Can AI feedback help us learn better?

Making mistakes and learning from feedback are part of life. Feedback is an essential part of the learning process, also at school: It can show students where they are in their learning process and what they can change in order to achieve their learning objectives. For example, it can draw their attention to a mistake in a mathematical formula or to the fact that the central thesis in their essay is not yet clearly defined. Well-crafted feedback can be very effective. Studies have shown that some learners make more rapid progress when they receive feedback on their answers than by simply watching an explainer video. This is because feedback supports active learning. In this sense, then, the issue at hand is not about assessing students' performance with AI but about promoting their learning process.

AI feedback can support students in learning

Even when everything is changing – from reading and writing to maths and solving problems, basic skills remain essential or are becoming even more important. Therefore, AI should not be seen as a replacement but an opportunity – a tool with which we can improve teaching and learning. We only have to use it right. Especially regarding more complex tasks, such as writing longer texts in German or English, teachers are faced with time constraints or high workloads. AI offers great potential here. By providing an opportunity for more feedback, it can take pressure off teachers and offer students more opportunities to learn from mistakes.

Even in situations where students are supposed to rely on themselves to do all the work, such as when doing their homework, AI can also make a difference. AI feedback can help students put in more work (such as by revising a text one more time) and also when they feel stuck, thereby providing assistance not only with homework, but also for tasks that may arise at the university or the workplace. 

AI feedback can also help students reach a better understanding of each step of their task when they encounter difficulties. Initial studies have shown that AI-generated feedback can help with the revision of argumentative essays in English classes, for instance. A growing number of evidence also indicates that, with regards to quality, AI feedback is in many cases barely distinguishable from feedback provided by a teacher.

Active learning remains key

Of course, AI must not replace the writing process. It is important that students think actively and do not outsource the learning process to AI. This also matters when dealing with feedback. For example, what happens when learners are confronted with mistakes in the AI-generated feedback? Initial study results indicate that learning is still possible, as long as students notice mistakes in AI feedback by actively engaging with its content. However, there is reason to assume that motivation for learning declines the more mistakes the feedback contains. As with most research findings, these initial results need to be taken with a pinch of salt for now. Further research is needed to shed more light on the challenges of erroneous AI feedback in different subjects and learning settings.

Are we putting too much or too little trust in AI feedback?

Some studies show that AI feedback is often underestimated. When feedback is labelled as 'generated by AI', many learners as well as teachers assess it more negatively. Some students even change their mind about the feedback in retrospect, as soon as they find out it was not provided by a human but generated by AI. These findings indicate a lack of trust in AI. 

On the other hand, AI feedback lacks several social aspects that may have both good and bad consequences. For example, when learners are afraid of receiving a negative assessment from a human instructor, they might be more likely to accept AI feedback. At the same time, AI feedback can potentially be less motivating. More studies are needed because they could contribute to developing AI-assisted learning scenarios, which make use of the benefits and drawbacks of AI feedback in addition to feedback provided by teachers and peers. At the same time, this would give teachers more time for social interaction in the classroom.