Adaptive Education: Can AI Empower the Online Education Industry?
-
Adaptive learning systems are AI-based personalized teaching systems. This article objectively analyzes whether AI-powered online education is a gimmick or a genuine breakthrough from the perspective of learning processes.
The recent outbreak of the novel coronavirus has made online education a 'necessity.' Where there is demand, there is a market. At this critical juncture, many internet companies engaged in online education have stepped up to help students overcome the challenges.
On the afternoon of January 27, Alibaba Group's Youku and DingTalk jointly launched the 'Study at Home' initiative, allowing primary and secondary school students across the country to attend free online classes via Youku and DingTalk apps.
Tencent's education platform introduced 'Tencent Classroom' to help educational institutions quickly set up online live and recorded teaching solutions.
TAL Education Group, the parent company of Xueersi, announced a donation of 20 million yuan to the Wuhan Charity Federation while also offering free synchronized live classes for students in Hubei and other affected areas who cannot attend school.
Almost overnight, online education companies took center stage, seizing the opportunity.
According to iResearch's '2019Q3 China Online Education Industry Data Report,' China's online education market was projected to reach 313.36 billion yuan in 2019, a 24.5% year-on-year increase. The market is expected to grow at a steady rate of 18-21% over the next three years, with a gradual slowdown in growth momentum.
In China's education sector, 'AI empowerment' is no longer a foreign concept. Platforms like Zuoyebang, Yuanfudao, and Xueba Jun use speech or image recognition technologies to assist students and parents. New Oriental has launched an intelligent oral and writing platform, while Hujiang introduced the 'Uni Smart Learning System,' promoting personalized learning tailored to individual needs.
In China, planning and exploration of AI empowerment in education began in 2016.
According to a report by Zhiyan Consulting, '2020-2026 China AI Education Industry Market Operation and Investment Opportunity Analysis,' AI education financing has grown rapidly since 2016. In 2016 and 2017, AI education accounted for 4% and 7% of all education investment cases, respectively.
By 2018, this figure surged to 19%, and in Q1 2019, it had already reached half of 2018's total, with expectations to surpass it by year-end.
Adaptive learning systems are AI-based personalized teaching systems. In the West, the most notable example is Knewton's adaptive education, which has been in the field for over a decade and gained widespread recognition in the last three years. Over 3,000 schools and universities in Europe and the U.S. use adaptive education technologies.
In China, since 2017, much of the education industry's funding has been chasing 'AI + education.'
Squirrel AI (formerly YiXue Education) raised 270 million yuan in its angel round as early as 2017. Additionally, in November 2017, SmartStudy secured 200 million yuan in Series B+ funding, with upgrading its intelligent teaching system as a primary use of the funds.
On October 30, Jike Big Data received 100 million yuan in Series B funding. Founded in 2014, Jike Big Data focuses on EI (Educational Intelligence) systems using image recognition and natural language processing to assist schools in managing student assignments and exams.
Today, the 'AI + education' landscape covers the entire industry chain—'learning, practicing, correcting, testing, and evaluating'—from language oral exams and automated grading to adaptive learning, virtual learning assistants, and expert systems.
Whether it's Squirrel AI's 'Intelligent Adaptive Learning System' or Hujiang's 'Uni Smart Learning System,' both advocate personalized learning tailored to students' needs to improve efficiency.
However, concepts and slogans are cold. After more than two years of 'AI + education,' beyond the noise of capital and the restlessness of entrepreneurs, the promised revolution has yet to arrive. Concerns and skepticism persist.
Some argue that 'AI + education overhypes AI's capabilities, focusing only on knowledge delivery while deviating from the essence of education.' Others claim that 'real-time interaction with AI, through voice or text, effectively breaks the boundaries of learning time.'
Amid the debate, this article objectively analyzes whether AI-powered online education is a gimmick or a genuine breakthrough from the perspective of learning processes.
Yu Minhong once stated publicly:
'Two key elements in education are: first, improving student learning efficiency, ultimately leading to better grades; second, enhancing teaching quality, resulting in teachers being liked by students and further boosting their performance. Beyond improving learning efficiency, resource connectivity, sharing, and distribution are equally important.'
Breaking down the learning process, current products cover peripheral aspects like 'photo-based problem-solving' (e.g., Zuoyebang, Yuanfudao) and 'oral assessments' (e.g., Liulishuo, Boxfish).
These tools leverage advanced AI but are limited to auxiliary learning processes, not directly enhancing teaching quality or outcomes.
Adaptive AI learning, however, targets secondary and core learning processes, fundamentally improving learning concepts and methods.
So, how do these intelligent education systems work?
An analysis of Squirrel AI's (China's first AI education company) adaptive learning system reveals its core components: data, algorithmic models, and teaching research teams.
Li Qilin, co-founder and CTO of Afanti, once said, 'Data is a significant barrier.'
Currently, educational institutions source data from traditional teaching behaviors and online student interactions.
For example, Squirrel AI combines user state assessment and knowledge recommendation engines to build data models, accurately identifying students' knowledge gaps and recommending tailored content. As of May 2019, Squirrel AI had 2 million registered users, with potential for further growth in China.
Compared to the West, China's adaptive education started later but has clear advantages.
Professor Tom Mitchell (now part of Squirrel AI's team and Dean of Carnegie Mellon's School of Computer Science) noted:
'In the future, those with larger and higher-quality data will likely prevail.'
Undoubtedly, AI is a powerful tool, but in 'AI + education,' only those who truly understand education can create products aligned with its principles. For instance, Yuanfudao's founding team came from NetEase, and Wang Jianhong, founder and CEO of Aixue Tang, was previously principal of Huanggang Middle School's online division before Tsinghua University became its largest institutional shareholder.
In building knowledge bases, Squirrel AI's teaching team breaks down subjects into ultra-nano-level knowledge points.
On overcoming cross-provincial educational barriers, Li Haoyang stated:
'By dissecting knowledge points finely enough, we can largely overcome these obstacles. Despite varying textbooks, the core knowledge points are similar, reducing differences at this level.'
During the pandemic, leading online education platforms launched free public courses. Xueersi, Yuanfudao, and Zuoyebang reported impressive results, with Yuanfudao claiming 5 million simultaneous live viewers and Zuoyebang surpassing 10 million sign-ups for its free courses.
Although society as a whole is doing its best to engage in public welfare during the pandemic, traffic acquisition remains an unchanging theme.
Liu Baoshu, founder of Aoyi Education, stated:
"The DAU growth rate of major platforms has reached 100 or even 1000 times. Even a small platform I'm familiar with, without any advertising, has achieved a 10-fold increase in DAU. The sudden surge in traffic has caused their servers to crash across the country."
Currently, online education has broken spatial limitations to gain a large number of new users. However, after the pandemic, the existing shortcomings in the online education industry will not disappear.
Therefore, how to effectively operate the current influx of new traffic, ensure course quality and service, and actively maintain engagement with students and parents will be key to retaining users after the pandemic.
For AI+education, users are the absolute source of data, making this an opportune time to actively manage the new traffic.
On the evening of February 15th, Li Haoyang, founder of Squirrel AI (formerly YiXue Education), posted on social media: "Due to the pandemic, online student numbers have surged, but headquarters is offering free services, which has actually increased costs."
To cut costs, he "made the tough decision to reduce all salaries by 65% for five months, with core executives taking zero salary." He claimed this move was to "ensure we have sufficient funds for major battles when the sun rises again, which is the greatest care for our employees."
As an upgraded form of online education, adaptive education faces challenges in training models with increased user behavior data, continuously optimizing nanolevel knowledge points, and properly managing relationships with government, schools, and teachers - all of which will determine survival after the current boom period.