Abstract:

The public and higher education in Hungary migrated to distance/digital form in mid-March. Higher education was given a week to prepare for this change, whilst public education only received a weekend. A significant proportion of students and teachers were not prepared for this kind of change and were not well equipped to switch to distance learning in such a short amount of time. Not everybody owned a desktop computer/laptop, web-camera, microphone, drawing-tablet or even smartphone. Previously, one computer would be enough for a family, but now children and parents needed to learn/work in parallel. Furthermore, not everyone had (a good) internet access to take part in a video conference (sometimes in parallel with the other members of the family), and several hour long daily video conferences do not fit a daily/monthly mobile data limit.

We tried to organize our Artificial Intelligence (AI) course in the least restrictive way, to overcome the previously mentioned shortcomings and limitations. Like thousands of universities, we work based on the Russel-Norvig AIMA book. Fortunately many classroom lectures can be found on Youtube in this topic. We linked these videos in the corresponding Moodle course for our AI course, so students would not be limited to the official time of our lecture. We have announced consultation dates and redirected email communications to Teams (Microsoft365) which is available all of our students. The questions asked here were often answered outside of consultation hours as well.

The exercises in this course are typically centered around algorithms. It’s worth demonstrating these algorithms in the process, so for these a text description or an audio file would not have been enough. Therefore, we presented these algorithms with animations and explained the solutions of the exercises. The students were very happy with these videos because they were able to watch the steps of the method again and again. Of course to see a video is not enough in order to learn it in practice, the students would learn by practicing these algorithms. In the last years we have created several online tests, and now we took the opportunity to expand these with some new question types and updated others to increase the variety of answers. The students were able to practice without restrictions, during consultations we could discuss any problematic cases, and during the exam they were expected to solve similar tasks.

During the exam, the same test was given as in previous years. We introduced a few restrictions to reduce the chances of collaboration, but of course this exam was an open-book one. The large number of practical examples (50-100 per task type) and the comprehensive nature of the theoretical questions may have helped to produce realistic results, although they could not guarantee them.

While previously it was a mandatory part of the subject to create one or two AI-related software (solving a puzzle, recommending an action), this has deterred many people from taking the exam. This year, this limit was abolished, but the pass-limit was raised. Writing and defending a program could increase the result of the test. This year, almost all students passed the exam, even those who refrained from taking the exam in the previous year. Whilst it’s unfair to attribute this passing rate to our innovations, we do wish to keep a few of the tools (such as short explanatory videos) and apply them to other subjects as well.

Keywords:

Distance learning, e-learning, Moodle, artificial intelligence.