From Modelling to Assessing Algorithmic Abstraction – the Missing Dimension

ion is the most fundamental idea of computer science (CS), manifested in every aspect of the discipline in multiple forms. In algorithmic problem solving, which lies at the heart of CS, it is manifested in two ways: 1) Modeling the problem and its components as computational entities, while focusing on essential core details of the problem and ignoring others. 2) Devising the solution by constantly employing algorithmic abstraction. Namely, the solution process involves moving up and down between abstraction levels, where black boxes are closed or opened, and details of the solution are ignored or considered, depending on the current phase of the problem-solving process. Models of this process can serve educators to effectively teach algorithmic problem solving, including algorithmic abstraction. Such models are often used also for assessment, for example, to examine the teaching of algorithmic problem solving or to evaluate students’ performance regarding algorithmic abstraction. In two different studies, we explored the teaching and learning of algorithmic abstraction in different contexts and at different age levels. In each of these studies an appropriate model of assessment was designed by means of combining top-down (deductive) and bottom-up (inductive) analyses. Thus, these models expressed grounded elements, namely, each of them included unique aspects of algorithmic abstraction that were revealed during the respective study. In this theoretical paper, we generalize these models and propose a new highly expressive two-dimensional model for evaluating algorithmic abstraction.