AI is being used in the US to automate grading and feedback, as well as to create a dashboard that allows teachers to keep a finger on the pulse of the class to know exactly how they are doing Picture: John Amis, AP.
AI is being used in the US to automate grading and feedback, as well as to create a dashboard that allows teachers to keep a finger on the pulse of the class to know exactly how they are doing Picture: John Amis, AP.

Tech news: Artificial intelligence teachers making their way into the classroom

By Louis Fourie Time of article published Aug 3, 2021

Share this article:

UNFORTUNATELY, not all teachers are equally good at what they do, well-trained or committed to their work. This is perhaps one of the major contributing factors to the poor performance of some schools in South Africa.

On the other hand, in the high-performing schools the teachers often appear to be overworked due to numerous meetings, class planning and preparation, endless grading of papers, giving meaningful feedback, administrative functions, and extramural activities.

Artificial intelligence assistance

The schools in the quaint town of Sewell in New Jersey, a north-eastern state in the US, also experienced the problem of overburdened teachers.

The town with its 37 000 inhabitants decided to test an artificial intelligence (AI) driven learning system that automatically grades handwritten tests, analyses results, tracks students’ progress and delivers new content and lectures based on the individual needs of the learners.

The Bakpax system is a computer vision system that uses AI to assess photos of the children’s completed handwritten assignments and tests, convert the handwriting to text, interpret what the learner meant to say, auto-grade the work, and deliver corrections and suggestions within a few seconds. Due to a machine-learning capability, the system’s auto-grader teaches itself how to score and thus improves itself over time.

Improved learner performance

What is impressive is that the AI system assisted in improving the results of the learners and simultaneously freed up the teachers’ time, by grading assignments and typing maths equations on the worksheets through clicking and dragging – a tedious and time-consuming task. The learners further indicated that the learning (for example maths) was considerably easier and they enjoyed the learning more.

The use of AI in our daily life is nothing new. We have become used to Alexa, Google, and Siri assisting us with information in our homes. Although voice recognition and execution is pretty awesome, AI becomes really powerful when it enables machines to plan and make decisions such as the systems used by medical professionals to assist them in making diagnoses or a crime system predicting the next crime area. We are collecting so much data about locations, habits, behaviour, browsing histories and many more, that AI becomes very valuable in detecting patterns, making decisions based on the data, and instructing systems to adjust accordingly.

This is what the educational grading tool Bakpax does due to its deep roots in adaptive machine learning technology. It analyses the data of learners’ progress and personalises the learning experience, speed of learning and content.

The power of machine learning

For many years, researchers have tried to reimagine learning through AI, but it was only since the machine-learning revolution of the past decade that real progress has been made. Increasingly, algorithms were making their way into classrooms to take over many repetitive tasks.

However, it was the AI capability to scrape data off paper that made a major difference in the educational environment. Bakpax uses this capability to automate the grading and feedback, as well as to create a dashboard that allows the teacher to keep a finger on the pulse of the class to know exactly how they are doing.

One of the benefits, say the developers and the schools using the app, is that the app helped to mitigate the digital exclusion experienced by many learners in underserved areas. All that was needed for learners to do their school work was a phone. Since they could merely write their work in normal handwriting, take a picture with their cellphone, and submit it to the AI system, no expensive computer or tablet was needed.

Robot teachers

AI is starting to take over more and more tasks in the classroom, like grading, the optimising of coursework, the preparation of exams, and tracking of learners’ progress. Researchers believe that AI will soon incorporate more humanlike interfaces to enable learners students to communicate with the system as they normally would with a teacher.

It is an established fact that the best education is delivered one-to-one by a knowledgeable educator. But this is too labour intensive and expensive and not practical in countries like South Africa where large numbers of learners need to be educated. This is where AI can assist – often at a minimal cost.

Computer tutoring is not new and has been with us since the 1960s. Over the years many approaches have been tried such as rule-based AI, cognitive theory, and decision trees, but it was the machine-learning revolution that changed education.

Learning algorithms detect patterns in the data about the performance of learners with regard to specific learning material in the past and optimise teaching strategies accordingly. They adapt to the individual learner’s performance based on the interaction of the learner with the system. The result is that learners perform significantly better than conventional classes and instruction from human tutors. Researchers say the learners are doing better because the computer is more patient and insightful than teachers. The learners also love the immediacy of the system when giving feedback.

In the case of Bakpax, data is collected over time, which allows the teachers to determine problems not only on an individual level, but also on a class level, which allows them to compare the performance of various classes with one another.

Machine-learning solutions are increasingly becoming popular. Riiid, a South Korean company, says they are using reinforcement learning algorithms to increase the scores of learners in standardised entrance tests by 20% or more with just 20 hours of study. Acuitus, a US company, created a personal digital tutor that focuses on the teaching of concepts and understanding. Acuitus says it can therefore train experts in months rather than years.

End of the road for teachers?

Inevitably the question arises if AI will eventually replace teachers in the schooling system. Despite the amazing progress made by AI over the past few years and even the improved results of learners, it will probably not replace humans but rather assist us to do more. We will still need schools, classrooms and teachers to form and motivate learners and to teach social skills, teamwork and soft subjects like art, music, and sports – at least for now.

The problem is that in some cases parents have discovered during the Covid-19 pandemic and home schooling that they could often get better educational lessons for their children through the internet than they were getting at school. It is thus possible that as algorithms improve and become more complex, machine learning become more powerful, and digital tutors and even robots become more common in our daily lives, that some underperforming teachers will become obsolete. The Fourth Industrial Revolution requires new skills and a new adaptive curriculum.

Prof Louis C H Fourie is a technology strategist.

*The views expressed here are not necessarily those of IOL or of title sites.


Share this article: