Generating Feedback for Multiple Choice Questions with the Help of AI

Sammanfattning: Quality education is one of the United Nations Sustainable Development Goals. In pursuit of reaching this goal there are multiple available techniques today, where one of them is Artificial Intelligence, AI. How it can be used is not established but a potential area of use is explored in this thesis. It is investigated whether an AI, today available to the public, can be used to generate feedback for human produced multiple choice questions. It also reflects on to what extent the feedback produced is sufficient to be helpful to a student when learning. This area of study is of great importance due to feedback being a great contributor to learners' understanding. However, feedback needs to be available when a learner needs it, and it has to be specific for the learner. This is a big challenge and hard to achieve since feedback is hard to produce, and it is therefore of great importance to find a viable option. Multiple questions with feedback were provided and were sent into an AI with three different prompts, to produce new improved feedback. Ratings of the questions were also provided. The generated feedback was evaluated in two ways, it was rated according to criterias for good feedback and a survey was sent out. The survey gave outside opinions on what prompts were preferred and why. The results showed that feedback can be produced with the different prompts and that it was better and often preferred over the feedback created by humans. The feedback produced was mostly correct, but not always, making the AI not always trustworthy. The results also showed that the feedback produced was mostly considered sufficient. This shows that AI can be used to enhance learning.