbot app will also periodically remind students to complete assignments and notes, in case they forget.

“We make sure users know their weaknesses and allow them to work on [those areas],” says Kulkarni.

Founder and CEO Wilson Wang summarizes Opet’s core value proposition. “In essence, we are using the Duolingo method to teach high school curriculum.” The Duolingo technique relies on its data to boost learners’ metrics and restructure their learning experience. For the Opet chatbot, users can rate questions by clicking on a thumbs up or thumbs down button. Their feedback affects the likelihood of questions being recommended to other users.

Students as data sources

Photo credit: Pixabay

Kulkarni has other plans for Opet’s recommendation engine to propose university courses that students might do well in.

Data from user activity such as the number of quizzes solved, average response time to a question, and overall learning curve are collected and saved into a “feature matrix.” In turn, the matrix can be sent to universities as proof of students’ capabilities, possibly creating a new way for them to study in institutions, particularly those that might be out of reach for the less affluent. This could help students who don’t have access to such educational advantages, according to Kulkarni.

The app can also give students who have a clear goal an extra push. If they want to pursue a Bachelor of Business Administration (BBA), for example, the Opet app can assess the requirements for the course, such as adeptness in math and economics. The chatbot can then recommend that they “do questions and quizzes to help acquire a BBA,” shares Kulkarni.

The same AI can also be tapped to endorse courses that users may be suitable for but had never considered. These recommendations will pop up even if students have already stated a preference for another course of study.

Changing the education sector

While recommendation engines have been mostly used to seek and gather people’s opinions, AI and machine learning has enabled thoughtful and helpful suggestions to appear on our screens. But the use of this technology has been largely limited to consumers and apps for entertainment.

Opet believes that using such tech for education can do a lot of good, especially in improving users’ learning capabilities and bridging the class divide that can limit academic opportunities for disadvantaged students. This could completely change the way the education system works.

Find out more about Opet and how it merges education with blockchain here