Abstract:

As Artificial Intelligence (AI) expands into more and more areas, there is a naturally growing need for more professionals in this area. At the Faculty of Informatics at the University of Debrecen AI has been taught for nearly 30 years. At BSc level we only have one semester to teach this subject. Thus, we need to provide the students with comprehensive, well-grounded knowledge and approach on which they can later build upon through self-education, so that even years down the road they can use these tools to further their knowledge.

In order for our students to acquire all the necessary competencies, we have developed a multi-part exam. In the first half, students have to demonstrate their knowledge through an online test. This test contains theoretical questions and exercises in almost equal proportions. Some of the latter are presented in the form of MCQ (Multiple Choice Questions) tests, simplifying the verification process. As a significant part of the curriculum is covered by various algorithms, we had to test the students’ ability to understand and apply them. Therefore, we have implemented these algorithms and run them for specific, randomly generated inputs. These gave us the questions and the corresponding correct solutions. We generated the incorrect answers manually and collected all the answers into a common file for each type of questions. Finally, using a short script, we put together randomly selected correct and incorrect answers for each question and stored them as individual XML files to match Moodle's own format. This enabled us to generate a wide variety of different tests from a limited number of questions. Some of these generated test have been added to a training server, whist the remainder have been added to the exam server. The former was available to the students as a revision tool for the exam, and some of our students also actively used these tests for both self-control and learning.

The second part of the exam was an oral presentation about the implementations of a logical puzzle and suggesting the next move in a two-person game. In both cases, there was a widely available corresponding AI framework that students could use, so they only needed to focus on formulating the specific tasks.

At the end of the exams, we measured how satisfied our students were with these practice and examination tools, which are quite unique in our faculty, as they are not naturally suited to the electronic study system and so were somewhat strenuous to implement. We asked eleven questions about the use of the tests, out of which eight were on the 5 point Likert scale. These questions were designed to fit the elements of the Technology Acceptance Model, so that we could measure the students' opinions on online tests, to see if they should be used in education, or we should rather choose different tools.

In this article we describe the structure of the course, the system of examinations, the questionnaire and the answers it received.

Keywords:

Artificial Intelligence, MCQ, Moodle.