A STUDY OF THE EFFICIENCY OF AN ALTERNATIVE PROGRAMMING PARADIGM TO TEACH THE BASICS OF

Although computers and programs have now become essential in many sciences as analysis or measurement tools, many still find the learning of computer science extremely difficult. Kaäsboll reported in 2002 [1] that between 25 and 80 % of students in first-year programming courses either failed or gave up world-wide. A synthesis of different works in Cognitive Psychology and Didactics allowed us to summarise the intrinsic difficulties of programming in two main objectives: • to build a viable model of the structure of the task. • to build a viable model of the computer which will perform the task. The reason that makes learning programming that difficult is that students, contrary to physics for instance, have no effective starting model, and therefore no ground to understand the brutal error feedback they are faced to in their early experiences of programming. These internal difficulties are even more stressed by the use of “professional ” languages and environments as learning environments, because they: • are built on rather complex models of the computer. • give support to evaluation but few to conception, leading to “blank page ” syndromes. • make it very difficult to separate the construction of these two models. We plan to study the efficiency of an alternative programming paradigm, “programming by examples”. Programming by examples is neither static nor abstract, but enables the programmer to interact with a graphical model of the state of the task, through concrete examples. We detail the MELBA learning environment, built on a constructivist approach and example-based techniques to provide a visual and interactive support to construct the models of the task and of the computer, and analyse the efficiency of the system and of its underlying method through a study on 65 first year students in bio-informatics. 1