Teachers replaced- Can AI give students feedback?

I work in a Research project that tests AI in schools. It could give students the opportunity to have fair assessment, but there are some obstacles…

Assessment is a very important tool for teachers to measure the performance of their students, their progress in the course, to give them feedback on that and motivate them to perform on higher levels. Especially after Covid and the increased usage of Online Learning resources, assessment became quite important to make sure every student reaches the standards that are required of people their age. But naturally, Online teaching made it very hard for educators to give substantial feedback to their students. In a paper from November this year, Indian Researchers claim:

“Currently, with the online learning environment, automated feedback is becoming an emerging focus area of research. Most effective and frequently observed feedback practices are teacher feedback, self-feedback, peer feedback, automated/ computer-based feedback, and feedback modes are oral, written, e-mail, audio, video, rubrics, and web-based” (Read the paper here).

However, even the highest quality feedback is wasted, if the students do not implement it to enhance their school performance. To reach the students, we have to grant them Independence and personalized learning methods in order to motivate them. A high number of students, however, uses Online-platforms like YouTube or they read Blog Posts, and use search engines to study. As mentioned, this makes it harder for teachers to observe and evaluate the learning process, especially if you decide to not give homework to your students every week.

Also, it is undeniable that humans have biases towards other members of our species. If the teacher likes you or not, influences your assessment. If your teacher has biases against black students, you won’t be evaluated fairly as a black student. So what can we do to improve that?

Now, I currently work and study as a Research Assistant at the Christian-Albrechts University in Germany. We test an AI system that gives students feedback on their English Essays. But how exactly shall this machine learn to give efficient feedback to students?

Researchers use Machine Learning technology for such projects. Machine Learning is a Data analysis tool, that uses Database systems to learn an ability that humans and animals naturally have: Experience. They shall find natural patterns in a database, and learn from their own algorithms, to give better prognosis and assessments for instance. On our project, we go from school to school, collect the Essays from students and feed our algorithm with Data. The more samples we have in our database, the more advanced and efficient the algorithms become adaptively. Now my job is to implement the database where we store our samples, from which our Machine Learning algorithm can improve its knowledge, find patterns in feedback techniques, and learn from its own algorithms to develop an efficient assessment method. Good software to implement databases are MySQl or Apache Cassandra. You can use the potential of Machine Learning to make better decisions through Data with programs like Matlab, for instance, or Python.

There are different methods for Machine Learning. We differentiate between supervised and unsupervised Learning. A supervised Learning method, develops a predictive model based on Input and Output Data, so it analyzes both what comes in and what comes out in Data. Unsupervised Learning shall find hidden patterns and inner structures in Data. It groups and interprets Data based on only the Input Data. To find new patterns, you could use an explorative unsupervised method like Clustering to find out what exactly makes an Essay good. Right now, that is partly a subjective definition. By analyzing several assessments of teachers, we could find patterns that Essays with positive feedback all have.

Exciting, isn’t it?

What exactly we do in our project, I can’t tell you yet. We are still at the beginning anyway and I am not the first person to receive news about groundbreaking results, if there is one. However, the future might hold some exciting advancements prepared.

At last, I want to stress that with all that we do not seek to replace teachers, but help them and the students to improve the learning experience in schools and motivate the students. Yet there can be some obstacles we will have to tackle along the way. I don’t know about you guys, but I am always happy and flattered when my Professor gives me positive feedback. I constantly try to impress him and make him proud. Obviously, a machine couldn’t cause the same motivation for me. Sometimes the opinion of humans is more important to us than receiving a fair assessment from an AI. I study the subjects I study because I had a teacher who motivated me to do that.

We will have to find a degree between using AI to improve assessment in schools and yet don’t shift the focus from a healthy teacher-student relationship. Naturally, if an AI gives an assessment instead of teachers, that might undermine their authority in front of the students.

So what do you think? AI in schools- necessary or dangerous?

References and sources:

https://mathworks.com/discovery/machine-learning.html