AI & ML in education: Real-life use cases

During an 8:00 am Economics lecture in my MBA course, one of my friends sat half asleep in the first row. The professor asked him a question, and even without having any clue about the answer, he just blurted “Profit Maximization”. “Correct”, said the professor, and moved on!

After all, that one term seemed to be the default correct answer for every question in MBA. I could relate to that as an analogy because during this pandemic, we can safely say that today, “Artificial Intelligence and Machine Learning” have similarly emerged as the default solution to most technological problems.

Digitalization in the education sector

Although industry leaders in education have been advocating the move towards digitalization for a while, the COVID-19 pandemic has forced companies to finally give impetus to their digitalization efforts and leverage emerging technologies such as AI and AR/VR. Technavio’s market research had predicted a CAGR of 48% in AI’s market in the US education sector during 2018-2022.^[1] Imagine if someone had told them about COVID-19!

We are already aware of the superpower of AI and ML - even a simple Google search provides clear evidence of the surge in EdTech funding worldwide. But what gaps are being filled by this combination of AI and Education?

There is huge potential in leveraging technology for the bottom of the pyramid and solve real-world problems. The challenge, as of today, lies in developing the infrastructure of developing economies. While the stiff competition and constant innovations have reduced cost, effective implementation requires:

• mass acceptance of digitalization
• trust amongst stakeholders and
• proactive allocation of research budgets by states to understand the gaps

Real-life use cases of AI & ML in education

Let us look at some use cases from across the world where AI and ML are already creating a difference.

Virtual assistance

In 2015, a Computer Science Professor at Georgia Tech, Ashok Goel, was teaching an online course on Artificial Intelligence. At the end of the course, the students got to know that Jill Watson, one of their favorite teaching assistants answering their questions most of the time, was actually a bot! The team had used all the chats generated over a year to feed an AI program that was able to respond to FAQs with 97% accuracy.^[2]

One of the key problems that faculties face in digital teaching is to provide live feedback/support to student queries. By leveraging AI and ML, chatbots can act as virtual assistants and solve real-time queries. This allows a teacher to dedicate more time from admin tasks to actual lesson planning. At Nagarro, we have Hermoine, our Edu Bot, which connects to the data lake and LMS (Learning Management System) to provide solutions to student queries about their next assignment, their mistakes in the last quiz, faculty feedback, and much more.

Digital exams and assessment

The lockdown impacted examinations, and in most cases, schools had to either cancel or postpone the exams. Within some time, institutes realized that online learning would be the new normal, and they must adapt to move ahead. Three solutions were commonly used - graduating everyone, providing online assignments, and extrapolating students’ marks from their previous exams. Evaluation, especially for high stake exams that require access to resources and strict invigilation, continues to remain a problem for institutions to conduct online.

Artificial intelligence provides a fair response to this issue. Solutions looking to engage and evaluate online exam environments will likely gain ground soon through retinal tracking, environment stimulus tracking, and IP tracking. The data generated through such digital examinations combined with the power of machine learning will auto-generate evaluation papers as well as a course of action for each student to help teachers focus on the facilitation part.

Companies like Mercer Mettl and Littlemore Innovation are providing digital exam solutions using AI and ML. “Our AI has been trained with more than 2.8 million proctored assessments that can detect over 18 dynamic digressions with 95% accuracy. Our platform has conducted more than 20 million online tests over the years, and the numbers continue to grow. The robust technology based on the power of AL and ML has a proven track record of delivering results at all levels. We have recently been successful in conducting over one lakh proctored assessments in a day”, says Mr. Siddhartha Gupta, CEO, Mercer | Mettl.

Special education

The wide spectrum of autism always had less research in learning theories than needed. The current shift increases the pressure on a facilitator to overcome barriers and provide individual attention to kids with special needs. While the lockdown has created mental health issues, they are amplified even more in this segment because of the rapid transition to the digital medium.

