Abstract:

In a broad sense, questionnaires are among the most versatile educational resources. Although they have been widely used for student assessment with traditional methods, some technological advances such as educational e-learning platforms, software, or simulation tools, offer new opportunities to enhance the teaching and learning process.

Apart from their usefulness for competence assessment, questionnaires are a very powerful methodology for student self-assessment and continuous improvement, resulting in better academic results and personal satisfaction.

In the specific case of Data Science, questionnaires alleviate some of the problems encountered in teaching related subjects such as Statistics, Operations Research, and Artificial Intelligence.

These problems are mainly based on three facts:
- Traditional "problem sheets" are a limited set of exercises, which students can learn to solve, but cannot practice with new data to strengthen their knowledge.
- Often the exercises are generic or unrelated to their study domain, and students lose motivation for the course.
- The solved exercises do not always include detailed explanations and links to the underlying theory.

In this context, the professors' reflections of the Data Science Laboratory for Teaching Innovation at Rey Juan Carlos University on how to improve our teaching work to overcome the above obstacles have led us to develop a novel framework based on customized questionnaires called DSEXAMS. This framework is part of a teaching innovation project that promotes the generation of flexible and specific content (e.g., student profile, domain, or language), disseminating it, and sharing it with the educational community.

DSEXAMS has its foundations in the {exams} R package, which combines the power of R as a tool for data analysis and visualization with LaTeX, a powerful tool for editing mathematical texts, tables, and graphics. This promotes the creation of exercises of any level of difficulty both in the statement and in the content, far exceeding the limitations of educational platforms (e.g., Moodle and Sakai).

DSEXAMS creates questions to develop specialized exercise "templates". These templates are randomly generated including exercises, their solutions (using probability distributions to produce data or sampling for exercise options), and domain-specific feedback. Templates can be exported to other platforms and formats (e.g., pdf, Moodle, or Kahoot). This approach can be used to create Virtual Classroom quizzes to stimulate students who will never repeat exercises as in the hackneyed "problem sheets".

As a result of the project, DSEXAMS will expand R by creating a new package {dsexams} that will be published in public repositories.

Keywords:

Randomized questionnaires, Data Science teaching, Self-assessment, Autonomous learning, R Software.