Artificial intelligence and machine learning can be of great help in this situation. AI not only provides highly customized learning for special children, it also learns from the responses to work on specific areas and improves them while considering the student’s learning speed and time.^[3] This will help schools to compensate for the extra effort needed in the new normal by using it as a tool for teachers, in physical and digital classrooms, to make data-driven decisions in responding to kids with special needs.

Otsimo, an award-winning application, boasts of multiple success stories in this field. According to Zafer Elcik, the founder of Otsimo, “We use AI for our speech therapy application. It is used to align the filters that come after the child has completed an exercise, and the filter comes up as part of the reward mechanism. Machine learning is used for voice recognition. That is how the app truly understands what the child is saying and then fills up the progress bar accordingly.”

Customized learning

Tailormade courses are the new reality. AI and ML empower education to be individually-driven. The programs can predict the student’s mood, progress, and learning to alter the module and assignments. A lot of ed-tech companies have sprung up in this space over the last few years. Be it K12, upskilling, coding, placement prep, or early childhood education – players are making the most of the COVID situation at all the levels.

Companies like Embibe, Interview Bit, Knewton, Duolingo are using the power of AI to customize the learning path for all the students. As shared by Abhimanyu Saxena, co-founder, InterviewBit & Scaler Academy, “We have been extensively using AI-based proctoring tools in our to ensure the authenticity of solutions submitted by applicants. For instance, in the Scaler Academy entrance test, the entire length of the test is tracked by proctoring tools that record everything. After the exam, all the submitted solutions are run through the tools that check plagiarism, the text, and also the semantic flow of solutions. We strongly believe that AI proctoring tools ensure a much more reliable and accurate examination process. Some of the companies that we work with are also using AI proctoring tools to conduct assessments on potential employees as part of their recruitment process.”

Enrollment and admission process

Imagine going to a school and getting suggestions for course curriculum based on your goals and student enrollment records data to help create a customized path for you! At Berkley, assistant professor Dr. Zachary Pardos is developing the same model for the students by using machine learning. This model will not only help students carve their path but will also focus on making suggestions to develop specific skills using artificial intelligence.^[4]

Bank loans

In 2019, the student loan debt in the US alone stood at $1.6 trillion. ^[5] To give an idea of how big that number is: the combined GDP of Poland, Sweden, and Belgium^[6] is less than $1.6 trillion! Artificial intelligence models can help banks and governments in predicting defaults on education loans.^[7] But that’s a problem solving for banks, you’d think. How does it help the education sector? “The program can suggest a suitable solution for students to avoid loan default,” confirms Prof. Jayadev M from Indian Institute of Management, Bangalore. Shaping Edu is a project at Arizona State University that is trying to achieve just that.

Use cases in Nagarro

At Nagarro, we have our own suite of AI ML tools for the education domain.

This includes:

• an AI-based scheduling tool that helps institutes improve their timetables and save cost
• an AI-based trailer generator which helps in auto-video based content creation
• a class-mood tracker, and many more in the pipeline

Some of them are a part of our Blended Learning solution.

While there is still a lot of potential in exploring AI and ML’s usage, especially in solving the bottom-of-the-pyramid problems and in providing universal access to education, it is still safe to say that the pandemic has provided a boost to the efforts in moving towards the right direction. It will be interesting to see how the adaptability rate increases in the new normal.

Are you interested in exploring our nonprofits & education offerings? Click here to know more about us or reach out to us to explore how you can benefit from our services.


References:
[1] https://www.technavio.com/report/artificial-intelligence-market-in-the-us-education-sector-analysis-share-2018?tnplus
[2] http://dilab.gatech.edu/a-suite-of-online-learning-tools/
[3] https://www.forbes.com/sites/ilkerkoksal/2018/04/17/ai-has-already-started-reshaping-the-special-education/#530adcf229d5
[4] https://dl.acm.org/doi/10.1145/3298689.3347030
[5] https://fred.stlouisfed.org/series/SLOAS#0
[6] https://www.worldometers.info/gdp/gdp-by-country/
[7] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3510012